Metabolomics profiling reveals novel markers for leukocyte telomere length

Abstract

Leukocyte telomere length (LTL) is considered one of the most predictive markers of biological aging. The aim of this study was to identify novel pathways regulating LTL using a metabolomics approach. To this end, we tested associations between 280 blood metabolites and LTL in 3511 females from TwinsUK and replicated our results in the KORA cohort. We furthermore tested significant metabolites for associations with several aging-related phenotypes, gene expression markers and epigenetic markers to investigate potential underlying pathways. Five metabolites were associated with LTL: Two lysolipids, 1-stearoylglycerophosphoinositol (P=1.6×10(-5)) and 1-palmitoylglycerophosphoinositol (P=1.6×10(-5)), were found to be negatively associated with LTL and positively associated with phospholipase A2 expression levels suggesting an involvement of fatty acid metabolism and particularly membrane composition in biological aging. Moreover, two gamma-glutamyl amino acids, gamma-glutamyltyrosine (P=2.5×10(-6)) and gamma-glutamylphenylalanine (P=1.7×10(-5)), were negatively correlated with LTL. Both are products of the glutathione cycle and markers for increased oxidative stress. Metabolites were also correlated with functional measures of aging, i.e. higher blood pressure and HDL cholesterol levels and poorer lung, liver and kidney function. Our results suggest an involvement of altered fatty acid metabolism and increased oxidative stress in human biological aging, reflected by LTL and age-related phenotypes of vital organ systems.

Keywords: biological aging; glutathione; metabolomics; oxidative stress; telomere length.

Figure 1. Telomere length, metabolite and phenotype interrelationships

Nodes represent variables where rectangles represent metabolites, circles represent phenotypes, pentagons represent expression levels and hexagons represent DNA methylation levels. Links between nodes represent significant correlations (red negative, blue positive). Thicker edges indicate stronger correlations.

Figure 2. LTL prediction performance

The figure shows the prediction performance (mean square error on Y axis) of three different Lasso models, based on metabolites only (red), clinical variables only (blue) and metabolites with clinical variables combined (green), dependent on the amount of regularization (lambda on x axis).