Genomics of 1 million parent lifespans implicates novel pathways and common diseases and distinguishes survival chances

Listed are SNPs or close proxies (r² > 0.95), which have been previously reported to associate with lifespan or extreme longevity. At or near – Gene, cluster of genes, or cytogenetic band in close proximity to lead SNP; Chr – Chromosome, Position – Base-pair position on chromosome (build GRCh37); A1 – the effect allele, increasing lifespan in discovery; Freq1- Frequency of the A1 allele in the replication sample, or if missing, the discovery sample; Sex – sex of the individuals or their parents used in the discovery and replication; Beta1 – the log_e(protection ratio) for carrying one copy of A1 under additive dosage model, inferred for discovery (see Materials and methods); SE – Standard Error, calculated from reported P value and inferred effect estimates for discovery, assuming a two-sided test; Years – Years of lifespan gained for carrying one copy of the A1 allele; P – P value reported by original study for discovery, one-sided P value for the Wald test association between imputed dosage and cox model residual for the replication. For discovery, except Pilling et al’s SNPs (Pilling et al., 2017), where we re-calculated effects directly from individual UKBB data ourselves, effects sizes have been converted to a common scale to enable comparison. Study – original study that identified the candidate SNP; Sample – independent sample used to replicate the results (a = Full dataset, b = LifeGen excluding UK Biobank). Loci showing nominal replication (p < 0.05) are bolded.

At or near – gene, cluster of genes, or cytogenetic band near lead SNP; Proxy – the rsID of the nearest (r2) SNP reported by Deelen et al.; Chr – Chromosome; Position – Base-pair position (GRCh37); A1 – the effect allele, A0 – the reference allele, Freq1 – the frequency of A1 allele; Beta1 – the log hazard ratio (in self) for a carrier of 1 copy of A1; SE – standard error; P – P value for test of association between proxy and lifespan (for IVM replication this is one sided); Discovery – the combined GWAS of UKBB genomically British, UKBB other and LifeGen; Replication – the GWAMAs of Deelen et al. (2014) and Broer et al. (2015), recalibrated (using APOE) to log hazard ratios, combined using inverse-variance meta-analysis, and adjusted for sample overlap; Alpha – the ratio of effect size in replication to discovery (note as this was calibrated on APOE, that result was necessarily 1).

At or near – Gene, cluster of genes, or cytogenetic band in close proximity to lead variant. Variant – rsID, longevity allele. Parent – Parent. Age range – Lower limit to upper limit of age in analysis. N – Number of lives used for the analysis (e.g. a parent aged 55 contributed to analysis of 40–50 and 50–60, but not 60–70). Deaths – Number of deaths within the age range. Beta – log_e(protection ratio) for 1 copy of effect allele in self in the age band (i.e. 2 x observed due to kin cohort method). SE – Standard error. Z – Test statistic for test of H₀. P – P value of two sided test of association.

At or near – Gene, cluster of genes, or cytogenetic band in close proximity to lead variant. Variant – rsID, longevity allele. Beta – Moderator effect estimate of sex (categorical variable, being male) or age (ordinal variable, mean age in age band) on lead SNP effect on lifespan. SE – Standard error. P – P value for association of SNP lifespan effect size with age or sex. Q – Benjamini-Hochberg FDR-corrected P value. Bolded lines contain sex or age-specific effects passing a 5% FDR threshold.

Genome-wide significant disease SNPs from the GWAS catalog are listed with the amount of lifespan variance explained (LVE), with disease-protective alleles signed positively when increasing lifespan and signed negatively when decreasing lifespan. SNPs with limited evidence of an effect on lifespan are greyed out: an FDR cut-off of 1.55% is applied simultaneously across all diseases, allowing for one false positive among all significant SNPs. Secondary pleiotropic SNPs (i.e. those associating strongly with another one of the diseases, as assessed by PheWAS in UK Biobank) are coloured, as less relevant to the disease in question. Of these, turquoise SNPs show one or more alternative disease associations in the same direction and at least twice as strong (double Z statistic) as the principal disease, while brown SNPs show one or more significant associations with alternative disease in the opposite direction that explains the negative association of the disease-protective SNP with lifespan. Of specific interest is the SNP near MAGI3, which is reported as a breast cancer SNP but associates more strongly with CVD in UK Biobank and shows no evidence of sex-specific effects on lifespan. However, we do not classify it as a CVD SNP as its main effect on lifespan is likely due to protection from autoimmune disease by a nearby missense variant (rs6679677_C, r² > 0.6, 95% CI log OR type one diabetes –0.74 to –0.46 (Cooper et al., 2008); rheumatoid arthritis −0.66 to −0.50 (Raychaudhuri et al., 2008), and carrying these diseases can reduce life expectancy up to 13 years (Myllykangas-Luosujärvi et al., 1995; Livingstone et al., 2015). Similarly, the HLA-DQA1 locus also associates most strongly with autoimmune disease and is therefore absent from the analysis. At or near – Gene, cluster of genes, or cytogenetic band in close proximity to lead variant. Chr – Chromosome. Position – Base-pair position on chromosome (build GRCh37). A1 – Allele protecting from disease or disease risk factors. Freq1 – Frequency of the disease-protective allele in the discovery + replication sample. Years – Years of lifespan gained for carrying one copy of the A1 allele. P – P value for association with lifespan under CES assumption (left), P value for genome-wide significant association with disease as reported in the GWAS catalog (right). Q – Benjamini-Hochberg FDR-corrected P value for association with lifespan. LVE – Lifespan variance explained, signed positively when A1 increases lifespan and negative when A1 decreases lifespan. Pleiotropic – SNP shows evidence of pleiotropy, see definition above. Trait – Disease trait reported in GWAS catalog. Beta1 – log OR for having the reported disease, or unit increase in risk factors associated with disease, per copy of A1 allele. PMID – PubMed identification number of the study reporting the disease association. Z estimates – Z statistic for association with disease in unrelated, Gen. British UK Biobank samples. Missing statistics indicate the SNP is not present in the CES meta-analysis summary statistics and its LVE has been imputed from the closest proxy (min. r² > 0.9) or proxies if equally close.

A polygenic risk score was made for each subject using GWAS results that did not include the subject sets under consideration. Parent survival information (age and alive/dead status) was stratified by sex and age. Sample – Out-of-sample subsets of UK Biobank individuals used for PGRS association (E and W: English and Welsh). N – Number of parental lifespans reported by sample individuals and used for the analysis (e.g. a parent aged 55 contributed to analysis of 40–50 and 50–60, but not 60–70). Deaths – Number of parental deaths within the age range reported by sample individuals. Ages – Lower limit to upper limit of age in analysis. Beta – log_e(protection ratio) for one standard deviation in polygenic score in self in the age band (i.e. 2 x observed due to kin cohort method). SE – Standard error. P – P value of two sided test of association.

UKBB phenotypes included 29 self-reported non-cancer disease fields for the participants and each of their parents, which included 474 integer-value coded diseases. These 474 diseases were aggregated and meta-analysed into four major mortality-increasing disease groups, namely CVD, diabetes, neurological and pulmonary disorders. Cancer was the fifth major disease group and was coded as either the occurrence or absence of cancers instances throughout the participant’s lifetime. Codes not shown were excluded from the analysis.

A polygenic risk score was made for each subject using GWAS results that did not include the subject sets under consideration. Disease associations have been identified in the subjects, their siblings, or parents, at FDR 5%. See Figure 10—source data 1 for disease code grouping. Sample –– Out-of-sample subsets of UK Biobank individuals used for PGRS association (E and W: English and Welsh). Kin – Family member for which the disease was reported. Trait – Disease reported by UK Biobank subject. N – Number of individuals tested. Cases – Number of reported individuals or kin carrying the disease. Beta – log OR of NOT carrying the disease per standard deviation of PGRS (i.e. positive beta indicates the PGRS protects from disease). SE – Standard Error. P – Two-sided P value. Q – Benjamin-Hochberg adjusted P value.