A new study of 1,828 people found that men and women do not just age at different rates. They age in fundamentally different ways at the cellular level.
The dominant model for how sex influences aging has been quantitative: women age more slowly than men, but through the same underlying biological process. The evidence for this view has seemed compelling. Women live longer in virtually every country on earth. Men die earlier from cardiovascular disease and cancer. If the process is the same but the speed differs, then interventions that slow aging should benefit both sexes equally, just applied with the right timing.
A new study published in Nature Communications has tested that model at cellular resolution and found it incomplete in a specific and important way.
Researchers at Duke-NUS Medical School in Singapore, led by Jacques Behmoaras with first author Harry Park and collaborators including Antonio Bertoletti and Jan Gruber, analyzed single-cell transcriptome data from 1,828 healthy individuals aged 19 to 97. The dataset comprised 3.8 million peripheral blood mononuclear cells (PBMCs), the immune cells that circulate in human blood and serve as the front line of immune defense. The sample was deliberately diverse: 35% of participants were Asian, making this one of the most ethnically representative immune aging studies conducted to date.
Rather than averaging gene expression across cell types or comparing young adults to old ones, the team tracked how every major immune cell population changed at every age, allowing them to map the actual shape of the aging trajectory rather than simply measuring the distance between two endpoints.
The shape they found was not a smooth decline. It was punctuated.
Two peaks where everything changes
The most striking structural finding in the study was the identification of two discrete peaks of change in immune cell gene expression that appeared at specific ages rather than accumulating gradually.
The first peak appeared around age 40. At this point, CD4 T cells, the helper cells that coordinate the immune system’s response to threats, showed dramatic shifts in gene expression. These were not subtle adjustments. The magnitude of change at 40 was comparable to what most aging research would expect to see across decades of gradual decline.
The second peak appeared after 60. Here, CD8 T cells, the cytotoxic killer cells that directly attack infected and cancerous cells, showed their own distinct surge of change. Again, the magnitude was acute and concentrated rather than spread evenly across years.
Between these peaks, the immune system showed relative stability. The aging process was not constant. It accelerated at specific biological moments.
“Prominent peaks of differential gene expression around 40, mainly in CD4 T cells, and after 60, mainly in CD8 T cells, were accompanied by an age-dependent decline in RNA and protein homeostasis and inflammatory polarization with sex-dependent kinetics,” the researchers wrote.
The sex-dependent kinetics is where the most medically significant finding lies.
How men and women diverge at those peaks
At both inflection points, male and female immune cells followed different programs.
Women showed a pattern of sustained and escalating immune activity through old age. After 60, female CD8 T cells showed heightened activation signatures that persisted and intensified into late life. Beyond the CD8 compartment, late-life aging signatures appeared across multiple immune cell types in women: CD4 T cells, natural killer cells, and B cells all showed pronounced female-specific changes in the decades after 60. The female immune system, in this data, was characterized by sustained remodeling, a system that kept responding, reorganizing, and activating throughout old age.
This pattern connects directly to well-established clinical observations. Women are approximately twice as likely as men to develop autoimmune conditions, diseases in which a chronically active immune system turns against the body’s own tissues. The sustained late-life immune activation seen in women’s cells provides a cellular substrate for why that vulnerability persists and intensifies with age.
Men showed a fundamentally different trajectory. The most distinctive male-specific signal appeared not in old age but earlier, in the CD4 T cells, where fluctuations in immunometabolism, the energy metabolism of immune cells, appeared in mid-life. These early fluctuations were linked to a specific epigenetic change: hypomethylation of the SSH3 gene at chromosome 11q13, a region associated with cardiovascular and metabolic disease. The epigenetic mark appeared to set the direction of male immune aging early, shaping the downstream trajectory through a pathway that differed fundamentally from what was happening in women of the same age.
This is not a quantitative difference. A faster version of the female aging program would not produce early CD4 immunometabolic shifts linked to chromosome 11q13 hypomethylation. These are qualitatively different biological processes.
What deep learning revealed about the gap
The most direct evidence that male and female immune aging are biologically distinct came from the study’s application of deep learning-based biological age clocks.
A biological age clock is a model trained to predict a person’s age from molecular data, typically gene expression or epigenetic markers. When such a clock performs well, it means the molecular data contains consistent age-related signals. When it performs poorly, it means the molecular data is too variable or noisy for the model to find reliable patterns.
The researchers built three types of age clocks from the immune cell data: one trained on men and women combined, one trained only on men, and one trained only on women. They then tested each clock’s accuracy at predicting biological age.
The sex-stratified clocks outperformed the combined model substantially. When men’s immune data was used to train the clock and then tested on men, and women’s data used to train and test on women, the predictions were significantly more accurate than when a single model tried to handle both.
This result has a straightforward interpretation: the biological features that predict age in men’s immune cells are not the same features that predict age in women’s immune cells. The systems are following different trajectories, tracking different variables, and ultimately converging on different cellular states in old age. A single clock that tries to capture both is averaging over two distinct programs and doing justice to neither.
“Deep learning-based biological age clocks stratified for sex outperformed sex-combined models, learning from transcriptional immune trajectories,” the researchers wrote.
Why this matters for medicine
The practical implications of this finding extend well beyond academic interest in aging biology.
Clinical trials for aging-related interventions, whether drugs, lifestyle modifications, or immune therapies, are typically designed and analyzed as if aging is a unified process. Age is treated as a covariate to control for, and sex is treated as a demographic variable to balance across groups. The assumption embedded in this approach is that the thing being intervened upon is the same thing in men and women.
If male and female immune aging follow distinct biological programs, that assumption fails at the foundational level. An intervention that targets the late-life CD8 activation pathway driving female aging might have no meaningful effect on the early CD4 immunometabolic pathway driving male aging, and vice versa. Clinical trials that pool men and women may produce null results not because an intervention fails, but because its effects in one sex are diluted or cancelled by the absence of effects in the other.
The researchers describe this as an opportunity for what they call precision geromedicine, the application of personalized approaches to the medicine of aging, stratified not just by individual genetics but by sex-specific biological aging trajectories.
“We thus unravel a nonlinear PBMC aging whereby targeting sex-specific pathways might allow precision geromedicine,” the team concluded.
The dataset’s ethnic diversity, with 35% Asian participants, also addresses a persistent limitation of aging research, which has historically drawn almost entirely from European-ancestry populations. The consistency of the sex-specific aging patterns across ethnic groups in this dataset suggests the findings reflect fundamental biology rather than population-specific effects, though the researchers note that further validation in more diverse cohorts remains important.
What the study cannot establish
The study used a cross-sectional design, meaning it compared different people at different ages rather than following the same individuals over time. Cross-sectional studies can identify age-related patterns but cannot definitively establish that any observed change reflects aging within individuals rather than differences between generations that happened to be in the sample.
The data came from peripheral blood immune cells specifically. The immune system in blood does not fully represent the immune landscape in tissues, where many of the most consequential aging-related changes take place. Whether the sex-specific trajectories observed in circulating immune cells correspond to similar patterns in tissue-resident immune populations is an open question.
The specific epigenetic changes linked to male CD4 immunometabolism, particularly the SSH3 hypomethylation on chromosome 11q13, require follow-up experimental work to establish whether they drive the observed aging patterns or simply track alongside them.
What the study establishes with unusual scale and resolution is that the biological program of immune aging in healthy humans is not a single process running at two speeds. It is two distinct processes, with two distinct trajectories, two distinct sets of cellular changes, and two distinct sets of downstream disease risks. The unified model of aging, however useful as a simplification, does not capture what is actually happening inside the immune system as men and women grow old.
The study, “Sex-specific trajectories of nonlinear immune aging at single-cell level”, was authored by Sopena-Rios, M., Ripoll-Cladellas, A., Omidi, F. et al. at Duke-NUS Medical School and collaborating institutions, and published August 21, 2026 in Nature Communications.
Source: Duke-NUS Medical School, Singapore. DOI: 10.1038/s43587-026-01099-x
