Consistent signatures in the human gut microbiome of longevous populations

ABSTRACT

Gut microbiota of centenarians has garnered significant attention in recent years, with most studies concentrating on the analysis of microbial composition. However, there is still limited knowledge regarding the consistent signatures of specific species and their biological functions, as well as the potential causal relationship between gut microbiota and longevity. To address this, we performed the fecal metagenomic analysis of eight longevous populations at the species and functional level, and employed the Mendelian randomization (MR) analysis to infer the causal associations between microbial taxa and longevity-related traits. We observed that several species including Eisenbergiella tayi, Methanobrevibacter smithii, Hungatella hathewayi, and Desulfovibrio fairfieldensis were consistently enriched in the gut microbiota of long-lived individuals compared to younger elderly and young adults across multiple cohorts. Analysis of microbial pathways and enzymes indicated that E. tayi plays a role in the protein N-glycosylation, while M. smithii is involved in the 3-dehydroquinate and chorismate biosynthesis. Furthermore, H. hathewayi makes a distinct contribution to the purine nucleobase degradation I pathway, potentially assisting the elderly in maintaining purine homeostasis. D. fairfieldensis contributes to the menaquinone (vitamin K2) biosynthesis, which may help prevent age-related diseases such as osteoporosis-induced fractures. According to MR results, Hungatella was significantly positively correlated with parental longevity, and Desulfovibrio also exhibited positive associations with lifespan and multiple traits related to parental longevity. Additionally, Alistipes and Akkermansia muciniphila were consistently enriched in the gut microbiota of the three largest cohorts of long-lived individuals, and MR analysis also suggests their potential causal relationships with longevity. Our findings reveal longevity-associated gut microbial signatures, which are informative for understanding the role of microbiota in regulating longevity and aging.

Introduction

As the global population ages, most countries in the world encounter significant challenges, with longevity remaining as one of the cherished goals for their nationals and citizens. Previous studies have described growing evidence highlighting geographical disparities in life expectancy at county, race, and ethnic levels,¹ which may be associated with genetics,^2,3 environmental factors, and socioeconomic factors.⁴ Dysbiosis is one of the 12 hallmarks of aging,⁵ and it contributes to multiple pathological conditions associated with age-related diseases,⁶ such as diabetes, hypertension, atherosclerosis, and osteoporosis.^7,8 These diseases affect the quality and quantity of life, but the gut microbiome as we age is not completely understood. Thus, identifying longevity or aging-specific species and pathways could have clinical implications for both monitoring and modifying gut health to achieve healthy aging.

In recent years, the gut microbiota of centenarians has garnered considerable research attention. Several studies have characterized the gut microbiota in Italian centenarians,^9–11 Japanese centenarians,¹² and Chinese centenarians,^4,13–16 providing potential insight into gut microbial composition associated with aging. For example, Akkermansia was found to be enriched in the gut of centenarians,^4,11,12,14 while Bacteroides and Faecalibacterium were more abundant in healthy younger controls.^4,14 Another study evaluated gut microbiomes associated with aging across over 9,000 individuals from three distinct populations spanning the ages 18–101 years, and found that healthy aging is characterized by a depletion of core genera found across most humans, primarily Bacteroides.¹⁷ However, few mechanisms that link longevity- or aging-correlated pathophysiology with specific microbes, functional pathways, or microbial metabolites have been identified.^12,18,19 For instance, centenarian-derived Odoribacteraceae strains can produce isoalloLCA that exerted potent antimicrobial effects against gram-positive multidrug-resistant pathogens.¹² There are limitations or unanswered questions in these studies. One is the heterogeneity that may affect the reproducibility of relevant association findings. Second, we lack more mechanistic and causal insights into the microbes and functions that are associated with human longevity. To uncover consistent signatures and provide clues for mechanistic findings, high-quality metagenomic data from diverse geographic locations and cohorts will be highly informative.

In the present study, we integrated eight longevous cohorts including one recruited by us from Jiaoling County, a longevity town of China. We utilized the shot-gun metagenomic sequencing data from a total of 1156 fecal samples to obtain accurate species-level functional profiling and aimed to identify longevity- or aging-associated microbial functions. Moreover, most observational research cannot infer the causal relationship between gut microbiota and longevity or aging. Mendelian randomization (MR) has attracted wide attention by inferring the causal relationship between variables by means of the instrumental variable of genetic variation.²⁰ Therefore, this study also assessed the causal relationships between gut microbiota and longevity using MR. These results provide valuable insights into the generalizable microbial signatures of long-lived populations and expand our understanding of bacteria-host interactions associated with longevity or aging.

Methods

Ethics approval and consent to participate

The Meizhou cohort was recruited from Jiaoling County, Meizhou City, Guangdong Province, China. Participants must meet the following criteria: (1) have lived in Jiaoling for at least five consecutive years at the time of sampling; (2) with fecal samples; (3) did not undergo antibiotic treatment within 3 weeks before the fecal samples collected; and (4) without severe disease (diabetes, inflammatory bowel disease, rheumatoid arthritis, cancer, etc.). Two hundred and sixteen participants were enrolled, consisting of 35 long-living elderly (95–105 years old), 142 younger elderly (60–89 years old), and 39 young people (20–59 years old). This study was approved by the ethics committee of the Third Affiliated Hospital of Sun Yat-sen University (Ethics number: [2019]02-010-02) and informed consent was obtained from all study participants. All subjects in the investigated Meizhou population were on an omnivorous diet. Common diets include rice, porridge, pork, pig liver, chicken, egg, tofu, green vegetables, etc.

Sample collection, DNA extraction, and sequencing

Fecal samples of Meizhou cohort were freshly collected from each subject and immediately frozen at −20°C, transported to the laboratory on an ice pack and stored at −80°C until analysis. Fecal genomic DNA was extracted with the QIAamp fast DNA stool mini kit (Qiagen, Hilden, Germany). DNA concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Scientific). Sequencing libraries were constructed using the Nextera DNA Sample Prep kit (Illumina), and paired-end sequencing (2 × 150 bp) was performed using the Illumina NovaSeq platform.

Fecal metagenomic datasets

This paper analyzes seven existing, publicly available cohorts (C1–C7) and one our recruited Meizhou cohort (C8). The sample information of eight cohorts was summarized in Supplementary Table 1. The raw sequencing data used in this study for the Japanese centenarians (C1)¹², Qidong county (C2, China),⁴ Romagna region (C4, Italy),¹⁰ Dujiangyan (C6, China),¹⁶ and Hainan (C7, China)¹⁵ are available at the NCBI database under accession numbers PRJNA675598, PRJNA613947, PRJNA553191, PRJNA624763, and PRJNA772518, respectively. The raw sequencing files for Sardinian (C3, Italy) centenarians⁹ are available at the ENA database under accession number PRJEB25514. The raw metagenomic data from Rugao City (C5, China)¹³ are available at the CNGBdb database under accession numbers CNP0002519, and our dataset from the Jiaoling County (C8, China) is also available at the CNGBdb database under accession number CNP0004699. These eight datasets adopted a shot-gun metagenomic sequencing strategy on the Illumina sequencing platform, with around 2 × 150 bp paired-end reads.

Metagenomic data processing

Low-quality reads were filtered and trimmed by fastp version 0.20.0 with the default parameters. Human reads were removed by mapping the reads to the human reference genome (hg38) with Bowtie2 version 2.4.1.²¹ The composition of microbial communities for eight datasets was computed using the MetaPhlAn version 4.0.2²² with the database mpa_vJan21_CHOCOPhlAnSGB_202103. The functional profiling was performed using HUMAnN v3.8,²³ and required databases were downloaded from https://huttenhower.sph.harvard.edu/humann_data/. We used the “humann_join_tables” function to merge the “genefamilies” and “pathabundance” files, respectively, from eight datasets. At last, we identified 638 metabolic pathways using a total of 1156 samples from the MetaCyc database (https://metacyc.org/). Besides, UniRef90 genefamilies abundance profiles were converted to enzyme commission (EC) abundance profiles using the “humann_regroup_table” function and ‘uniref90_level4ec’ option, as previously described.²⁴ We identified the EC families that were differentially expressed in different age groups.

Random forest classifier for longevity

We constructed a classifier to discriminate samples of extremely long-lived individuals (95+ years old) and non-long-lived individuals (younger elderly and young adults) based on a random forest (RF) model with an abundance of species. Five trials of a 10-fold cross-validation approach were applied to evaluate the performance of the classifier, as previously described.²⁴ The receiver operating characteristic curves (ROC) for both the discovery cohort (C1–C7) and validation cohort (C8) were plotted, or merging them into a large discovery cohort (C1–C8). The best selected variables were identified based on the maximum area under the curve (AUC) using the pROC package in R software (v4.2.3).

Statistical analysis for metagenomic data

α-Diversity was estimated based on the species profile of each sample according to the Shannon index. PCoA was performed based on the Bray-Curtis distance to compare the component of species in three or two age groups of samples from each cohort (β-diversity). The PERMANOVA analysis was performed using the “adonis” function in the “vegan” package, to determine the differences between samples from different age groups. Differential abundances of species, pathways, and enzymes were identified by the Kruskal–Wallis test in three age groups or the two-tailed Wilcoxon rank-sum test in two age groups of individuals. When multiple testing was necessary, the p values were adjusted using the Benjamini–Hochberg correction to control the false discovery rate. The analyses and visualizations were implemented in R software (v4.2.3).

Mendelian randomization (MR) analysis

MR analysis was performed to verify causal associations between gut microbiota and longevity-related traits (Supplementary Figure S12).

Five publicly available GWAS datasets of longevity-related traits were used in this study. The GWAS data of Human Healthspan consist of 300,477 British-ancestry individuals from the UK Biobank (UKB).²⁵ The GWAS data of longevity were derived from 36,745 individuals of European ancestry in multiple studies, encompassing 11,262 cases and 25,483 controls, and the cases were individuals who lived to age above the 90th percentile or 99th percentile.³ The GWAS data of lifespan consist of 1,012,240 European-ancestry individuals, including 512,047 mother and 500,193 father lifespans.²⁶ The GWAS data of parental longevity were collected for 389,166 UKB participants of European descent with data recorded on parents’ current ages or parents’ ages of death.²⁷ Pilling et al. identified all common genetic variants associated with longer parental lifespan, including seven traits (mother’s age at death, father’s age at death, mother’s attained age, father’s attained age, combined parental age at death, combined parental attained age, and both parents in top 10%).²⁷ One of the primary aging-linked physiological hallmarks is the onset of frailty,²⁸ which is becoming increasingly common with aging demography. The GWAS data of frailty included 175,226 individuals of European descent and used the frailty index (FI) to measure frailty.²⁹

Genetic variants associated with the gut microbiome were obtained from four datasets. The first dataset was conducted by the MiBioGen consortium and integrated 16S rRNA gene sequencing profiles and genotyping data from 18,340 participants across 24 cohorts.³⁰ Most of the participants had European ancestry (N = 13,266). Both genetic data and gut microbiota data were incorporated, and association estimates for a total of 211 bacterial taxa were calculated. After excluding 15 taxa of bacteria without specific names (unknown family or genus), gut microbiota was divided into 196 bacterial taxa including 9 phyla, 16 classes, 20 orders, 32 families, and 119 genera. The second dataset was obtained from 7,738 participants of the Dutch Microbiome Project (DMP), and the gut microbiota was identified by shotgun metagenomic sequencing of stool samples.³¹ It contained 207 taxa (105 of which are species) and 205 functional pathways that reflect the composition and activity of gut microbiota. The third dataset was obtained from 8956 German individuals by Ruhlemann et al.,³² which carried out a GWAS involving 430 taxa that reflect the composition of gut microbiota. The fourth dataset was a large-scale population-based cohort of 5,959 Finnish individuals enrolled in the FINRISK 2002 (FR02) cohort, which carried out a GWAS involving 473 taxa that provided insights into the composition of gut microbiota.³³

Selection of instrumental variables (IVs) in MR

We extracted the gut microbiota taxa as exposure data, including 196 bacterial taxa from the MiBioGen consortium, 412 bacterial traits (207 taxa and 205 pathways) from the DMP, 430 taxa from 8956 German individuals, and 473 taxa from 5,959 Finland individuals. To ensure the authenticity and accuracy of the causal relationship between gut microbiota and longevity, the following quality control procedures were implemented to select IVs. First, we selected IVs for each gut bacterial trait by using a loose cutoff of p < 1 × 10⁻⁵. Second, the independent IVs with the lowest P-value for each trait (r² < 0.001 and distance = 10,000 kb) were retained to reduce the influence of correlations among SNPs. Third, we calculated the F-statistic to evaluate the strength of the IVs. SNPs with F-statistics <10 was disregarded to avoid weak IV bias. Fourth, the screened SNPs were used as IVs to harmonize with summary statistics of longevity and the palindromic SNPs and fuzzy alleles were removed. Finally, we removed SNPs with p < 1 × 10⁻⁵ of outcome in harmonized data to avoid strong correlations between SNPs and outcome.

Statistical analysis in MR

We primarily employed the inverse-variance weighted (IVW) method for our analysis. To assess the heterogeneity among SNPs, we conducted Cochran’s Q test. If significant heterogeneity was observed (p < 0.05), we adopted the random-effects model; otherwise, the fixed-effects model was utilized.³⁴ Additionally, we performed sensitivity analyses to evaluate the robustness of our findings. The weighted median method was used to estimate the potential causal effects when IVs violated standard assumptions to provide a reliable estimate.³⁵ The MR–Egger method was used to detect directional pleiotropy, and the intercept of p > 0.05 was deemed to be no horizontal pleiotropy.³⁶ Besides, the MR pleiotropy residual sum and outlier (MR-PRESSO) method was also applied to test for possible bias from horizontal pleiotropy and outlier variant removal.³⁷ Furthermore, the leave-one-out test was conducted to confirm that MR estimates were not driven by strong effect SNPs. The results were visually analyzed using forest plots and scatter plots. All MR analyses were performed in the R software (v4.2.3) using “Mendelian Randomization,” “TwoSampleMR,” and “MRPRESSO” packages. We considered suggestive evidence of a potential causal association when p < 0.05.

Results

The diversity and composition of the gut microbiota in eight longevous populations

The number of samples and details of the groups for the datasets ranging from C1 to C8 are presented (see Figure 1(a)). C1 (N = 330), C2 (N = 348), and C8 (N = 216) are the three largest cohorts, each comprising over 200 samples. Initially, the α-diversity of samples was compared separately across different age groups within each dataset (Figure 1(b)). The result indicated that for C1 and C8, the α-diversity in younger elderly (E) group was significantly higher than that in young (Y) group, while for C5 and C6, the α-diversity in centenarian or nonagenarian (C) group was significantly higher than that in the young (Y) group. Subsequently, the samples from the eight datasets were consolidated into three age groups (C, E, and Y), and we observed that the α-diversity of fecal samples was significantly higher in the C and E groups compared to the Y group, indicating that gut microbial diversity is more abundant in healthy elderly individuals compared to younger adults. We further merged the samples from the E and Y groups, considering them as the non-longevous group, while the C group alone was classified as the longevous group. The results revealed no significant difference in α-diversity between these two groups.

Figure 1.

The sample information of eight longevous populations and microbial alpha-diversity of the fecal samples from different age groups. (a) A brief description of samples collected from eight longevous populations (C1 to C8). Centenarians or nonagenarians comprised the C group, the younger elderly were considered as the E group, and young adults were categorized as the Y group. (b) The violin plots illustrate the comparisons of alpha-diversity among fecal samples from different age groups within each individual population as well as when merged into the overall population. Samples from distinct groups are distinguished by different colors, and the corresponding p-values are indicated in the plots.

We conducted a comparative analysis of the identified species in the C, E, and Y sample groups, revealing that 338 species were common to all three groups (Supplementary Figure S1 and Supplementary Table S2). This finding potentially indicates that these species are core members of the gut microbiota, regardless of the age of the individuals. Moreover, we found 39 species uniquely present in the C group, 34 species uniquely present in the E group, and 25 species uniquely present in the Y group. Specifically, the 39 species, including Akkermansia sp. BIOML A40, Desulfovibrio fairfieldensis, Raoultibacter massiliensis, and Klebsiella oxytoca, were exclusively detected in the gut microbiota of centenarians or nonagenarians (C group). Additionally, another 71 species such as Limosilactobacillus fermentum and Neglecta timonensis were shared between the C and E groups, suggesting that they constitute the core components of the elderly microbiome. These results indicate that there are many species that are specifically prevalent in elderly and extremely elderly individuals, as compared to younger adults.

The results of β-diversity analysis revealed significant differences in gut microbiota composition between different age groups across six cohorts, including C1, C2, C3, C4, C6, and C8 (p < 0.05, PERMANOVA, Supplementary Figure S2). In contrast, no significant differences in microbial composition were observed between different age groups in C5 and C7 (p = 0.391 and p = 0.197, respectively, PERMANOVA). Therefore, in most population cohorts (6 out of 8), the composition of the gut microbiota undergoes significant changes with healthy aging or natural aging. Subsequently, we investigated the microbial composition at the genus level, and the results indicated that Bacteroides, Phocaeicola, Bifidobacterium, Blautia, Prevotella, Faecalibacterium, Alistipes, unclassified Clostridia, and unclassified Ruminococcaceae are the major dominant genera in the gut microbiota of longevous individuals (Supplementary Figure S3). Among non-longevous individuals (including younger elderly and young adults), the primary dominant genera were Bacteroides, Phocaeicola, Prevotella, Faecalibacterium, unclassified Lachnospiraceae, Alistipes, Roseburia, Bifidobacterium, and Clostridium. Longevous individuals exhibited a greater abundance of genera such as Alistipes, unclassified Clostridia, unclassified Ruminococcaceae, Parabacteroides, and Akkermansia, while non-longevous group harbored a notable higher abundance of Phocaeicola, Prevotella, Faecalibacterium, unclassified Lachnospiraceae, Roseburia, Clostridium, etc.

The significantly altered species within the gut microbiota of long-lived individuals from eight longevous populations

To further investigate how gut microbiota species vary with healthy aging, we identified the species that exhibited significantly different relative abundances between three or two age groups within each dataset. Specifically, for the C1 to C8 datasets, 300, 158, 120, 105, 31, 86, 42, and 63 species displayed notable differences in relative abundance, respectively (Supplementary Table S3). Subsequently, we analyzed the overlap of differentially abundant species across each dataset and discovered six species that consistently exhibited altered abundances in long-lived individuals, as compared to younger elderly and young individuals, in at least six of the eight datasets (Supplementary Table 4, Figure 2(a)). Notably, Eisenbergiella tayi consistently displayed altered abundance in seven longevous datasets, and was enriched in the long-lived group (Supplementary Table 5, Figure 2(b)). Five species, namely Methanobrevibacter smithii, Hungatella hathewayi, Ruthenibacterium lactatiformans, Enterocloster lavalensis, and Faecalibacterium prausnitzii, consistently exhibited altered abundance in six longevous datasets. Among these, the first four species were all more abundant in the long-lived group (Figure 2(c-d), Supplementary Figure S4A-B). In contrast, Faecalibacterium prausnitzii was relatively depleted in long-lived people across six datasets (Supplementary Figure S4C). In addition, Roseburia faecis, Eubacterium rectale, and Fusicatenibacter saccharivorans were significantly depleted in long-lived elderly compared to younger elderly and young individuals in five out of eight datasets. Conversely, Escherichia coli was consistently enriched in the gut microbiota of long-lived individuals in five datasets (Supplementary Table 4). Akkermansia muciniphila was consistently enriched in the gut microbiota of long-lived individuals in three datasets, specifically C1, C2, and C8, which represent the three largest cohorts (Supplementary Figure S5). These consistently altered species may be regarded as signatures in the human gut microbiome that are associated with longevity.

Figure 2.

The most consistently altered species in relative abundance between different age groups across eight longevous cohorts. (a) The heatmap shows the consistently altered species in relative abundance between different age groups across at least six out of the eight longevous cohorts. The average abundances of each species were exhibited in the heatmap. The abundance of eisenbergiella tayi differed significantly between different age groups in seven of the eight cohorts, whereas the abundance of the other five species displayed significant differences in six of the eight cohorts when comparing different age groups. (b-d) boxplots show the relative abundance of E. tayi (b), M. smithii c), and h. hathewayi (d) in each age group across eight longevous populations, respectively. The relative abundance is log 10 transformed. Samples from distinct groups are distinguished by different colors, and the corresponding significance level are indicated above each plot. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns, not significant.

Gut microbiota biomarkers for long-lived populations

To evaluate whether gut microbiota species could serve as potential biomarkers for distinguishing long-lived populations from non-long-lived populations, we employed the random forest (RF) classifier. Following feature selection utilizing 10-fold cross-validation, the area under the receiver-operating characteristic curve (AUC) achieved 0.86 for the discovery cohort (C1 to C7) and 0.79 for the validation cohort (C8), respectively (Figure 3(a-b)). The top 10 species selected based on the highest accuracy were Neglecta timonensis, Desulfovibrio fairfieldensis, Clostridium scindens, Anaerotruncus massiliensis, Clostridia bacterium, Eisenbergiella tayi, Pyramidobacter piscolens, Methanobrevibacter smithii, Hungatella hathewayi, and Bariatricus massiliensis (Figure 3(c)). When all eight datasets (C1 to C8) were considered as the discovery cohort, the area under the receiver operating curve (AUC) reached 0.85 (Figure 3(d)). The top 10 species selected based on the highest accuracy also comprised six of the listed species: Neglecta timonensis, Desulfovibrio fairfieldensis, Anaerotruncus massiliensis, Clostridium scindens, an unnamed Clostridia bacterium, and Eisenbergiella tayi, all of which were enriched in the long-lived populations (Figure 3(e)). Besides, Faecalibacterium prausnitzii, Alistipes onderdonkii, Blautia glucerasea, and Clostridium sp AF20-17LB were the remaining four species identified as features, and among these, A. onderdonkii was observed to be more abundant in long-lived populations. N. timonensis and D. fairfieldensis consistently ranked as the top two species in terms of accuracy, and their relative abundance in each dataset was also displayed (Figure 3(f-g)). Meanwhile, the abundances of A. massiliensis and C. scindens, which ranked third and fourth in terms of accuracy, were also shown for each group of samples (Supplementary Figure S6). These four species were significantly enriched in the long-lived group of individuals, at least in the two largest cohorts of C1 and C2. Overall, these results indicate that gut species have the potential to serve as biomarkers for longevity. Specifically, the overlap of six species enriched in long-lived populations under two RF classifier methods suggests that they could be regarded as robust bacterial markers associated with longevity or aging.

Figure 3.

The random forest (RF) classifier was employed to select species marker to distinguish longevous individuals from non-longevous individuals. (a-b) receiver operating characteristic (ROC) curve for the discovery cohort (C1 to C7, longevous, n = 399; non-longevous, n = 541) and validation cohort (C8, longevous, n = 35; non-longevous, n = 181) using the species abundance of gut microbiota. (c) The mean decrease accuracy (MDA) score of potential ten species markers in the RF model. (d-e) ROC curve for the discovery cohort (C1 to C8, longevous, n = 434; non-longevous, n = 722) using the species abundance of gut microbiota (d), and the corresponding MDA score of potential ten species markers in the RF model (e). Species names were colored according to the direction of enrichment. (f-g) boxplots shows the relative abundance (log10 transformed) of neglecta timonensis (F), and desulfovibrio fairfieldensis (G) in each sample group within eight cohorts. N. timonensis was not identified in C5 cohort, while D. fairfieldensis was not identified in C6 and C7 cohort.

Altered microbial pathways associated with differential species in long-lived populations

The species that underwent significant alterations in long-lived individuals from eight longevity populations were identified, and subsequently, we delved into the microbial pathways involving these species. First, we observed that Eisenbergiella tayi made a notable contribution to the pathway of PWY-7031: protein N-glycosylation (bacterial) and PWY-6749: CMP-legionaminate biosynthesis I (Supplementary Table 6). The relative abundances of PWY-7031 and PWY-6749 were significantly higher in the long-lived group compared to the elderly and young groups, especially for PWY-7031 (Figure 4(a)). E. tayi made a dominant contribution to the PWY-7031 pathway, while the other involved species comprised Bacteroides cellulosilyticus, Campylobacter jejuni, and Campylobacter concisus (Figure 4(b)). For PWY-6749, E. tayi also contributed significantly to this pathway, along with Escherichia coli, Klebsiella oxytoca, Clostridium citroniae, Methanobrevibacter woesei, and Campylobacter jejuni (Supplementary Figure S7). Second, we observed that Hungatella hathewayi made a distinct contribution to the P164-PWY: purine nucleobases degradation I (anaerobic) pathway, which was also more abundant in the long-lived group compared to the elderly and young groups (Figure 4(c-d)). Several Clostridium species, including C. clostridioforme, C. bolteae, and C. citroniae, were also involved in the P164-PWY. Third, Desulfovibrio fairfieldensis was notably enriched in the gut of long-lived individuals as mentioned above, and it significantly contributed to the pathway of PWY-7374: 1,4-dihydroxy-6-naphthoate biosynthesis I (Figure 4(f)). The abundance of PWY-7374 was significantly higher in the long-lived group compared to the elderly group (p = 3.2e-15) and the young group (p = 1e-16) (Figure 4(e)). Fourthly, Methanobrevibacter smithii was greatly involved in nine pathways that were enriched in the long-lived group (Supplementary Table 7), such as METHANOGENESIS-PWY (methanogenesis from H₂ and CO₂), PWY-5209 (methyl-coenzyme M oxidation to CO₂), PWY-8113 (3PG-factor 420 biosynthesis), PWY-6160: 3-dehydroquinate biosynthesis II (archaea), and PWY-6165: chorismate biosynthesis II (archaea) (Supplementary Fig. 8A). The detailed metabolic pathways and enzymes of PWY-6160 and PWY-6165 were displayed in Supplementary Fig. 8B. We can observe that the PWY-6165 pathway incorporates PWY-6160, as 3-dehydroquinate is an intermediate product in the chorismate biosynthesis II pathway. The abundance of the PWY-6165 pathway exhibited a notable difference between the long-lived group and the other two groups (Figure 4(g), p = 7e-12, and p = 4e-08, respectively). Methanobrevibacter smithii made a dominant contribution to PWY-6165 pathway, followed by Methanosphaera stadtmanae and Methanobrevibacter woesei (Figure 4(h)). Methanobrevibacter smithii also made an extremely dominant contribution to the PWY-6160 pathway (Supplementary Fig. 9).

Figure 4.

The relative abundance of four key microbial pathways in the C, E, and Y groups, along with the contributions of the species involved in these pathways. (a, c, e, g) boxplots show the relative abundance of PWY-7031 (a), P164-PWY (c), PWY-7374 (e), and PWY-6165 (g) in the samples of three age groups. The corresponding p-values between each pair of groups are indicated above each plot. (b, d, f, h) stacked bar plots show contributions of the species involved in the PWY-7031 (B), P164-PWY (d), PWY-7374 (f), and PWY-6165 (h). A total of 1156 samples were divided into three groups. Group C, n = 434; group E, n = 574; group Y, n = 148.

Next, we further explored the enzymes that are involved in the target pathways, along with the species that produce them. First, we focused on the PWY-7031: protein N-glycosylation (bacterial) pathway and discovered that pglC (EC 2.7.8.36), pglA (EC 2.4.1.290), pglJ (EC 2.4.1.291), pglH (EC 2.4.1.292), and pglI (EC 2.4.1.293) are the core biosynthetic enzymes involved in this pathway (Supplementary Table 8, Figure 5(a)). PglC is a glycosyl-1-phosphate transferase in the N-linked glycosylation pathway, while PglA, PglJ, PglH, and PglI are the four glycosyltransferases. For PWY-6749, the main biosynthetic enzymes include legB, legC, legH, legG, legI, and legF (Supplementary Table 9). The relative abundances of the five core enzymes of the PWY-7031 pathway, as well as the CMP-legionaminate synthase (legF, EC 2.7.7.82) from the PWY-6749 pathway, were significantly different among the three age groups, exhibiting enrichment in the long-lived group (Figure 5(b)). Then, we examined which species were involved in these important enzymes. Eisenbergiella tayi and Bacteroides eggerthii made a dominant contribution to the pglH enzyme (EC 2.4.1.292), followed by several other Bacteroides species, such as B. stercoris, B. cellulosilyticus, B. thetaiotaomicron, and B. uniformis (Figure 5(c)). Eisenbergiella tayi and Bacteroides cellulosilyticus contributed significantly to EC 2.7.7.82, while other notable contributors were Methanobrevibacter woesei, Campylobacter jejuni, Clostridiales bacterium CHKCI006, and Campylobacter coli (Figure 5(d)). These findings indicate that E. tayi is involved the PWY-7031 and PWY-6749 pathways through specific enzymes, such as PglA, PglJ, PglH, PglI, and legF.

Figure 5.

Several key microbial pathways and their involved enzymes. (a) The description of the PWY-7031 and PWY-6749 pathway routes in the MetaCyc database, and several key enzymes (EC numbers) were labeled. (b) Boxplots show the comparisons of five key enzymes involved in the PWY-7031 and two key enzymes involved in the PWY-6749 pathways between three age groups. (c-d) the contribution of species to EC 2.4.1.292 (c) and EC 2.7.7.82 (d). (e) The metabolic pathway of P164-PWY: purine nucleobases degradation I (anaerobic). (f) The metabolic pathway of PWY-7374: 1,4-dihydroxy-6-naphthoate biosynthesis I. (g) Boxplots show the comparisons of EC 2.4.1.292 and EC 2.7.7.82 between three age groups. (h-i) the contribution of species to EC 1.21.4.2 (h) and EC 3.2.2.26 (i). Group C, n = 434; group e, n = 574; group Y, n = 148.

Moreover, we examined the P164-PWY, which is the purine nucleobase degradation I (anaerobic) pathway, in which Hungatella hathewayi is involved (Figure 5(e)). Two core enzymes of the P164-PWY pathway are glycine reductase (EC 1.21.4.2) and acetate kinase (EC 2.7.2.1), which catalyze the transformation of glycine into acetyl phosphate and then further converted into acetate in the next step. Regarding the PWY-7374 (1,4-dihydroxy-6-naphthoate biosynthesis I) pathway, which involves Desulfovibrio fairfieldensis, two of the core enzymes are chorismate dehydratase (EC 4.2.1.151) and futalosine hydrolase (EC 3.2.2.26) (Figure 5(f)). The abundances of both EC 1.21.4.2 and EC 3.2.2.26 were notably more abundant in the samples from the long-lived group compared to the elderly and young group (Supplementary Table 10, Figure 5(g)). By analyzing the species involved in EC 1.21.4.2, we found that Hungatella hathewayi is one of the most dominant contributors in samples from long-lived individuals (Figure 5(h)). Similarly, by analyzing the species involved in EC 2.7.2.1, we found that Hungatella hathewayi is also one of the contributors. Therefore, Hungatella hathewayi could involve in the pathway purine nucleobases degradation I by the specific enzymes of glycine reductase and acetate kinase. Desulfovibrio fairfieldensis is one of the contributors involved in EC 4.2.1.151, although the biggest contributors are three Alistipes species, including A. putredinis, A. finegoldii, and A. onderdonkii (Supplementary Fig. 10). Desulfovibrio fairfieldensis is also the most distinct contributing species to EC 3.2.2.26 in samples from the long-lived group (Figure 5(i)). These results suggest that Desulfovibrio fairfieldensis could involve in the pathway 1,4-dihydroxy-6-naphthoate biosynthesis I by the specific enzymes of chorismate dehydratase and futalosine hydrolase. Moreover, EC 2.2.1.10 (aroA) and 1.4.1.24 (aroB) were the two important enzymes of PWY-6160 and PWY-6165 pathways (Supplementary Fig. 8B), and Methanobrevibacter smithii potentially contribute to these two enzymes (Supplementary Fig. 11).

MR suggest the causal relationship between gut microbiota and longevity

Although we have identified species and pathways associated with longevity or aging through metagenomic analysis, the causal relationship between the gut microbiome and longevity or aging still requires further exploration. After conducting an MR analysis utilizing the most comprehensive GWAS summary data currently available (Supplementary Table 11 , Supplementary Fig. 12), many significant associations were uncovered. We focused our attention on certain associations that those genera and species were consistently enriched in gut microbiota of long-lived individuals across multiple cohorts (Supplementary Table 12). At the genus level, Hungatella, Erysipelatoclostridium, Anaerotruncus, and Desulfovibrio consistently demonstrated enrichment in the gut microbiota of extremely long-lived individuals across at least five distinct cohorts (Supplementary Fig. 13). Hungatella was significantly positively correlated with the parental longevity of mother’s age at death (IVW, OR = 1.036, p = 0.009, Figure 6(a)). Erysipelatoclostridium exhibited a positive correlation with the parental longevity of mother’s attained age (IVW, OR = 1.017, p = 0.044, Figure 6(a)). Anaerotruncus exhibited a negative correlation with parental longevity of combined parental age at death (IVW, OR = 0.952, p = 0.025; Weighted median, OR = 0.941, p = 0.02, Figure 6(a)). Notably, Desulfovibrio was significantly linked to multiple longevity-related traits (Figure 6(b)). Desulfovibrio from the MiBioGen consortium dataset had a positive impact on parental longevity, specifically increasing the combined parental age at death (IVW, OR = 1.032, p = 0.036). However, in the German individual’s dataset, Desulfovibrio showed a negative association with lifespan (IVW, OR = 0.978, p = 0.001), parental longevity of both parents in top 10% (IVW, OR = 0.993, p = 0.038) and the combined parental age at death (IVW, OR = 0.985, p = 0.016). Concurrently, it displayed a positive correlation with the combined parental attained age (IVW, OR = 1.016, p = 0.001), and the father’s attained age (IVW, OR = 1.013, p = 0.001).

Figure 6.

MR analysis suggest the causal relationships between gut microbiota and longevity-associated traits. (a) Forest plot shows the significant influence of hungatella, erysipelatoclostridium, anaerotruncus on specific longevity-associated traits. (b) Forest plot shows the significant influence of desulfovibrio abundance on multiple longevity-associated traits. MR methods include IVW, MR Egger, and weighted median.

Moreover, Alistipes was found to be enriched in the gut microbiota of long-lived populations across the three largest datasets: C1, C2, and C8 (Supplementary Fig. 14). The Alistipes abundance was significantly associated with parental longevity of combined parental age at death (Supplementary Table 12). More specifically, Alistipes senegalensis was positively associated with lifespan (IVW, OR = 1.039, p = 0.008), while Alistipes shahii abundance in stool was linked to increased longevity of age above the 90th percentile (IVW, OR = 1.18, p = 0.007). In addition, there were no significant associations detected between longevity-correlated traits and species such as Neglecta timonensis, Desulfovibrio fairfieldensis, Clostridium scindens, Anaerotruncus massiliensis, Eisenbergiella tayi, Methanobrevibacter smithii, and Hungatella hathewayi. However, we found Akkermansia muciniphila was negatively correlated with longevity (>99th percentile, OR = 0.803, p = 0.03) but was positively correlated with three traits of parental longevity (both parents in top 10%, OR = 1.012, p = 0.047; mother’s age at death, OR = 1.021, p = 0.033; combined parental age at death, OR = 1.025, p = 0.012) in the IVW method (Supplementary Fig. 15). Therefore, the enrichment of specific species and genera in long-lived populations may have a causal influence on longevity.

Discussion

In this study, we examined the gut microbiome of eight longevous populations from China, Japan, and Italy. To the best of our knowledge, this is the first large-scale integrated metagenomic cohort study of long-lived individuals to date, comprising a total of 1156 fecal samples. We identified many altered species and their associated functional pathways that may contribute to longevity or aging. Moreover, we employed MR analysis to infer the potential causal relationship between gut microbiome and longevity.

The first notable finding of our study is that when combining data from eight longevity cohorts, the alpha-diversity of gut microbiota in long-lived populations was observed to be higher compared to that in younger populations. There are many species that are specifically prevalent in the gut microbiota of elderly and extremely elderly individuals compared to young individuals, such as Limosilactobacillus fermentum, Christensenellaceae bacterium NSJ-44, and Neglecta timonensis. Some of these species may be associated with longevity or specific to aging. Additionally, a diverse gut microbiota ecosystem may suggest flexible adaptability to perturbations, such as illness, and could potentially serve as a marker of longevity. The second key finding is that Eisenbergiella tayi, Methanobrevibacter smithii, Hungatella hathewayi, Ruthenibacterium lactatiformans, and Enterocloster lavalensis were consistently enriched in the gut microbiota of extreme elderly individuals (centenarians or nonagenarians), exhibiting alterations across the six to seven longevity-related cohorts. Upon analyzing the metabolic pathways involving these species, we discovered that E. tayi significantly contributes to PWY-7031 (bacterial protein N-glycosylation) and PWY-6749 (CMP-legionaminate biosynthesis I). N-glycosylation is one of the major glycosylation pathways, initiated in the endoplasmic reticulum by the oligosaccharyltransferase, and it has roles in folding, quality control, stability, transport, and function of proteins.³⁸ The first bacterial N-linked protein glycosylation pathway (encoded by the pgl gene cluster) was described in Campylobacter jejuni.^39,40 However, protein N-glycosylation is not restricted to pathogens such as C. jejuni but also exists in commensal organisms such as certain Bacteroides species.⁴¹ Notably, glycation-induced biological products are associated with aging, neurodegenerative disorders, diabetes and its complications, skin photoaging, osteoporosis, and progression of some tumors.⁴² In addition, E. tayi and Campylobacter are also involved in the CMP-legionaminate biosynthesis I pathway. This pathway involves unique GDP-sugar intermediates, rather than the UDP-sugar intermediates, and LegF could efficiently catalyze production of CMP-legionaminic acid.⁴³ Legionaminic acids are analogs of sialic acid that occur in cell surface glycoconjugates of several bacteria including Campylobacter jejuni, Escherichia coli, and Salmonella enterica that modulate cellular interactions.⁴⁴ The exact mechanism of how E. tayi is involved in the bacteria–host interactions has yet to be characterized.

Notably, M. smithii was significantly correlated with age,¹⁵ suggesting that it may play a role in aging. Methanobrevibacter smithii made a significant contribution to nine pathways including methanogenesis from H₂ and CO₂, methyl-coenzyme M oxidation to CO₂, 3PG-factor 420 biosynthesis, 3-dehydroquinate biosynthesis II (archaea), and chorismate biosynthesis II (archaea). Methanogenesis is an anaerobic respiration that generates methane as the final product of metabolism, and in archaea, F420 is a key coenzyme. Chorismate is synthesized in five steps from 3-dehydroshikimate. Chorismate is an important intermediate that leads to the biosynthesis of several essential metabolites, including aromatic amino acids, vitamins E and K, ubiquinone, and certain siderophores.⁴⁵ In mammalian cells, vitamin K functions as an essential vitamin for the activation of several proteins involved in blood clotting and bone metabolism.⁴⁶

Moreover, Hungatella hathewayi made a distinct contribution to P164-PWY: purine nucleobases degradation I (anaerobic). Hungatella hathewayi involved the glycine reductase (EC 1.21.4.2) and acetate kinase (EC 2.7.2.1) that catalyze glycine into acetyl phosphate and next step into acetate, respectively, which may decrease the purine accumulation and down-regulate the levels of uric acid in serum. A previous study showed that the urine-degrading bacteria included Hungatella hathewayi, which modulate abundance of purines in cecum and circulation.⁴⁷ Therefore, the Hungatella hathewayi in the gut microbiota of long-lived elderly may help them maintain the global purine homeostasis, lowering gout risk.

When performing the gut species classifier for longevous populations, we discovered that Neglecta timonensis and Desulfovibrio fairfieldensis were among the top ten species that could distinguish longevous individuals from non-longevous individuals. Neglecta timonensis was previously isolated from the stool of an 88-year-old woman with type 2 diabetes,⁴⁸ but the specific functional pathways it is involved in remain unclear. For Desulfovibrio fairfieldensis, it is involved in the 1,4-dihydroxy-6-naphthoate biosynthesis I pathway (PWY-7374), specifically through the actions of enzymes such as chorismate dehydratase and futalosine hydrolase. Menaquinone is biosynthesized from chorismate via a classical menaquinone pathway involving 1,4-dihydroxy-2-naphthoate or via an alternative futalosine pathway involving 1,4-dihydroxy-6-naphthoate.⁴⁹ Menaquinone, broadly designated as vitamin K2, is endogenously synthesized by intestinal bacteria, can prevent age-related diseases such as osteoporosis-induced fractures⁵⁰ and cardiovascular disease.^51,52 Therefore, the presence of Desulfovibrio fairfieldensis in the elderly may contribute to preventing age-related diseases.

The genera Hungatella, Erysipelatoclostridium, Anaerotruncus, and Desulfovibrio were consistently enriched in the gut microbiota of at least five cohorts of extremely long-lived individuals. MR results indicated that Hungatella and Erysipelatoclostridium were significantly positively correlated with parental longevity, and on the contrary, Anaerotruncus was negatively associated with parental longevity. Desulfovibrio was significantly associated with lifespan and multiple traits of parental longevity. The potential causal relationship between these species and longevity could be attributed to the functional pathways in which they are involved, as we have previously analyzed.

In addition, Alistipes and Akkermansia muciniphila were consistently found to be enriched in the gut microbiota of the three largest cohorts of long-lived individuals. MR results revealed that the Alistipes abundance was positively correlated with the combined parental age at death. Two species within the Alistipes genus, A. senegalensis and A. shahii, were positively associated with lifespan and longevity, respectively. Interestingly, an increased relative abundance of fecal Alistipes has been causally linked to a decreased triglyceride concentration.²⁰ Alistipes may exhibit protective effects against liver cirrhosis and cardiovascular diseases.⁵³ Furthermore, Akkermansia muciniphila showed a positive correlation with three traits of parental longevity, but also showed a negative correlation with longevity (>99th percentile). Despite the existence of opposing causal relationships, A. muciniphila are more likely to act as a beneficial microorganism and promote health.⁵⁴ A. muciniphila can improve metabolic diseases,^55,56 and protect against lethal sepsis in animal models.⁵⁷ Moreover, oral gavage of A. muciniphila extends the lifespan of progeroid mice, suggesting their pro-health activities.⁵⁸ More experimental studies are needed in the future to uncover novel mechanisms.

Our study has two limitations. First, we lack individual-level data on other host factors, including economic, behavioral, and environmental factors, which can independently influence longevity apart from gut microbiome. Second, in our MR study, most participants enrolled in the GWAS dataset are of European descent, thus limiting the generalizability of our association findings to other racial populations. Therefore, further large-scale population studies incorporating diverse longevity phenotypes and individual-level data are necessary.

Conclusion

We found consistent signatures in the human gut microbiome of eight longevous populations. Eisenbergiella tayi, Methanobrevibacter smithii, Hungatella hathewayi, and Desulfovibrio fairfieldensis were consistently enriched in the gut microbiota of long-lived individuals, and they could regulate the host health or aging by corresponding functional pathways. The bacterial protein N-glycosylation involving E. tayi may play a significant role in regulating physiological and pathological processes during aging. The purine nucleobase degradation I pathway involving Hungatella hathewayi may assist elderly individuals in maintaining global purine homeostasis, thereby reducing the risk of gout. The 1,4-dihydroxy-6-naphthoate biosynthesis I involving Desulfovibrio fairfieldensis may synthesize vitamin K2 that prevents age-related diseases such as osteoporosis-induced fractures and cardiovascular disease. All the findings contribute to a more nuanced understanding of the intricate relationships between the microbiome and longevity and provide a unique point of intervention to improve late-life health.

Funding Statement

This study was supported by the National Key Research and Development Program of China [2020YFA0907800], the National Natural Science Foundation of China [32100039], the Shenzhen Science and Technology Program [KQTD2020082014582202, RCJC20231211085944057, and ZDSYS20220606100803007], and the Foshan Science and Technology Innovation Program [2120001010795].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The metagenomic dataset from Jiaoling county (Meizhou, Guangdong province, China) has been deposited into the CNGB Sequence Archive (CNSA) of China National GeneBank DataBase (CNGBdb) with accession number CNP0004699 and CNP0005686.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2024.2393756