Novel bile acid biosynthetic pathways are enriched in the microbiome of centenarians

Centenarians have a decreased susceptibility to ageing-associated illnesses, chronic inflammation and infectious diseases1–3. Here we show that centenarians have a distinct gut microbiome that is enriched in microorganisms that are capable of generating unique secondary bile acids, including various isoforms of lithocholic acid (LCA): iso-, 3-oxo-, allo-, 3-oxoallo- and isoallolithocholic acid. Among these bile acids, the biosynthetic pathway for isoalloLCA had not been described previously. By screening 68 bacterial isolates from the faecal microbiota of a centenarian, we identified Odoribacteraceae strains as effective producers of isoalloLCA both in vitro and in vivo. Furthermore, we found that the enzymes 5α-reductase (5AR) and 3β-hydroxysteroid dehydrogenase (3β-HSDH) were responsible for the production of isoalloLCA. IsoalloLCA exerted potent antimicrobial effects against Gram-positive (but not Gram-negative) multidrug-resistant pathogens, including Clostridioides difficile and Enterococcus faecium. These findings suggest that the metabolism of specific bile acids may be involved in reducing the risk of infection with pathobionts, thereby potentially contributing to the maintenance of intestinal homeostasis. The microbiota of centenarians (aged 100 years and older) comprise gut microorganisms that are capable of generating unique secondary bile acids, including isoallolithocholic acid, a bile acid with potent antimicrobial effects against Gram-positive—but not Gram-negative—multidrug-resistant pathogens.

Centenarians have a decreased susceptibility to ageing-associated illnesses, chronic inflammation and infectious diseases [1][2][3] . Here we show that centenarians have a distinct gut microbiome that is enriched in microorganisms that are capable of generating unique secondary bile acids, including various isoforms of lithocholic acid (LCA): iso-, 3-oxo-, allo-, 3-oxoallo-and isoallolithocholic acid. Among these bile acids, the biosynthetic pathway for isoalloLCA had not been described previously. By screening 68 bacterial isolates from the faecal microbiota of a centenarian, we identified Odoribacteraceae strains as effective producers of isoalloLCA both in vitro and in vivo. Furthermore, we found that the enzymes 5α-reductase (5AR) and 3β-hydroxysteroid dehydrogenase (3β-HSDH) were responsible for the production of isoalloLCA. IsoalloLCA exerted potent antimicrobial effects against Gram-positive (but not Gram-negative) multidrug-resistant pathogens, including Clostridioides difficile and Enterococcus faecium. These findings suggest that the metabolism of specific bile acids may be involved in reducing the risk of infection with pathobionts, thereby potentially contributing to the maintenance of intestinal homeostasis.
The microbiome has long been recognized as a key player in determining the health status of ageing individuals through its role in controlling digestive functions, bone density, neuronal activity, immunity and resistance to pathogen infection [3][4][5][6][7] . Microbial data from older individuals often show increased interindividual variability and reduced diversity, and are thus being linked to immunosenescence, chronic systemic inflammation and frailty. By contrast, centenarians (individuals aged 100 years and older) are less susceptible to age-related chronic diseases and have survived several bouts of infectious diseases [1][2][3] . It has been postulated that there are centenarian-specific members of the gut microbiota that, rather than representing a mere consequence of ageing, could actively contribute to resistance against pathogenic infection and other environmental stressors [3][4][5][6][7] . In this study, we aimed to identify such beneficial bacteria in the gut microbiota of centenarians.

Microbiome signature of centenarians
We recruited a cohort consisting of three age groups: centenarian (average age, 107 years old; n = 160), older (85-89 years old; n = 112) and young (21-55 years old, n = 47) (all Japanese individuals). All centenarians were recruited as part of the Japan Semi-supercentenarian Study 1 , most of whom lived in nursing homes (85.0%) with the remainder living at home (9.4%) or in hospitals (5.6%) (Supplementary Table 1). Centenarians generally reported reduced activities of daily living and mini-mental state examination scores, along with reduced red-blood-cell counts and serum albumin (Extended Data Fig. 1a-c and Supplementary Table 1). Consistent with the paradigm that ageing is accompanied by chronic inflammation secondary to decreased barrier integrity and immunosenescence 4 , a subset of centenarians showed signs of low-grade inflammation as evidenced by elevated serum C-reactive protein and faecal lipocalin (Extended Data Fig. 1c, d). Nevertheless, the majority of centenarians were free of chronic diseases such as obesity, diabetes, hypertension and cancer, and the prevalence of these diseases was not significantly increased compared to the group of older individuals (Extended Data Fig. 1e, f and Supplementary Table 1). We collected faecal samples from the three groups of participants and characterized the microbiome (excluding three centenarians and one older participant undergoing antibiotic treatment). Principal coordinate analysis revealed significant differences in the composition of the microbiome between centenarians and both control groups (Extended Data Fig. 2a). At the phylum level, we observed a significant enrichment of Proteobacteria and Synergistetes, a moderate enrichment of Verrucomicrobia and a depletion of Actinobacteria in centenarians compared to both control groups (Extended Data Fig. 2b, c), which is in partial agreement with previous centenarian studies, including that of a Sardinian cohort 3 . Such expansions of Proteobacteria are a frequent finding in patients with inflammatory bowel disease (IBD); however, in contrast to the reduced microbial α-diversity commonly observed in patients with IBD, centenarians had, on average, a higher Shannon index compared to young control participants (Extended Data Fig. 2d).
Several taxa showed differential relative abundances in centenarians versus the control groups (Extended Data Fig. 2e-g), which we categorized into three signatures based on the trajectory with age: (1) the first signature included taxa, the abundance of which was increased or decreased with age; (2) the second signature included taxa, the abundance of which was similar in centenarians and young control participants, but distinct from the older participants; and (3) the third signature included centenarian-specific taxa, the abundance of which was significantly different between centenarians and both the older and young control groups, but not between these two control groups. Notably, Alistipes, Parabacteroides, Bacteroides, Clostridium and Methanobrevibacter species were included in the third signature and were specifically enriched in centenarians (Extended Data Fig. 2g). One of the most-enriched species in centenarians was Clostridium scindens, which is known to possess the relatively rare 7α-dehydroxylation capacity needed to convert primary into secondary bile acids 8,9 . By contrast, key butyrate producers such as Faecalibacterium prausnitzii and Eubacterium rectale were selectively depleted in centenarians (Extended Data Fig. 2g). Some of these observations are in agreement with the Sardinian study, in which centenarians exhibited a decreased relative abundance of F. prausnitzii and E. rectale and an increase in Methanobrevibacter smithii 3 .
We also analysed stool samples from the lineal descendants and siblings of centenarians by 16S ribosomal RNA (rRNA) sequencing (n = 22 relatives of 14 centenarians; 48-95 years old) (Extended Data Fig. 3). Some bacterial species-such as Alistipes putredinis and Odoribacter splanchnicus-were more abundant in centenarians and their family members compared to the other groups (Extended Data Fig. 3d). Enrichment of these taxa may be due to host genetics, lifestyle and diet.

Centenarians have a unique bile acid profile
We next assessed faecal metabolites. The levels of short-chain fatty acids (SCFAs) such as propionic and butyric acid were decreased, whereas branched SCFAs such as isobutyric and isovaleric acid, as well as ammonium, were elevated in centenarians (Extended Data Fig. 4a, b and Supplementary Table 2). These metabolic alterations may be attributable to the simultaneous depletion of SCFA producers and enrichment of protein-fermenting organisms, such as A. putredinis 10 (Extended Data Fig. 2g). The increase in amino-acid-using bacteria is probably a consequence of reduced upper intestinal proteolytic capacity. Moreover, faecal pH was significantly higher in centenarians than both groups of control participants (Extended Data Fig. 4c), which may be due in part to the lower SCFA concentrations and reduced gastric juice production that is characteristic of ageing.
Metagenomic analysis identified an increase in the relative abundance of bile-acid-inducible (bai) operon genes in centenarians, although this trend was not apparent in the Sardinian cohort 3 (Extended Data Fig. 2h, i). Therefore, we next focused on the distribution of bile acids in faeces. Although the total bile acid load was not significantly different between groups, centenarians showed a unique distribution with lower levels of primary bile acids and increased levels of chenodeoxycholic acid (CDCA) metabolites ( Fig. 1a- Table 3). In particular, the levels of isoLCA, 3-oxoLCA, alloLCA, 3-oxoalloLCA and isoalloLCA were significantly elevated in centenarians, whereas they were comparably low in the older and young control groups (Fig. 1a, d and Extended Data Fig. 5e). Furthermore, faecal pH was positively associated with the concentrations of these bile acids (Extended Data Fig. 4d), potentially implying that such an intestinal milieu may promote the expansion of certain bacterial species and/or the expression and activity of enzymes involved in the production of these bile acids.
To validate our pathway predictions and identify unique bile-acid-producing bacterial strains, we followed up on a supercentenarian (CE91, who is more than 110 years old) who showed high faecal levels of these bile acids (Fig. 1a). We cultured faecal samples from CE91 and isolated 68 unique strains, which roughly recapitulated the microbiota structure of participant CE91 ( Fig. 2a and Supplementary  Table 4). We then incubated individual isolates with CDCA, LCA or 3-oxo-Δ 4 -LCA as starting substrates. Whereas the incubation with CDCA did not result in the production of target bile acids in any of the cultures, incubation with LCA or 3-oxo-Δ 4 -LCA afforded 12 out of 68 strains the capacity to generate 3-oxoLCA and 8 to generate isoLCA, suggesting that these strains possess 3α-HSDH, 3β-HSDH and/or 5BR ( Fig. 2b and Extended Data Fig. 7). Notably, after incubation with 3-oxo-Δ 4 -LCA, a marked accumulation of isoalloLCA was observed in the cultures of Parabacteroides merdae strain number (St)3, Odoribacter laneus St19, Odoribacteraceae spp. St21-St24 and, to a lesser degree, Bacteroides dorei St6, St7 ( Fig. 2b and Extended Data Fig. 7c), suggesting that these strains contain both 5AR and 3β-HSDH. Additionally, Parabacteroides goldsteinii St1, St2, Bacteroides thetaiotaomicron St9, Bacteroides uniformis St10-St13, Alistipes finegoldii St15, St16, Alistipes onderdonkii St17, St18, and O. laneus St20 cultures all showed the substantial accumulation of 3-oxoalloLCA, but little to no isoalloLCA (Fig. 2b), which is Article probably due to the carriage of 5AR but lack or insufficient activity of 3β-HSDH in these culture conditions. Collectively, a total of 20 Bacteroidales strains were found to be capable of transforming 3-oxo-Δ 4 -LCA into 3-oxoalloLCA, 8 of which were able to generate isoalloLCA ( Fig. 2b and Extended Data Figs. 7c, 8a). AlloLCA was consistently below the detection limit, suggesting that our 5AR-carrying isolates either lack or have insufficient 3α-HSDH activity in the tested culture conditions. Of note, incubation at pH 9 (representative of the gut environment of centenarians) enhanced the accumulation of 3-oxoalloLCA, isoalloLCA, isoLCA and 3-oxoLCA compared to that at pH 7 ( Supplementary Fig. 1), suggesting that alkaline pH stress may promote the activity or expression of enzymes involved in the production of these bile acids.
We next tested whether our isolates could produce appreciable quantities of isoalloLCA in vivo. Germ-free C57BL/6N mice were monocolonized with one of 12 selected Bacteroidales isolates and fed a 3-oxo-Δ 4 -LCA-containing diet. Upon faecal bile acid profiling, only Odoribacteraceae strains were found to induce robust accumulation of isoalloLCA (Fig. 2f). Although P. merdae St3 efficiently transformed 3-oxo-Δ 4 -LCA to isoalloLCA in vitro, only marginal levels of isoalloLCA were detected in vivo. Mice monocolonized with P. merdae St3-but not P. merdae Δ5AR-exhibited significant faecal accumulation of 3-oxoalloLCA (Fig. 2f), demonstrating that P. merdae 5AR retains its activity in vivo, albeit weakly. B. uniformis, B. thetaiotaomicron and B. dorei strains were capable of producing 3-oxoalloLCA but were not able to effectively convert it into isoalloLCA. Overall, these in vivo data suggest that the activity of commensal 5AR and 3β-HSDH is context-dependent and that the intestinal environment (at least in mice) favours isoalloLCA production by Odoribacteraceae.
Article therefore incubated C. difficile 630 with various concentrations of bile acids and tracked bacterial growth over time in vitro (Extended Data Fig. 10a, b). Notably, isoalloLCA potently inhibited the growth of C. difficile 630. The minimal inhibitory concentration required to prevent ≥90% growth (MIC 90 ) was 2.0 μM, far below that of the other bile acids tested (  Table 5a). IsoalloLCA also potently inhibited growth of toxigenic C. difficile VPI10463 and vancomycin-resistant E. faecium (VRE) (Fig. 3a and Extended Data Fig. 10a-c). Electron microscopy revealed that isoalloLCA was bactericidal, inducing morphologic alterations such as collapse, swelling and multiple cross walls in C. difficile and VRE (Fig. 3b). These patterns of damage are reminiscent of those induced by β-lactam antibiotics. Co-culturing with Odoribacteraceae St21 in conjunction with 3-oxo-Δ 4 -LCA supplementation resulted in significant growth inhibition of C. difficile and VRE, similar to that observed with isoalloLCA treatment (Extended Data Fig. 10d). By contrast, growth inhibitory effects were not observed when co-culturing was performed with Clostridium innocuum St51 (an isoLCA producer) or P. distasonis St4 (an isoLCA and LCA producer).
We then examined the effect of isoalloLCA on other Gram-positive, as well as Gram-negative, pathogens. IsoalloLCA strongly inhibited growth of all Gram-positive pathogens tested with MIC 90 values ranging from 0.5 to 3 μM in Wilkins-Chalgren Anaerobe (WCA) medium and from 3 to 6.25 μM in brain heart infusion (BHI) medium. By contrast, all members of our Gram-negative pathogen panel were resistant to isoalloLCA (Fig. 3a, Extended Data Fig. 10a-c and Supplementary Table 5a). Taken together, these data show that isoalloLCA has strong bactericidal and/ or bacteriostatic effects specifically on Gram-positive pathogens. The median absolute concentration of isoalloLCA in faeces of centenarians was 19.5 μM (7.33 μg ml −1 ) (Extended Data Fig. 5f), which is sufficiently above the in vitro MIC 90 values, suggesting that it is present at physiologically relevant concentrations in vivo.
We next examined the effect of isoalloLCA on C. difficile infection in vivo. Specific-pathogen-free (SPF) C57BL/6N mice were infected with C. difficile 630 and given an 0.1% isoalloLCA-containing diet. IsoalloLCA administration resulted in elevated faecal and serum levels of isoalloLCA without causing systemic toxicity, and induced a significant reduction in C. difficile shedding (Extended Data Fig. 11a-c). We next colonized germ-free C57BL/6N mice with faecal microbiota from a patient with a C. difficile infection and, as above, subsequent isoalloLCA administration significantly reduced C. difficile shedding by these mice (Fig. 3c). Next, we examined whether colonization with Odoribacteraceae St21 could help to fend off C. difficile in vivo. To this end, we used Cyp2c70 −/− Cyp2a12 −/− double-knockout mice 21 , which mimic the human bile acid profile with high levels of LCA derivatives and lack muricholic acids (Extended Data Fig. 11d-g). SPF Cyp2c70 −/− Cyp2a12 −/− double-knockout mice were infected with C. difficile 630, followed by repeated oral gavage with Odoribacteraceae St21. Odoribacteraceae administration significantly increased faecal isoalloLCA levels and markedly reduced C. difficile shedding to below the limit of detection (Fig. 3d). By contrast, wild-type mice undergoing the same treatment regimen did not produce isoalloLCA nor showed inhibition of C. difficile shedding, despite a similar extent of Odoribacteraceae engraftment (Fig. 3d). Together, these results suggest that Odoribacteraceae and

Effects of isoalloLCA on the commensal gut microbiota
We next investigated whether isoalloLCA affects common members of the human microbiota. A total of 42 prevalent gut microbiota members were selected from our culture collection, and each was incubated with isoalloLCA. IsoalloLCA did not appreciably affect the growth of most Gram-negative commensals such as Bacteroides. By contrast, it substantially interfered with the growth of Gram-positive commensals (Extended Data Fig. 12a, b and Supplementary Table 5b). However, MIC 90 values for commensal strains were generally higher than those for pathogens, and electron microscopy revealed that commensal morphology (Clostridium sporogenes, Clostridium indolis and Clostridium HGF2) was preserved when incubated with 2.5 μM isoalloLCA (1.25× MIC 90 for C. difficile) (Extended Data Fig. 12c). Moreover, culturing commensal strains in peptone-and amino-acid-rich BHI medium conferred increased resistance to isoalloLCA compared to culturing in WCA medium (Extended Data Fig. 12b, d and Supplementary Table 5a, b), whereas pathogens generally remained sensitive to isoallo LCA irrespective of the medium used (Fig. 3a). These results indicate that although the concentration at which isoalloLCA exerts antimicrobial effects varies substantially depending on environmental conditions, Gram-positive pathogens consistently remain more sensitive to it than commensals.
To further evaluate the effects of isoalloLCA within the context of a complex gut flora, we incubated human faecal microbiota from healthy volunteers with bile acids. After incubation with isoalloLCA, we observed a pronounced reduction in Gram-positive species such as Faecalibacterium, Bifidobacterium and Streptococcus, along with a corresponding increase in Gram-negative species such as Bacteroides and Alistipes (Extended Data Fig. 13a, b), consistent with the enrichment of Bacteroides and Alistipes and depletion of Streptococcus species seen in the microbiomes of centenarians (Extended Data Figs. 2g, 3c-e, 13c-e). We additionally examined the effect of isoalloLCA on human microbiota in the context of C. difficile infection by incubating stool samples from three patients who had a C. difficile infection with bile acids. IsoalloLCA altered the microbial community structure and exerted the strongest inhibitory effect on C. difficile in all three samples (Extended Data Fig. 13f-h). These results are consistent with the notion that isoalloLCA can directly affect the structure of intestinal microbial communities and protect against potential pathogens.

Association of gut microbiome with bile acid profile
We next conducted a more in-depth metagenomic analysis of the participants. 5AR and 3β-HSDH gene clusters were identified in 35 species (all Bacteroidales). These clusters can be categorized into three groups: V1, V2 and V3 (Fig. 4a). Cluster V1 contains 5AR together with both 3β-HSDH-I and 3β-HSDH-II, and is found in the genomes of

Article
Odoribacter, whereas cluster V3 contains 5AR and 3β-HSDH-II and is found in the genomes of Alistipes. By contrast, cluster V2 comprises 5AR and 3β-HSDH-I and is carried by diverse Bacteroides species (Fig. 4b). We found that clusters V1 and V3 were significantly associated with the levels of faecal isoalloLCA, 3-oxoalloLCA, alloLCA, 3-oxoLCA and isoLCA in our cohort ( Fig. 4c and Extended Data Fig. 14a, b). Moreover, V1-or V3-carrying Odoribacter and Alistipes species were more abundant in centenarians than other groups ( Fig. 4b and Extended Data Fig. 14a). By contrast, the abundance of cluster V2 was negatively associated with the concentration of these bile acids ( Fig. 4c and Extended Data Fig. 14b). These results suggest that carriage of 5AR together with 3β-HSDH-II (rather than 3β-HSDH-I) is required to produce isoalloLCA and related bile acids. Moreover, bacterial species carrying cluster V2 may compete with those carrying V1 or V3 (Extended Data Fig. 14c). Notably, several species such as Anaerotruncus spp. and M. smithii-which do not encode 5AR-were strongly positively associated with isoalloLCA and related bile acid concentration (Extended Data Fig. 14d), suggesting that bile acid production by 5AR-and 3β-HSDH-II-carrying species may be modulated by other microbial community members. Given the context-dependent activity of 5AR gene clusters and the need to better account for interindividual microbial community variation, we stratified the microbiomes of our cohorts into three microbial community types (termed community types 1, 2 and 3) 22 (Extended Data Fig. 15a-c). The community structure of the young group was predominantly type 2, whereas the community structure of the centenarians was predominantly type 3 followed by type 1 (Fig. 4d). Notably, individuals with community types 1 and 3 tended to have microbiomes with higher abundances of species carrying V1 and V3 relative to those carrying V2 ( Fig. 4e and Extended Data Fig. 15d). Moreover, higher levels of 3-oxoLCA, isoLCA, alloLCA, 3-oxoalloLCA and isoalloLCA were observed in individuals with community types 1 and 3 than those with type 2, whereas this trend was reversed for CDCA (Fig. 4f). Notably, both the faecal bile acid profile and community type of centenarians were generally stable between longitudinal samples (Extended Data Fig. 15e, f). Overall, these results further support the notion that the underlying microbial community structure affects the intestinal bile acid profile, and that community types 1 and 3 are associated with Japanese centenarians and may promote an environment that facilitates the expansion and function of isoalloLCA-producing species.

Discussion
In the present study, we identified centenarian-specific gut microbiota signatures and defined bacterial species as well as genes and/or pathways that promote the generation of isoLCA, 3-oxoLCA, 3-oxoalloLCA and isoalloLCA. Several reports have demonstrated that bile acids contribute to the protection against enteropathogenic infection [18][19][20] . To our knowledge, isoalloLCA is one of the most potent antimicrobial agents that is selective against Gram-positive microorganisms, including multidrug-resistant pathogens, suggesting that it may contribute to the maintenance of intestinal homeostasis by enhancing colonization-resistance mechanisms. There are a number of limitations to this study. In particular, there are several confounding factors in our cohorts, making it difficult to discern the major driving forces behind the observed enrichment of isoalloLCA-producing organisms in centenarians. In addition, the causal relationship between the unique bile acids of centenarians and longevity needs to be validated with longitudinal surveys, additional participants and long-term analyses of animal models. Regardless, we may be able to exploit the unique bile-acid-metabolizing abilities of the bacterial strains identified in this study to rationally manipulate the bile acid pool and combat diseases caused by Gram-positives such as antibiotic-resistant C. difficile and VRE.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-021-03832-5.

Data reporting
No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

Human sample collection
Faecal samples and blood tests from Japanese young and older participatns, centenarians, and lineal relatives of centenarians were obtained following a protocol approved by the Institution Review Board of Keio University School of Medicine (code 20150075 for young healthy donors; 20160297 for older cohorts (as part of the Kawasaki Ageing and Wellbeing project); and 20022020 for centenarians and lineal relatives of centenarians (as part of The Japan Semi-supercentenarian Study 1 ). The microbiome dataset of Japanese patients with IBD was obtained from previous studies 23 . Informed consent was obtained from each donor before participation. All experiments adhered to the regulations mandated by these Review Boards. All study procedures were performed in compliance with the relevant ethical regulations. The Japan Semi-supercentenarian Study 1

Metagenomic sequencing and 16S rRNA gene pyrosequencing of human stool samples
Faecal samples were suspended in an equal volume of PBS containing 20% glycerol and 10 mM EDTA and stored at −80 °C until use. After thawing, 100 μl of faecal suspension was gently mixed and incubated in 800 μl TE10 (10 mM Tris-HCl, 10 mM EDTA) buffer containing RNase A (final concentration of 100 μg ml −1 ; Invitrogen) and lysozyme (final concentration of 15 mg ml −1 ; Sigma) for 1 h at 37 °C. Purified achromopeptidase (final concentration of 2,000 U ml −1 ; Wako) was added and further incubated for 30 min at 37 °C. SDS (final concentration of 1%) and proteinase K (final concentration of 1 mg ml −1 ; Roche) was further added to the mixture and incubated for 1 h at 55 °C. High-molecular-mass DNA was extracted with phenol:chloroform:isoamyl alcohol (25:24:1 at pH 7.9), precipitated with isopropanol (equal volume to the aqueous phase), washed with 1 ml 70% ethanol and gently resuspended in 30 μl of TE buffer.
We collected faecal samples from the three groups and characterized the microbiome by both 16S rRNA amplicon and whole-metagenome shotgun sequencing, excluding samples from participants undergoing antibiotic treatment (three centenarians and one older participant). The 16S rRNA sequencing was performed using MiSeq according to the Illumina protocol. PCR was performed using 27Fmod 5′-AGRGTTTGATYMTGGCTCAG-3′and 338R 5′-TGCTGCCTCCCGTAGGAGT-3′ to the V1-V2 region of the 16S rRNA gene. Amplicons generated from each sample (around 330 bp) were purified using AMPure XP magnetic beads (Beckman Coulter). DNA was quantified using a Quant-iT Picogreen dsDNA assay kit (Invitrogen) and Infinite M Plex plate reader (Tecan) and then stored at 4 °C. The pooled amplicon library was sequenced using a MiSeq Reagent Kit v2 (500 cycles) and Miseq sequencer (Illumina; 2 × 250-bp paired-end reads). After demultiplexing the 16S sequence reads based on the sample-specific index, primer sequences were trimmed by Cutadapt v.1.15. The trimmed reads were uploaded to the DADA2 R package v.1.18.0 to construct amplicon sequence variants (ASVs) using the fil-terAndTrim function with standard parameters (maxN = 0, truncQ = 2 and maxEE = 2). Possible chimeric reads were removed with the remove-BimeraDenovo function of the DADA2. The taxonomic assignment of each ASV was determined by similarity searching against the NCBI RefSeq database using the GLSEARCH program.
Metagenomic sequencing libraries were prepared from 2 ng of input DNA using the Nextera XT DNA Library Preparation kit (Illumina) according to the manufacturer's recommended protocol. Libraries were pooled by equal volume, and insert sizes and concentrations for each pooled library were determined using an Agilent Bioanalyzer DNA 1000 kit (Agilent Technologies). Sequencing was performed on an Illumina NovaSeq 6000 with 151-bp paired-end reads to yield around 10 million paired-end reads per sample. Data were analysed using the Broad Picard Pipeline, which includes demultiplexing and data aggregation (https://broadinstitute.github.io/picard). The quality control for the metagenomic data was conducted using Trim_Galore! to detect and remove sequencing adapters (minimum overlap of 5 bp) and KneadData v.0.7.2 to remove human DNA contamination and trim low-quality sequences (HEADCROP:15, SLIDINGWINDOW:1:20), retaining reads that were at least 50 bp long. Metagenomic reads were assembled individually for each sample into contigs using Mega-HIT 24 , followed by an open-reading-frame prediction with Prodigal 25 and retaining predicted genes that had both a start and a stop codon. A non-redundant gene catalogue was constructed by clustering predicted genes based on sequence similarity at 95% identity and 90% coverage of the shorter sequence using CD-HIT 26,27 . Reads were mapped to the gene catalogue with BWA requiring a unique, strong mapping with at least 95% sequence identity over the length of the read 28 , counted (count matrix) and normalized to transcripts per kilobase per million (TPM matrix) using in-house scripts. The count matrix served as an input for binning genes into metagenomic species pan-genomes (core and accessory genes) using MSPminer with default settings 29 . We represented the abundance of each metagenomic species (MSP) in a sample as a median TPM for the 30 top representative core genes reported by MSPminer. Assembled genes were annotated at species, genus and phylum levels with NCBI RefSeq (version May 2018) as previously described 30 . To annotate phylogenetically MSPs that had no match to any species from NCBI RefSeq, we used PhyloPhlan with default settings 31 . The α-diversities were calculated using the Shannon index and the β-diversity was calculated using Bray-Curtis dissimilarity based on relative abundances at the species levels (Vegan package in R). The non-redundant gene catalogue was queried using USEARCH ublast 30 with proteins in the bai operon of C. scindens 8 or proteins in the bacterial isolates reported here as 5AR, 5BR, 3β-HSDH-I or 3β-HSDH-II, to identify and annotate homologous proteins with at least 40% identity and 80% coverage to the query sequence. An identical processing pipeline has been applied to a dataset describing the gut microbiome in Sardinian centenarians 3 .
In the subsequent analysis, we only used samples with at least 4 million reads after the quality-control step. Additionally, we discarded samples that were collected while the participant was undergoing any antibiotic treatment. To test the differential abundance of species or phyla and differences in the Shannon diversity index, we used linear random-effects modelling (centenarians versus young or older controls) or fixed-effects modelling (older versus young controls), as implemented in the lmer and lm functions in R. Furthermore, for the analysis of species differential abundance, we restricted the analysis to MSPs present in at least 10% of samples, zeros were replaced by half of the smallest non-zero measurement on a per-feature basis and log 10 transformation was applied to the relative abundances for normality. Linear modelling included fixed-effect covariates: sex (male or female) and cohort information (centenarian, older or young); random effects included participant information to account for more than one sample among a few centenarians. The permutational multivariate analysis of variance (PERMANOVA) analysis as implemented in the adonis function in the R package vegan was applied to the Bray-Curtis dissimilarity to identify the correlation between age group (centenarian, older or young) and sex information and the composition of the gut microbiome as a whole. To determine the clustering of samples based on their metagenomes into community types, we applied the Dirichlet multinomial mixtures algorithm 22 to MSPs abundance matrix (median reads per killobase for 30 top representative core genes reported by MSPminer). The appropriateness of community types was determined based on the lowest Laplace approximation score.

Faecal bile acid quantification using LC-MS/MS
Accurately weighed freeze-crushed faecal samples were resuspended in 20 times the volume of water (for example, 12.5 mg of faece in 250 μl water). Then, 250 μl of faecal suspension was homogenized in 747 μl of 0.27 N NaOH by ultrasonication for 1 h in a screw-cap glass vial containing 3 μl of deuterium-labelled internal standards (d 4 -CA, d 4 -GCDCA, d 4 -TCDCA, d 4 -CDCA-3S and d 4 -LCA; 50 μM each). After incubation for 1 h at room temperature, the pH was adjusted to 8.0 using 8 N HCl, mixed with 110 μl of 0.5 M EDTA/0.5 M Tris-HCl. The solution was centrifuged at 15,000 rpm and the supernatant was transferred onto a solid-phase extraction cartridge (Agilent Bond Elut C18, 100 mg per 3 ml, preconditioned with 1 ml of methanol and 3 ml of water, three times). The cartridge was washed with 1 ml of water and captured bile acids were eluted with 300 μl 90% ethanol. Then, 2 μl was used for LC/ electrospray ionization (ESI)-MS/MS injected to a Triple Quad 6500+ tandem mass spectrometer, equipped with an ESI probe and Exion LC AD ultra-high pressure liquid chromatography system (SCIEX). A separation column, InertSustain C18 (150 mm × 2.1 mm inner dimension, 3 μm particle size; GL Sciences), was used at 40 °C. A mixture of 10 mM ammonium acetate and acetonitrile was used as the eluent and the separation was carried out by linear gradient elution at a flow rate of 0.2 ml min −1 . The mobile phase composition was gradually changed as follows: ammonium acetate:acetonitrile (86:14, v/v) for 0.5 min; (78:22, v/v) for 0.5-5 min; (72:28, v/v) for 5-28 min; (46:54, v/v) for 28-55 min; (2:98, v/v) for 55-66 min; and (2:98, v/v) for 4 min. The total run time was 70 min. To operate the LC/ESI-MS/MS, the following MS parameters were used for the positive-ion multiple reaction monitoring (MRM) mode: ion spray voltage, 5,500 V; interface temperature, 500 °C; curtain gas, 30 psi; collision gas (nitrogen), 9 psi; nebulizing gas, 80 psi; and heater gas, 80 psi. For the negative-ion MRM mode: ion spray voltage, −4,500 V; interface temperature, 500 °C; curtain gas, 30 psi; collision gas (nitrogen), 9 psi; nebulizing gas, 80 psi; and heater gas, 80 psi. Samples were obtained using Analyst software v.1.71 and analysed using SCIEX OS-MQ software v.1.7.0.36606. The detailed set-up for each of the 43 bile acid compounds is listed in Supplementary Table 6.

Faecal SCFA, pH and ammonia measurements
The faecal SCFA concentration was determined by gas chromatography-mass spectrometry (GC-MS) (Shimadzu QP2020 system with a flame ionization detector), equipped with PAL RTC autosampler (CTC Analytics). Using the PAL RTC autosampler, 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4methylmorpholinium and n-octylamine (10 μl of each reagent at a concentration of 80 μM) were added to each faecal sample and reacted for 9 h before injection and analysis by GC-MS. Fused silica capillary columns (30 m × 0.25 mm, coated with 0.25-μm film thickness) were used with helium as the carrier gas. The injection port temperature was set to 250 °C. The initial oven temperature was held at 60 °C for 2 min and then ramped to 330 °C at a rate of 15 °C per min. MS parameters were set to: ion source temperature at 200 °C, interface temperature at 280 °C and loop time of 0.3 s. For the GC-MS measurement, 50 μl of faecal samples with a concentration of 0.5 μg μl −1 and 20 μg μl −1 prepared in ethanol were mixed with 10 μl of acetic acid-d 4 (80 μM). Samples were analysed and quantified using LabSolutions Insight GC-MS software (Shimadzu).
Faecal pH was measured from the supernatant from 0.1 mg μl −1 of faecal suspension in distilled water using a pH meter (Horiba). From the same faecal suspension, the faecal ammonia level was quantified using an enzymatic ammonia ELISA assay kit (Abcam) according to the manufacturer's protocol.

Isolation of bacterial strains from a centenarian
A faecal sample from a supercentenarian (CE91, Japanese woman, age over 110 years) was suspended in an equal volume (w/v) of PBS containing 20% glycerol, snap-frozen in liquid nitrogen, and stored at −80 °C until use. Then, 200 μl of thawed faecal suspension was serially diluted with PBS and 100 μl was seeded onto nonselective (Brucella agar with haemin, vitamin K1, lysed rabbit blood and defibrinated sheep blood (BHK-RS), Kyokuto) and selective (for Gram-negative bacteria, paromomycin and vancomycin supplemented BHK agar, Kyokuyo; for Clostridial bacteria, oxoid-reinforced Clostridial agar, ThermoFisher) plates and grown inside an anaerobic chamber (Coy Laboratory Products) under anaerobic conditions (80% N 2 , 10% H 2 and 10% CO 2 ) at 37 °C. Individual colonies emerged after 72 h and up to 10 days of incubation were picked. Isolated strains were identified by PCR amplification of the 16S rRNA gene region with universal primers (27Fmod 5′-AGRGTTTGATYMTGGCTCAG-3′; 1492R 5′-GGYTACCTTGTTACGACTT-3′) for Sanger sequencing and using the NCBI genome database. Individual isolates in the culture collection were given species name with >98.0% of 16S rRNA sequence homology, family name with >94.5% similarity and order name with >86.5% similarity. Bacterial isolates were cryo-preserved in 20% glycerol in optimal culture broth at −80 °C.
For sample preparation, 100 μl of the culture supernatant was transferred into a screw-cap glass vial containing 10 μl of deuterium-labelled internal standards (d 4 -CA, d 4 -CDCA and d 4 -LCA; 1 nmol ml −1 each). Then, 400 μl water was added and the sample was sonicated for 10 min and applied onto the solid-phase extraction cartridge (Agilent Bond Elut C18, 100 mg per 1 ml; preconditioned with 1 ml methanol and 3 ml water). The cartridge was washed with 1 ml water and captured bile acids were eluted with 1 ml 90% ethanol. After solvent evaporation, the remaining residue was dissolved in 100 μl 50% ethanol, of which 2 μl of the solution was injected to LC/ESI-MS/MS spectrometer (LC-MS8040 tandem mass spectrometer, equipped with an ESI probe and Nexera X2 ultra-high-pressure liquid chromatography system; Shimadzu). A separation column, InertSustain C18 (150 mm × 2.1 mm inner diameter, 2 μm particle size; GL Sciences), was used at 40 °C. Mixture A (10 mM ammonium acetate, 0.01% formic acid and 20% acetonitrile) and mixture B (30% acetonitrile and 70% methanol) were used as the eluent, and the separation was carried out by linear gradient elution at a flow rate of 0.2 ml min −1 . The mobile phase composition was gradually changed as follows:  Table 7.

Bacterial whole-genome sequencing
The extracted genomic DNA of 68 isolated strains was sheared to yield DNA fragments. The genome sequences were determined by the whole-genome shotgun strategy using PacBio Sequel and Illumina MiSeq sequencers. The library of the Illumina Miseq 2 × 300-bp paired-end sequencing was prepared using the TruSeq DNA PCR-Free kit (target length = 550 bp) and all of the MiSeq reads were trimmed and filtered with a more than 20 quality value using FASTX-toolkit (http:// hannonlab.cshl.edu/fastx_toolkit/). The library of the PacBio Sequel sequencing was prepared using the SMRTbell template prep kit 2.0 (target length = 10-15 kb) without DNA shearing. After the removal of the internal control and adaptor trimming by Sequel, error correction of the trimmed reads was performed using Canu v.1.8 with additional options (corOutCoverage = 10,000, corMinCoverage = 0, corMhap-Sensitivity = high). De novo hybrid assembly of the filter-passed MiSeq reads and the corrected Sequel reads were performed using Unicycler v.0.4.8, which contained checks for overlapping and circularization to generate circular contigs. The gene prediction and annotation of the generated contigs were performed using the Rapid Annotations based on Subsystem Technology (RAST) server 33 and Prokka software tool 34 . Default parameters were used unless otherwise specified.

Mutant generation
The deletion mutants (Δ5AR, Δ5BR and Δ3β-HSDH) of P. merdae St3 were generated by conjugation-mediated plasmid transfection and selection of double-crossover resolvents with a rhamnose-inducible ssBfe1 cassette 35 . Approximately 2-kb sequences flanking the coding region were amplified by PCR (PCR primers used in this study are listed in Supplementary Table 8) and assembled into the PstI and SalI sites of the suicide vector pLGB30 using the HiFi DNA Assembly (NEB) as per the manufacturer's protocol. Then, 1 μl aliquots of each reaction were transformed into electro-competent Escherichia coli MFDpir 36 . Transformants were conjugated with P. merdae St3 as follows. The donor (E. coli MFDpir) and recipient (P. merdae St3) strains were cultured in LB and BHI media, respectively, to an optical density at 600 nm (OD 600 ) of 0.5 and mixed at a ratio of 1:1. The mixture was dropped onto a BHI agar plate and incubated anaerobically at 37 °C for 16 h. Transconjugants were selected on BHI agar plates containing tetracycline (6 μg ml −1 ). Subsequently, to select for loss of the plasmid from the genome by a second crossover, transconjugants were plated on M9 agar supplemented with 0.25% (w/v) glucose, 50 mg l −1 l-cysteine, 5 mg l −1 haemin, 2.5 μg l −1 vitamin K1, 2 mg l −1 FeSO 4 ·7H 2 O, 5 μg l −1 vitamin B12 and 10 mM rhamnose. Successful deletions were confirmed by PCR and Sanger sequencing. From fresh colonies grown on BHK blood agar plates (Kyokuto), a primary suspension adjusted to an OD 600 of 0.63 was prepared in WCA or BHI medium (the composition of media is presented in Supplementary Table 9). Subsequently, the secondary suspension was prepared by diluting 100 μl of the primary suspension into a total of 2.4 ml of medium. Then, 10 μl of the secondary suspension was inoculated to a total of 200 μl of medium containing varying concentrations (3.175, 6.25, 12.5, 25 or 50 μM) of bile acids; DCA, LCA, 3-oxoLCA, 3-oxoalloLCA, isoLCA, alloLCA or isoalloLCA. The growth of bacteria was monitored every 0.5-1 h by OD 600 measurement using a microplate reader (Sunrise Thermo, Tecan) set at 37 °C with 60 s shaking before each time point and a PLATEmanager v.5/S software for the data collection. To determine the MIC, 10 μl of the secondary suspension was inoculated into a total of 200 μl of medium containing 0.25-50 μM isoalloLCA.

Electron microscopy
Bacterial cultures incubated with or without isoalloLCA were collected after incubation for 5 h for electron microscopy samples. For scanning electron microscopy, 10-30 μL of culture was spotted on a Nano Percolator membrane ( JEOL) and fixed in freshly prepared 2.5% glutaraldehyde solution. After overnight fixation at 4 °C, samples were washed in 0.1 M phosphate buffer (pH 7.4, Muto Pure Chemicals), fixed with 1.0% osmium tetroxide (TAAB Laboratories) for 2 h at 4 °C, and treated with a series of increasing concentrations of ethanol. Samples were dried up with a critical point dryer (CPD300, Leica Biosystems) and coated with about 2 nm thickness of osmium using a conductive osmium coater (Neoc-ST, Meiwafosis). Scanning electron microscopy images were acquired using the SU6600 (Hitachi High Tech) at an electron voltage of 5 keV.
For transmission electron microscopy, microbial pellets were prepared by centrifugation (13,000 rpm, 2 min) from 25 ml bacterial cultures. Pellets were fixed with 2.5% glutaraldehyde solution overnight at 4 °C. After washing with 0.1 M phosphate buffer, samples were fixed with 1.0% osmium tetroxide for 2 h at 4 °C, washed in distilled water and embedded into low-gelling-temperature Type VII-A agarose (Sigma-Aldrich). Samples were dehydrated by a series of increasing concentrations of ethanol to absolute ethanol, soaked with acetone (Sigma-Aldrich), with n-butyl glycidyl ether (Okenshoji), graded concentration of Epoxy resin with n-butyl glycidyl ether and also with 100% Epoxy resin (100 g Epon was composed of 27.0 g MNA, 51.3 g EPOK-812, 21.9 g DDSA and 1.1 ml DMP-30, all from Okenshoji) for 48 h at 4 °C. Polymerization of pure Epoxy resin was completed for 72 h at 60 °C. The ultrathin sections (70 nm) were prepared on copper grids (Veco Specimen Grids, Nisshin-EM) with an ultramicrotome (Leica UC7, Leica Biosystems) and stained with uranyl acetate and lead citrate for 10 min each. Transmission electron microscopy images were obtained using the JEM-1400plus ( JEOL) at an electron voltage of 80-100 keV.

Animal experiments
Germ-free female C57BL/6N mice were purchased from CLEA Japan or bred and maintained within the gnotobiotic facilities of JSR-Keio University Medical and Chemical Innovation Center. All animal experiments were approved by the Keio University Institutional Animal Care and Use Committee. All animals were maintained on a 12-h light-dark cycle and received gamma-irradiated pellet food (50 kGy radiated CL2, CLEA Japan). For quantification of microbial bile acid metabolism, an overnight bacterial culture (200 μl) was orally administered to 8-week-old germ-free female mice and the mice were switched from pellet food to powder diet (50 kGy radiated AIN-93G with bacteria-free mineral acid casein; Research Diets) with or without 0.1% 3-oxo-Δ 4 -LCA. Faeces samples were collected during the experiment to quantify the amount of bile acids using a SCIEX Triple Quad 6500+ system. The same method for human faecal sample preparation was applied to mouse faecal samples (diluted 20 times in 1× PBS).
For C. difficile infection, SPF C57BL/6N mice and Cyp2c70 −/− Cyp2a12 −/− mice (C57BL/6J background) 21 were made susceptible to colonization by administration of cefoperazone (0.4-5 g l −1 ) in drinking water for 4-5 days. After 2 days of pause from the antibiotic, mice were challenged orally with 6,000 C. difficile 630 spores. Infected mice were given an 0.1% isoalloLCA-containing diet or a daily administration of Odoribacteraceae St21 (200 μl of overnight bacterial culture) after 1 day of infection. Germ-free C57BL/6N mice colonized with the faecal microbiota from a patient with a C. difficile infection were obtained by administration of 200 mg μl −1 of faecal slurry followed by an 0.1% isoalloLCA-containing diet from the next day. The abundance of C. difficile was quantified by CFU after anaerobic culture of diluted faecal samples on BD BBL S-CCFA (Becton Dickinson)-selective agar plates at 37 °C for 24 h. Odoribacteraceae St21 was quantified by qPCR from DNA extracted from faeces (5′-ACGTGGACCATCAGTGAACT-3′, 5′-AGTTCTCCAAGTCCCTCTGC-3′).

Chemical synthesis of 3-oxo-Δ 4 -LCA and isoalloLCA
3-oxo-Δ 4 -LCA was prepared from 5 g of LCA through the following synthesis reactions. First, 35 ml of methanol and 1 l of HCl-infused methanol were added to LCA in a 300-ml round-bottom flask. The suspension was then stirred at 100 °C for 1 h for methyl esterification of the carboxylic arm of LCA. After cooling down the reaction mixture, the methanol was evaporated. The slurry was dissolved in 300 ml of chloroform and transferred to a separatory funnel containing 50 ml water. The layers were vigorously mixed and separated, and the aqueous layer containing HCl was discarded (3 × 50 ml). The combined organic layers were washed with saturated aqueous NaCl, dehydrated over Na 2 SO 4 for 15 min and filtered, and the chloroform was evaporated. To generate methylated 3-oxoLCA, crystals were dissolved in 40 ml of acetone and then reacted dropwise with 3.5 ml of Jone's reagent (CrO 3 in aqueous H 2 SO 4 ) while cooling with ice water. Then, 2 ml methanol was added to reduce toxic Cr 6+ to Cr 3+ . The reaction mixture was slowly poured into 600 ml of stirred ice water to crystalize the synthesized compound, then rinsed and collected by filtration. Crystals were dissolved in 300 ml of ethyl acetate and transferred to a separatory funnel with 10 ml of water to clean up H 2 SO 4 and residual chromium. The organic layers were collected, washed and dehydrated as described above, followed by evaporation of ethyl acetate. To further purify the methylated 3-oxoLCA, the reaction product was dissolved in 1 l of toluene and 100 ml of ethyl acetate for fractionation by silica-based liquid chromatography. Upon verification of fractions containing the target compound by thin-layer chromatography, fractions were combined and the eluent was evaporated.
To brominate the C4-carbon atom on steroid ring A, 30 ml of acetic acid was added to the methylated 3-oxoLCA and dropwise reacted with the same molar amount of bromine solution. The dark-blue mixture was slowly poured into 600 ml of stirred ice water to crystalize the brominated compounds (yellowish crystalline solid), after which it was rinsed and collected by filtration. Crystals were dissolved in 100 ml of chloroform and transferred to a separatory funnel with 10 ml of water for liquid-liquid extraction, followed by evaporation of the chloroform. The compound with bromide at the C4 carbon was purified by silica-based liquid chromatography (toluene:ethyl acetate, 49:1). Collected fractions were combined and the eluent was evaporated. For the debromination reaction to generate the C4-C5 double bond, 10 ml of N,N-dimethylformamide was added to the crystals, followed by 3.5× the amount of LiCO 3 and LiBr. The mixture was stirred and heated at 100 °C overnight. Methylated 3-oxo-Δ 4 -LCA was extracted by adding 50 ml of chloroform and filtering out the precipitates. The organic solution was cleaned up by liquid-liquid extraction, followed by the evaporation of the N,N-dimethylformamide and chloroform. Methylated 3-oxo-Δ 4 -LCA was further purified by silica-based liquid chromatography (toluene:ethyl acetate, 10:1).
Methylated 3-oxo-Δ 4 -LCA was also used for isoalloLCA synthesis. In brief, 3.6 g methylated 3-oxo-Δ 4 -LCA was dissolved in 5 ml dehydrated pyridine, mixed with 1.9 g hexamethyldisilazane and 3.6 g bromotrimethylsilane for 40 min while cooling. The reaction solution was cleaned up by liquid-liquid extraction followed by solvent evaporation. The crude product was dissolved in 100 ml of isopropanol and then mixed with 780 mg sodium borohydride (NaBH 4 ) for 4 h. The reaction was quenched with acetic acid and evaporated, extracted with ethyl acetate, and cleaned up with liquid-liquid extraction. The reaction product (methylated Δ5-3β-ol) was purified by silica-based liquid chromatography (benzene:ethyl acetate, 5:1). To generate methylated isoalloLCA by hydorgenation, methylated Δ5-3β-ol was dissolved in 20 ml ethyl acetate, followed by carefully adding 300 mg of 10% palladium-activated carbon and hydrogen gas. After 2 h of reaction, the mixture was filtered and methylated isoalloLCA was purified by silica-based liquid chromatography (benzene:ethyl acetate, 4:1). Finally, purified methylated 3-oxo-Δ 4 -LCA and methylated isoalloLCA were mixed with the same molar amount of KOH (2 M solution in 90% methanol) at 50 °C for 1 h, and then with 2 M HCl to neutralize the solution overnight to remove the methyl group for the synthesis of 3-oxo-Δ 4 -LCA and isoalloLCA. Crystals were dissolved in chloroform for the final clean-up by liquid-liquid extraction, followed by evaporation of chloroform. The purity of 3-oxo-Δ 4 -LCA and isoalloLCA was determined by thin-layer chromatography and LC-MS/MS. 3-Oxo-Δ 4 -LCA and isoalloLCA were also generated by Sundia MediTech.

Statistical analysis
A pairwise Wilcoxon rank-sum test (nonparametric) was used to evaluate the differences in the relative abundance of bai operon homologues and meta-16S rRNA analysis between different age groups. A Spearman's rank correlation was used to evaluate trends between the relative abundance of Bacteroidales species encoding 5AR, 5BR, 3β-HSDH-I or 3β-HSDH-II genes and the abundance of the secondary bile acids in stool samples. Overall nominal P values were adjusted for multiple testing using Benjamini-Hochberg correction and associations with FDR-adjusted P < 0.05 (unless stated differently) were considered to be significant. Statistical analyses were performed using GraphPad Prism software (GraphPad Software) and RStudio. One-way ANOVA with Tukey's test (parametric) and Kruskal-Wallis with Dunn's test (nonparametric) were used for multiple comparisons. A Mann-Whitney U-test (two-tailed) with Welch's correction (nonparametric) was used for all comparisons between two groups in the in vivo bile acid metabolism, co-culture pathogen inhibition, in vivo C. difficile inhibition experiments.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Shotgun sequencing data are deposited in NCBI under BioProject PRJNA675598. Genome sequences of the 68 strains isolated from a centenarian and 16S rRNA amplicon sequence data are deposited in the DNA Data Bank of Japan under BioProject PRJDB11902 and PRJDB11894, respectively. LC-MS/MS data are deposited in Metabolomics Workbench (https://www.metabolomicsworkbench.org/) under project ID PR001168 with study ID ST001851 for human faeces data and study ID ST001852 for in vitro data. Source data are provided with this paper.

Code availability
Code for all of the analyses is available on GitHub (https://gitlab.com/ xavier-lab-computation/public/centenarianmicrobiome).       The microbiota composition of centenarians was significantly different from that of both control groups (PERMANOVA false-discovery rate (FDR)-adjusted P < 0.05). b, c, Relative abundance across phyla. d, Shannon diversity index (*P < 0.05, linear model). e-g, Changes in the relative abundance (RelAb) of gut metagenome species (MSPs) between centenarian, older, and young participants grouped according to the following signatures of differential abundance: 1) ageing signature, 2) rejuvenation signature, and 3) centenarian signature. Each signature is accompanied by models depicting microbial relative abundance patterns in centenarian (C), older (O), and young (Y) groups that would fall into the given signature. Colour scale represents the coefficient from the linear model and indicates enrichment (red) or depletion (blue) of a species in the respective comparisons: centenarian compared to older, centenarian compared to young, and older compared to young; in each case, the latter group is used as a reference in the model. Differentially abundant species that are significant at FDR P < 0.05 are indicated with asterisks. The first signature ('ageing signature') included taxa whose abundance was increased or decreased with age (e). For example, Eubacterium siraeum and undefined Firmicutes species (msp_161, 213) were most abundant in centenarians, followed by the older and then the young controls, whereas Blautia wexlerae displayed the opposite trend, being most abundant in young controls, followed by the older participants, and finally the centenarians. These findings are in alignment with previous studies that suggest the relative abundances of these taxa reflect adaptation to ageing, and may be related to physical activity and diet 3,40 . The second signature ('rejuvenation signature') included taxa whose abundance was similar in centenarians and young controls, but distinct from the older participants (f). These species might reflect the maintenance of youth or possess reverse-ageing effects. Notably, R. gnavus and E. lenta were part of this signature, as they were comparably abundant in both centenarians and young controls. Notably, these species have been implicated in bile acid metabolism, particularly the biosynthesis of iso-bile acids 11 . The third signature ('centenarian signature') included centenarian-specific taxa whose abundance was significantly different between centenarians and both the older and young control groups, but not between these two control groups (g    In pilot studies, we found that 94 of 137 examined bile acids were minor components of centenarians' faeces (see Supplementary Table 3). We thus selected the remaining 43 bile acid compounds for follow-up quantitative analysis (see also Fig. 1a). a, Multi-dimensional scaling plot using Spearman's correlation highlights differences among the four groups' bile acid profiles. Each circle represents an individual participant from the indicated age group. P Responsible enzymes are indicated within boxes. The glycine or taurine conjugated primary bile acids are deconjugated (not depicted) and biotransformed into a variety of secondary bile acids by the gut microbiota. The predominant biotransformation is 7α-dehydroxylation of CDCA by bai operon genes, thereby converting it into lithocholic acid (LCA). In addition, bile acids can undergo oxidation and epimerization to generate oxo-(keto-), iso-(3β-hydroxy-), allo-(5α-H-), as well as cis-(indicated in a blue box) and trans-forms (indicated in a pink box). Brackets indicate predicted pathways to allo-form LCA production. Chemical structures are simplified by depicting only A and B steroid rings. b, 3-Oxo-Δ 4 -LCA (also termed 3-oxo-4,5-dehydro-LCA) and 3-oxoalloLCA are structurally similar to testosterone and 5α-dihydrotestosterone (DHT), respectively. Both DHT and 3-oxoalloLCA have A and B steroid rings in a planar (trans) conformation (indicated in a pink box). We predicted that alloLCA and isoalloLCA might be generated from 3-oxo-Δ 4 -LCA by the sequential action of a 5α-reductase (5AR) homologue and 3α-HSDH (for alloLCA) or 3β-HSDH (for isoalloLCA), through a 3-oxoalloLCA intermediate (see a), analogous to the 5AR-mediated conversion of testosterone into DHT by hydrogenating across the C4-C5 double bond, thereby forcing the A and B steroid rings into a planar conformation. c, Biosynthetic pathway of DCA and related bile acids by the gut microbiota based on ref. 11 . 3-OxoDCA can be generated from 3-oxo-Δ 4 -DCA by hydrogenation across the C4-C5 double bond such that the C5 hydrogen is in the β position. This reaction is mediated by a 3-oxo-5β-steroid 4-dehydrogenase (also termed 5β-reductase, 5BR) encoded by the BaiCD gene. We predicted that LCA and isoLCA might be generated from 3-oxo-    Extended Data Fig. 7 | Identification of bile acid-metabolizing bacterial strains isolated from the microbiota of a centenarian. In vitro bile acid metabolism by 68 CE91-derived isolates using 50 μM of CDCA (a), LCA (b), and 3-oxo-Δ 4 -LCA (c) as starting substrates in pH 9-adjusted media. Data was obtained by LC-MS/MS analysis of 48 h culture supernatants. A list of the 68 isolated strains is shown in the left panel. The closest species to each isolate was identified based on 16S rRNA sequence similarity to the National Center for Biotechnology Information Reference Sequence (NCBI RefSeq) database. Graphs with red backgrounds indicate trans-bile acids and blue backgrounds indicate cis-bile acids. Bile acid profiles after culturing in pH 7-adjusted media are shown in Supplementary Fig. 1. a, Incubation with CDCA did not result in production of target bile acids in any of the cultures, though C. scindens strains 59-60 (St59-60) and C. hylemonae St63 were able to produce LCA, albeit at low levels. b, When cultured with LCA, Gordonibacter pamelaeae St32 and E. lenta

Article
Extended Data Fig. 15 | Stratification into microbial community types with respective microbiome characteristics. a, Principal coordinate analysis of three microbial community types (1, 2, and 3) stratified using Dirichlet Multinomial Mixtures from the gut microbiomes of centenarian [CE, n = 176 (153 individuals)], older (n = 110), and young (n = 44) individuals. b, c, Relative abundances across phyla (b) and top differentially abundant species from each phylum (c) from the assembled gut metagenomes in each community type. Community type 1 is characterized by a high relative abundance of Firmicutes (for example, Oscillibacter spp.) and Proteobacteria (for example, Desulfovibrio spp.), whereas community type 2 exhibits relative enrichment of Actinobacteria and depletion of Proteobacteria. Community type 3 is structurally similar to type 1, but exhibits a higher abundance of Verrucomicrobia (for example, Akkermansia spp.) and Euryarchaeota (for example, Methanobrevibacter spp.). d, Species harbouring gene clusters V1, V2, and V3 from the assembled gut metagenomes in each community type. Each dot in c and d represents an individual from the centenarian (orange), older (blue), or young (grey) groups. Horizontal lines indicate the median; box boundaries indicate interquartile range (IQR); whiskers represent values within 1.5 x IQR of the first and third quartiles. Asterisks indicate significantly different abundance in the specified comparison at FDR P < 0.05 based on a Wilcoxon rank-sum test. ns, not significant. e, Longitudinal change in faecal bile acid composition from the same individual over the course of 1-2 years. Upper numbers indicate collection interval (days). Faecal bile acids were quantified by LC-MS/MS (μmol/g wet weight faeces), and the percent of each bile acid among total faecal bile acids in each sample was calculated. f, Stability of microbial community type (1, 2, and 3) in each individual over time.
Corresponding author(s): Kenya Honda Last updated by author(s): Jul 7, 2021 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

April 2020
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability Genome sequences of the 68 strains isolated from a centenarian and 16S rRNA amplicon sequence datasets have been deposited in the DDBJ under BioProject PRJDB11902 and PRJDB11894, respectively. Metagenomic data have also been deposited in the NCBI under Bioproject PRJNA675598. LC-MS/MS data were deposited in Metabolomics Workbench under project ID PR001168 with study ID ST001851 for human faeces data and study ID ST001852 for in vitro data. All deposited data will be released before publication. Source data are provided with this paper. European Nucleotide Archive (accession number PRJEB25514) was used for the Sardinian metagenome data.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No statistical methods were used to predetermine sample size for the human or mouse studies. For the human cohorts, no sample size calculation was conducted as we were not testing for end clinical outcomes nor testing any intervention. Sample sizes therefore represent the maximum number of samples we could obtain during the recruitment period. The number of animals studied per treatment group was based on our previous knowledge of the reproducibility, balancing statistical robustness and animal welfare.
Data exclusions We excluded subjects undergoing antibiotic treatment (3 centenarians, 1 elderly) as this is known to heavily influence the composition of microbiome as well as the metabolomics data. Two human subjects who had high serum C-reactive protein levels and died shortly after the data collection were excluded entirely from this study. These criteria were pre-established in the laboratory.

Replication
Each hypothesis was tested with multiple types of experiment. Reproducibility of C. difficile infection level in animal studies was verified by performing 2-3 independent experiments, which yielded comparable results (Main Fig. 3c, d and Extended Data Fig. 10c).
Randomization For the human cohorts no random allocation was used as our study was observational and did not test any intervention. For animal studies, mice were randomized into separate cages upon arrival from the vendor. Sex-matched littermates were used and experiments were intended to test a single variable. For experiments using Cyp2a12/Cyp2c70 double knockout mice, animals were allocated across treatment groups to balance age, litter, and body weight.

Blinding
All experiments were not strictly blinded because the measurements were quantitative. Analysis of metabolite levels were conducted in an objective manner by more than two researchers.

Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.