Genetic determinants of proteomic aging

Abstract

Changes in the proteome and its dysregulation have long been known to be a hallmark of aging. We derived a proteomic aging trait using data on 1459 plasma proteins from 44,435 UK Biobank individuals measured using an antibody-based assay. This metric is strongly associated with four age-related disease outcomes, even after adjusting for chronological age. Survival analysis showed that one-year older proteomic age, relative to chronological age, increases all-cause mortality hazard by 13 percent. We performed a genome-wide association analysis of proteomic age acceleration (proteomic aging trait minus chronological age) to identify its biological determinants. Proteomic age acceleration showed modest genetic correlations with four epigenetic clocks (R_g = 0.17 to 0.19) and telomere length (R_g = -0.2). Once we removed associations that were explained by a single pQTL, we were left with three signals mapping to BRCA1, POLR2A and TET2 with apparent widespread effects on plasma proteomic aging. Genetic variation at these three loci has been shown to affect other omics-related aging measures. Mendelian randomisation analyses showed causal effects of higher BMI and type 2 diabetes on faster proteomic age acceleration. This supports the idea that obesity and other features of metabolic syndrome have an adverse effect on the processes of human aging.

Fig. 1. Performance and feature contributions of the aging clock model.

A Performance of the elastic net model for chronological age trained on Olink data in the UK Biobank cohort (N = 44,435). B Distribution of elastic net model coefficients for individual Olink proteins (N = 1459).

Fig. 2. Proteomic aging stratifies individuals into divergent mortality trajectories.

Kaplan-Meier plot showing the relationship between predicted proteomic age and all-cause mortality. Time (years) since study baseline assessment visit. Oldest 10% (blue line) indicates the oldest decile of proteomic age (adjusted for chronological age; P = 1.61×10^–120). 95% confidence intervals are shown in lighter shading.

Fig. 3. GWAS of proteomic age acceleration.

Manhattan plot of GWAS results for proteomic age acceleration as outcome. The dashed orange line denotes the genome-wide significance threshold (P = 5×10^–8).