Aging announces itself not with damage but with absence. Long before any disease or symptom appears, the body begins quietly losing specific cells: tendon cells, kidney cells, the stem cells that maintain muscle. Two atlases assembled by Junyue Cao’s laboratory at Rockefeller University, built from more than 20 million mouse cells, show that this disappearance follows an order that is reproducible from animal to animal. The same cell types vanish at the same ages, and later the same immune populations expand. If aging were random wear and tear, the pattern would differ every time. It does not, and that reproducibility is the strongest evidence yet for an old and contested idea: that aging is less a breakdown than a program.
The first atlas, published in Science in January 2025, is called PanSci. It profiles the gene activity of more than 20 million single nuclei from 623 tissue samples spanning 13 organs plus the brain, collected from sex-balanced wild-type mice at 6, 12, and 23 months of age, with additional immunodeficient strains sampled at 3 and 16 months. The team identified 239 organ-specific main cell types, plus 31 in the brain, and 3,925 finer-grained subclusters. A single graduate student produced the entire dataset using EasySci, a low-cost combinatorial indexing technique developed in the lab, a scale of work that would normally require a consortium.
What the census revealed was not gradual decay but two distinct acts. The first act, spanning roughly 3 to 12 months of mouse age, is depletion. Between 3 and 6 months, the losses concentrate in metabolic and structural cells: brown fat cells, the nuclei of fast-twitch muscle fibers, and tendon cells. Between 6 and 12 months the losses widen to maintenance crews across the body: naïve T cells, intestinal macrophages, more tendon cells, the cells that wrap blood vessels, and multiple epithelial populations in the kidney and intestine. The second act, beginning around 12 months, reverses the direction. Instead of losing cells, the body begins expanding them, and the expansion is dominated by the immune system: cytotoxic T cells, gamma-delta T cells, plasma cells, and a population the authors call aging-associated B cells. A final wave around 16 months adds patrolling monocytes and more aging-associated B cells. In the full dataset, 174 subclusters expanded while 56 were depleted, with minimal temporal overlap between the two phases.
Cao maps these mouse stages onto human decades: the first losses in the twenties and thirties, the depletion of maintenance cells in the thirties and forties, the shift to immune expansion around age 40 to 50, and the full flowering of aging immune populations in the late fifties and beyond. In the second atlas, published in Science in February 2026, the team profiled chromatin accessibility, the physical packaging of DNA that determines which genes each cell can read, in more than 10 million cells across 21 mouse tissues. They resolved 536 main cell types and 1,828 subtypes and found that some of the most dramatic changes happen early: muscle satellite cells, tendon cells, and immune progenitors in the bone marrow lose more than half their numbers between 1 and 5 months of age, the mouse equivalent of young adulthood. These are the cells a body would need most to repair itself later, and they are gone first.
The chromatin data supply the strongest argument for order. Of 1,341,077 regulatory regions the team could measure, 279,401 changed with age, and the same regions changed in the same direction across mice, including well-defined losses and gains around specific genes. Changes were synchronized across organs, suggesting a systemic signal rather than independent organ decay. If aging were stochastic molecular damage, the affected regions would differ from mouse to mouse. Instead the pattern repeats, which is why Cao describes aging as a remodeling of the body’s cellular community driven by coordinated signals, with secreted immune molecules called cytokines as prime candidates for the conductor’s role.
The same atlas quantifies a dimension aging research has long underweighted: sex. About 40 percent of aging-associated changes in cell populations differed between males and females, with tens of thousands of chromatin regions altered in only one sex. Females showed broader immune activation with age, a pattern that may help explain why autoimmune diseases are more common in women. The finding is correlational, but it points to a mechanism with testable consequences.
The practical payoff is a target map. If aging is an ordered loss of specific vulnerable cell types, then the cells that vanish early are candidate early-warning markers, and the signals that drive the loss are candidate drug targets. Cao’s lab has made the atlas publicly available at epiage.net, and the group is already working on interventions aimed at the aging process itself rather than at individual diseases. The claim is not that aging can be stopped. The claim is narrower: that the body’s decline is structured, that its earliest events are disappearances, and that a process with this much internal order is one that intervention might hope to steer.
The atlases are mouse data, and mouse aging is not human aging. The orderliness is a description of what happens, not proof of a deliberate mechanism; the same reproducible sequence could arise from a cascade of secondary effects rather than an evolved program. And the human parallel, abrupt shifts in blood protein signatures around middle age observed in earlier studies, is suggestive rather than decisive. What the two atlases establish is a fact with consequences: the aging body does not break evenly. It empties specific rooms first, and the order of that emptying is the same every time.
Sources:
- Zhang, Z. et al. “A panoramic view of cell population dynamics in mammalian aging.” Science 387, eadn3949 (2025). DOI: 10.1126/science.adn3949. Preprint: bioRxiv 2024.03.01.583001.
- Lu, Z. et al. “Organism-wide cellular dynamics and epigenomic remodeling in mammalian aging.” Science 391, eadw6273 (2026). DOI: 10.1126/science.adw6273. Preprint: bioRxiv 2025.05.12.653376.
- Wickelgren, I. “Why Aging May Be a Program, Not a Breakdown.” Quanta Magazine, August 14, 2026.
- Rockefeller University. “Study reveals how unexpected shifts in cell populations are revising our understanding of the aging process” (December 11, 2024) and “Scientists map how aging reshapes cells across the entire mammalian body” (February 2026).
- Barnett, S.N. & Noseda, M. “Chromatin dynamics shape aging across organs.” Science 391, 869-870 (2026). DOI: 10.1126/science.aef5650.