26We use a multi-stage genome-wide association of 1 million parental lifespans of genotyped 27 subjects and data on mortality risk factors to validate previously unreplicated findings near 28 CDKN2B-AS1, ATXN2/BRAP, FURIN/FES, ZW10, PSORS1C3, and 13q21.31, and identify 29 and replicate novel findings near GADD45G, KCNK3, LDLR, POM121C, ZC3HC1, and ABO. 30 We also validate previous findings near 5q33.3/EBF1 and FOXO3, whilst finding contradictory 31 evidence at other loci. Gene set and tissue-specific analyses show that expression in foetal 32 brain cells and adult dorsolateral prefrontal cortex is enriched for lifespan variation, as are 33 gene pathways involving lipid proteins and homeostasis, vesicle-mediated transport, and 34 synaptic function. Individual genetic variants that increase dementia, cardiovascular disease, 35and lung cancer -but not other cancers-explain the most variance, possibly reflecting modern 36 susceptibilities, whilst cancer may act through many rare variants, or the environment. 37Resultant polygenic scores predict a mean lifespan difference of around five years of life 38 across the deciles. 39
Results
95Genome-wide association analysis 96 97We carried out GWAS of survival in a discovery sample of 635,205 parents (68% deceased) 98 of unambiguously British ancestry subjects from UK Biobank, and a replication sample of 99 377,035 parents (47% deceased) of other European ancestry subjects from UK Biobank and 100 26 additional populations cohorts (LifeGen ; Table S1). In each sample, we performed a sex-101 stratified analysis and then combined the allelic effects in fathers and mothers into a single 102 parental survival association in two ways. First, we assumed genetic variants with common 103 effect sizes (CES) for both parents, maximising power if the effect is indeed the same, 104 secondly, we allowed for potentially different effect sizes (PDES), maximising power to detect 105 sexually dimorphic variants, including those only affecting one sex. 106 107We find fourteen genomic regions containing SNPs with genome-wide significant (P < 5x10 -108 8 ) association in the discovery cohort, for one or both analyses (Fig. 1a). Ten of these loci 109 have been previously reported using similar data (6), but only 4 have been successfully 110 replicated so far (3-5, 10). We calculate the effects of previously replicated SNPs to be −1.06 111 (near APOE), −0.42 (CHRNA3/5), −0.76 (LPA), and +0.56 (HLA-DQA1) years of life gained 112 per minor allele, estimated from the meta-analysis of discovery and replication cohorts. We 113 also find evidence for replication (P < 0.05, one-sided test) for an additional 3 loci near 114 ATXN2/BRAP (−0.28), FURIN/FES (−0.25), and CDKN2B-AS1 (−0.25), which were previously 115 identified at genome-wide significance by Pilling et al(6) ( Table 1). 116 117While we were unable to replicate the remaining seven loci, this may be due to lack of power 118 for SNPs at or near 13q21.33 and C20orf187, where 95% confidence intervals (CIs) for 119 replication effect overlap with...