AI protein modeling flags bacteria that may help produce urolithin A
Sequence screening and AlphaFold 3 modeling prioritized six bacterial enzyme candidates that may catalyze a key step in producing the gut-microbial metabolite urolithin A.
Summary
Urolithin A is a metabolite gut bacteria produce from dietary ellagitannins found in pomegranates, berries, and nuts. It has been associated with anti-aging and anti-inflammatory effects, but few bacteria and enzymes responsible for its production are known. Researchers searched sequence databases using three Enterocloster enzyme references, narrowing 570 initial candidates to 65 non-redundant three-subunit candidates. They then used AlphaFold 3 to model enzyme complexes and compare their binding pockets. Screening excluded candidates below 75% pocket residue similarity and previously reported Enterocloster homologs, prioritizing six candidates across three bacterial phyla. Dehalobacter restrictus showed the strongest non-Enterocloster pocket conservation; Aromatoleum sp. was also highlighted. These are computational candidates for experimental testing, not newly confirmed producers or demonstrated health interventions.
Detailed Summary
Urolithin A is a gut-microbial metabolite associated with anti-aging and anti-inflammatory effects. Gut bacteria produce it from dietary ellagitannins and ellagic acid; a key final step converts urolithin C to urolithin A through 9-OH dehydroxylation. According to the source, Enterocloster bolteae is the only confirmed bacterium performing this conversion for which the responsible enzyme's protein sequences are available. Other Enterocloster species have similar operons and can metabolize urolithin C in vitro, although their specific enzyme activity was not characterized in the cited work.
The researchers combined sequence searches with structural prediction using enzyme references from E. bolteae, E. asparagiformis, and E. citroniae. Each reference enzyme has a 787-residue molybdenum-binding subunit, a 288–289-residue FAD-binding subunit, and a 163-residue iron-sulfur-binding subunit. Multi-database searches identified 570 initial candidates. Filtering for sequence characteristics, completeness of the three-subunit complex, and redundancy yielded 65 non-redundant candidates.
A maximum-likelihood phylogeny placed the three Enterocloster references in a distinct lineage. The sequence alignment contained approximately 1.73% gap and ambiguous positions. Lacrimispora sp. 210928-DFI.3.58 clustered next to the reference clade, consistent with a previously published suggestion that it may be misclassified outside Enterocloster. Other nearby candidates included Dehalobacter restrictus, Aromatoleum sp., Anaerovorax odorimutans, Cloacibacillus evryensis, and Zhenpiania hominis.
For structural screening, the researchers modeled the candidate enzymes with AlphaFold 3 in the presence of cofactors and urolithin C, generating 25 models per species. They retained models in which the substrate's C9 hydroxyl was the hydroxyl closest to the molybdenum center. Pocket residues were defined within 4.4 Å of the ligand and/or molybdenum center, and only residues appearing in at least half of the geometry-passing models for each species entered subsequent analysis. Candidates below 75% pocket residue similarity were excluded, and previously reported Enterocloster homologs were removed to prioritize novel candidates. Pocket structural similarity was also assessed using RMSD.
The source reports six prioritized candidates spanning three bacterial phyla. Dehalobacter restrictus showed the strongest non-Enterocloster pocket conservation, and Aromatoleum sp. was also highlighted for predicted functional similarity to the references. The supplied excerpt does not provide the complete final shortlist or detailed residue-level conservation and RMSD results.
These findings provide a shortlist for experimental characterization. Conserved predicted binding pockets do not establish catalytic activity, production of urolithin A by the organisms, or health benefits. The supplied material reports no experimental validation of the newly predicted candidates.
Key Findings
- Sequence screening identified 570 initial candidates and yielded 65 non-redundant, complete three-subunit enzyme candidates after filtering.
- Pocket-based screening prioritized six candidates across three bacterial phyla; the workflow excluded candidates below 75% pocket residue similarity and previously reported Enterocloster homologs.
- AlphaFold 3 generated 25 models per species, retaining poses in which the substrate's C9 hydroxyl was the hydroxyl closest to the molybdenum center.
- Pocket residues were defined within 4.4 Å of the ligand and/or molybdenum center, with a requirement for occurrence in at least 50% of geometry-passing models.
- Dehalobacter restrictus showed the strongest non-Enterocloster pocket conservation; Aromatoleum sp. was also highlighted for predicted functional similarity.
- Lacrimispora sp. 210928-DFI.3.58 clustered adjacent to the Enterocloster references, consistent with an earlier suggestion of taxonomic misclassification.
- The findings identify computational candidates for experimental testing, not newly validated urolithin A-producing bacteria.
Methodology
The reported workflow combines sequence screening, phylogenetics, and computational structural analysis. Three Enterocloster references were screened against multiple databases using BLASTP and TBLASTN, with redundancy removal using CD-HIT and checks for complete three-subunit complexes. MAFFT alignments and IQ-TREE 2 phylogenies were used to assess relationships. AlphaFold 3 generated 25 models per species with cofactors and urolithin C. Models were filtered to retain poses with the C9 hydroxyl closest to molybdenum. Pocket residues were defined within 4.4 Å of the ligand and/or molybdenum center and retained for analysis if present in at least 50% of geometry-passing models. Screening excluded candidates below 75% pocket residue similarity and previously reported Enterocloster homologs. Pocket structural similarity was assessed by RMSD.
Study Limitations
The candidate prioritization rests on computational predictions, and the supplied material reports no experimental validation of the newly predicted enzymes or organisms. Conserved binding pockets and plausible substrate poses do not prove catalysis or urolithin A production. The reference set comprises three closely related Enterocloster enzymes. The supplied excerpt ends early in the Results, preventing verification of detailed pocket conservation values, substitutions, RMSD results, and the complete final shortlist. No conflicts-of-interest statement appears in the supplied text.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe:
