Identification of 12 genetic loci associated with human healthspan

Abstract

The mounting challenge of preserving the quality of life in an aging population directs the focus of longevity science to the regulatory pathways controlling healthspan. To understand the nature of the relationship between the healthspan and lifespan and uncover the genetic architecture of the two phenotypes, we studied the incidence of major age-related diseases in the UK Biobank (UKB) cohort. We observed that the incidence rates of major chronic diseases increase exponentially. The risk of disease acquisition doubled approximately every eight years, i.e., at a rate compatible with the doubling time of the Gompertz mortality law. Assuming that aging is the single underlying factor behind the morbidity rates dynamics, we built a proportional hazards model to predict the risks of the diseases and therefore the age corresponding to the end of healthspan of an individual depending on their age, gender, and the genetic background. We suggested a computationally efficient procedure for the determination of the effect size and statistical significance of individual gene variants associations with healthspan in a form suitable for a Genome-Wide Association Studies (GWAS). Using the UKB sub-population of 300,447 genetically Caucasian, British individuals as a discovery cohort, we identified 12 loci associated with healthspan and reaching the whole-genome level of significance. We observed strong (|ρg| > 0.3) genetic correlations between healthspan and the incidence of specific age-related disease present in our healthspan definition (with the notable exception of dementia). Other examples included all-cause mortality (as derived from parental survival, with ρg = −0.76), life-history traits (metrics of obesity, age at first birth), levels of different metabolites (lipids, amino acids, glycemic traits), and psychological traits (smoking behaviour, cognitive performance, depressive symptoms, insomnia). We conclude by noting that the healthspan phenotype, suggested and characterized here, offers a promising new way to investigate human longevity by exploiting the data from genetic and clinical data on living individuals.