Plasma metabolomic profiles associated with mortality and longevity in a prospective analysis of 13,512 individuals

Abstract

Experimental studies reported biochemical actions underpinning aging processes and mortality, but the relevant metabolic alterations in humans are not well understood. Here we examine the associations of 243 plasma metabolites with mortality and longevity (attaining age 85 years) in 11,634 US (median follow-up of 22.6 years, with 4288 deaths) and 1878 Spanish participants (median follow-up of 14.5 years, with 525 deaths). We find that, higher levels of N2,N2-dimethylguanosine, pseudouridine, N4-acetylcytidine, 4-acetamidobutanoic acid, N1-acetylspermidine, and lipids with fewer double bonds are associated with increased risk of all-cause mortality and reduced odds of longevity; whereas L-serine and lipids with more double bonds are associated with lower mortality risk and a higher likelihood of longevity. We further develop a multi-metabolite profile score that is associated with higher mortality risk. Our findings suggest that differences in levels of nucleosides, amino acids, and several lipid subclasses can predict mortality. The underlying mechanisms remain to be determined.

Fig. 1. Schematic of the study design.

We examined the associations of circulating metabolites with all-cause, cardiovascular, and cancer mortality among participants with available metabolomics data in Nurses’ Health Study, Nurses’ Health Study II, and Health Professionals Follow-up Study. We further assessed the associations between metabolites and longevity, defined as reaching age 85 years. In addition, we developed a multi-metabolite profile score for all-cause mortality and examined its association with mortality and longevity. Results were externally replicated in PREDIMED.

Fig. 2. Metabolome-wide associations for all-cause mortality, cardiovascular mortality, cancer mortality, and longevity.

a Volcano plot for the associations between metabolites and all-cause mortality and HR (95% CI) per 1-SD increment for top 40 metabolites (20 positive and 20 negative). b Scatter plot for analysis stratified by case/control groups. c Scatter plot for log(HR) (β coefficient from Cox proportional hazard regression) for cardiovascular mortality versus log(HR) for all-cause mortality. d Scatter plot for log(HR) for cancer mortality versus log(HR) for all-cause mortality. e Scatter plot for log(OR) (β coefficient from logistic regression) for longevity versus log(HR) for all-cause mortality. All results were from the multivariable Cox model stratified by study cohorts, original sub-studies, and the case/control status in the original sub-study and adjusted for age, fasting status, body mass index, race, multivitamin use, smoking status, physical activity, diabetes, hypertension, antihypertensive medication use, hypercholesterolemia, lipid-lowering medication use, total energy intake, alcohol intake, and Alternate Healthy Eating Index. Metabolites with * indicate representative names. Source data are provided as a Source Data file.

Fig. 3. Associations of metabolite groups and modules with all-cause mortality and longevity.

a Association between knowledge-based metabolite groups and all-cause mortality. Metabolite Set Enrichment Analysis was used to estimate enrichment scores based on estimates from multivariable Cox models. b Metabolome-wide association for all-cause mortality by WGCNA derived modules, and association for each module, estimated using multivariable Cox models. c Metabolome-wide association for longevity by WGCNA derived modules, and association for each module, estimated using multivariable logistic regression models. All results were from the multivariable model stratified by study cohorts, original sub-studies, and the case/control status in the original sub-study, and adjusted for age, fasting status, body mass index, race, multivitamin use, smoking status, physical activity, diabetes, hypertension, antihypertensive medication use, hypercholesterolemia, lipid-lowering medication use, total energy intake, alcohol intake, and Alternate Healthy Eating Index. Abbreviation: WGCNA, weighted gene co-expression network analysis. Source data are provided as a Source Data file.