The Genetics of Aging is Not the Same in Males and Females

Today our paper is out in Nature 🥳

It took roughly two decades of mouse aging data, a team spread across two continents, and a mapping method we had to invent because nothing existing was quite right. Here's what we found, and why I think it matters.

The Setup

We used a population of 6,438 genetically diverse mice and tracked every one of them from puberty to death. The oldest mouse died at 1,456 days. The youngest died at 46. We had complete lifespan data, body mass at five time points, and genotype data at 891 markers covering ~10.6 million sequence variants.

The question we were trying to answer wasn't just which genes affect lifespan, it was when do they act, in which sex, and how do they interact with each other over time. That's a very different (and more interesting) question, and it required a completely different method.

The Actuarial Approach

Standard QTL mapping asks: does this variant associate with mean lifespan across all mice? Our approach asks: does this variant associate with mean lifespan of mice that survived to at least X days? We repeated this at 72 truncation points, from T₄₂ to T₁₁₀₀, in 15-day steps.

This is what I'd call survivorship-stratified mapping, and it turns out to matter enormously. A locus that only kills mice after day 900 is completely invisible if you just look at all-cause mortality from birth. You'd miss it entirely. Conversely, a locus that drives early male mortality looks like a large effect in young survivorships but fades completely when you condition on living past ~725 days.

What We Found

29 Vita loci, regions of the mouse genome that influence lifespan directly, as well as 30 Soma loci, regions that modulate the correlation between body mass and life expectancy.

That second category is new. We used correlated trait locus (CTL) mapping to ask: does a genotype change the relationship between how big you are and how long you live, not just one or the other independently? Nineteen Soma loci mediate higher mortality in heavier young mice (the classic 'big mice die young' effect). Eleven flip that: in older mice, being heavier is protective, and these loci mediate that trade-off.

The dynamics of Vita loci are genuinely complex. Some act only early in life (before T₅₀₀). Some kick in only after T₈₆₀. Some have effects that reverse polarity with age: a genotype that extends life at T₄₀₀ shortens it at T₉₀₀. We categorized these into four types: durable, RAM (Rate of Ageing Modulator), age-restricted, and reversal. Vita4a in males is a striking example of the reversal class: offset mortality waves between different haplotypes create a biphasic effect that swings 15 days across just 45 days of survivorship.

The Sex Split

This is the result I find most striking. The epistatic interactions among all 59 Vita and Soma loci are almost entirely sex-specific. 78 links in males, 72 in females, and only 2 overlap. Two. The genetic architecture of aging in males and females is, for practical purposes, separate. Mapping sexes together without modelling a gene-by-sex interaction would have been actively misleading for multiple loci. Vita2b is the most egregious case.

From Maps to Mechanisms

We didn't stop at maps. For Vita9b, a locus that modulates late-life mortality in both sexes, we tested 15 C. elegans orthologues using RNAi knockdown. Four had significant effects on motility (a validated lifespan proxy). One of them, acds-10 (mouse: Acad11), boosted aged motility to levels comparable to daf-2 knockdown, which is the IGF-1 receptor, a well-established longevity intervention in worms. We followed that up with Mendelian randomization in humans: APEH, another candidate, associates with paternal age at death in UK Biobank.

That's the pipeline: mouse loci, worm knockdowns, human GWAS, Mendelian randomization. It's not clean or fast, but it connects the map to the mechanism.

The Paper, The Code and The Data

Read the paper: https://www.nature.com/articles/s41586-026-10407-9
All code is on GitHub: https://github.com/DannyArends/UM-HET3/
Data (lifespan and body weight): https://genenetwork.org/
Supplementary files: files.genenetwork.org/current/umhet3_2025/