Identification of 12 genetic loci associated with human healthspan

Abstract

Aging populations face diminishing quality of life due to increased disease and morbidity. These challenges call for longevity research to focus on understanding the pathways controlling healthspan. We use the data from the UK Biobank (UKB) cohort and observe that the risks of major chronic diseases increased exponentially and double every eight years, i.e., at a rate compatible with the Gompertz mortality law. Assuming that aging drives the acceleration in morbidity rates, we build a risk model to predict the age at the end of healthspan depending on age, gender, and genetic background. Using the sub-population of 300,447 British individuals as a discovery cohort, we identify 12 loci associated with healthspan at the whole-genome significance level. We find strong genetic correlations between healthspan and all-cause mortality, life-history, and lifestyle traits. We thereby conclude that the healthspan offers a promising new way to interrogate the genetics of human longevity.

Fig. 1

The incidence of the most prevalent chronic diseases, risk of death (the mortality rate) and healthspan for UKB participants. The disease incidence increases approximately exponentially with age at approximately the same rates. Disease incidence rates are calculated independently, participants that have more than one condition during follow-up period are counted for every disease they have, except for healthspan which is defined as the first event occurred. Shaded area represents 95% confidence interval

Fig. 2

Discovery GWAS of healthspan in GCW-British individuals. The trait is a form of Martingale residual of the Cox-Gompertz proportional hazards model of healthspan as described in section Cox-Gompertz proportional hazards model and healthspan. The loci are tagged by SNPs from Table 2, labeled by the nearest gene symbol, replicated SNPs marked in bold

Fig. 3

Genetic correlation between GWAS of the healthspan and the diseases used to produce the healthspan phenotype in the UKB discovery cohort. The significant correlations marked in bold (p < 0.05 after Bonferroni correction)

Fig. 4

Thirty-five traits with significant and high genetic correlations with healthspan (|r_g|  ≥ 0.3; p ≤ 4.3 × 10⁻⁵). PMID references are placed in square brackets. Note the absence of genetic correlation between the healthspan and Alzheimer disease traits (r_g = −0.03)