A cellular map of human aging: 14 modules instead of a single biological age
A preprint identified 14 groups of similar age-related signals associated with cells from different organs
On August 13, the authors combined data from 49 388 UK Biobank participants with a single-cell tissue atlas. Of 128 organ and cell type combinations, 107 produced usable age estimates. The relationships among these estimates formed 14 modules.
A recent review of protein markers of aging highlighted a limitation of any single plasma-based age measure: it combines signals from different processes. In a new preprint, researchers examined which indirect estimates of cellular aging change together within the same person.
The researchers used Tabula Sapiens, a single-cell atlas of human tissues that shows which genes are characteristic of different cell types in different organs. For each organ and cell type pair, the team selected these genes, matched them to plasma proteins, and constructed an age estimate. This estimate indicates how far a participant’s protein profile differs from the expected profile of people of the same age.
The authors then compared these deviations within the same individuals. Estimates for cells lining the lungs, kidneys, and intestines formed a shared epithelial module. Estimates for immune cells from the blood, spleen, and bone marrow formed several immune modules. Models for liver cells, pancreatic glandular cells, and bone marrow stem cells formed separate programs.
The age estimates grouped primarily by cell type rather than by organ. Estimates within the same module were more closely aligned than estimates paired only because they came from the same organ. The mean correlation ranged from 0,59 to 0,82 within modules, compared with 0,32 to 0,55 for pairs linked only by organ. Similar cells from different organs can therefore follow the same pattern, while several aging programs can coexist within one organ.
In an independent cohort with repeat samples collected about 15 years apart, the module estimates showed moderate to strong agreement between measurements. Protein signatures associated with epithelial, vascular, and stromal cells changed most strongly around age 50, while the largest change for antibody-producing B-cells occurred at about age 65. All 14 modules were statistically associated with the risk of death and future diagnoses in UK Biobank.
Previous work had already used blood proteins to construct age estimates for individual cell types. The current preprint adds the broader structure connecting these estimates. Aging can be described as a profile of several cellular programs that may change together or independently within the same person.