Atlas of 375,000 Cells Maps How Blood Vessel Linings Age

Endothelial cells line every blood vessel and control vascular tone, permeability, clotting, and the growth of new vessels. Their dysfunction appears to precede overt cardiovascular disease, metabolic syndrome, and neurodegeneration, which makes them an attractive early target for aging biology. Yet no one had built a harmonized, cross-tissue view of how human endothelial gene expression changes through adulthood. This study fills that gap and offers a baseline for detecting and eventually treating vascular aging.

Dobner and colleagues assembled the Human Aging Endothelial Cell Atlas (HAECA) from 34 publicly available single-cell and single-nucleus RNA-sequencing studies, screened from 59 candidates. The final atlas holds 375,829 endothelial cells from 290 nondiseased donors across 12 tissues: adipose, brain, breast, heart, intestine, kidney, liver, lung, muscle, ovary, skin, and uterus. After anchor-based integration, cells were annotated as arterial, capillary, venous, lymphatic, or angiogenic. Conserved markers emerged, including SULF1 in arteries, RGCC in capillaries, ACKR1 in veins, and PROX1 in lymphatics.

Donors were split into young and aged groups using each tissue's median age, with pseudobulk DESeq2 analysis and a cross-tissue meta-analysis. Aged endothelium consistently gained inflammatory and senescence-associated secretory genes and TNF-α signaling. It lost angiogenesis genes such as APLNR, COL4A1, ID3, and PLVAP, along with oxidative phosphorylation and MYC-target programs. A second analysis using fixed cutoffs (50 or younger versus 60 or older) confirmed 287 age-increased and 135 age-decreased genes. Veins showed the strongest decline in angiogenic signatures, including ANGPT2, EDN1, and LAMA4. Aged arteries showed coagulation signatures, and angiogenic capillaries trended toward lower abundance.

The atlas also exposed extracellular matrix and mechanical-sensing changes. Aged liver, lung, skin, and adipose cells showed higher expression of glycosylation genes, including GALNT family members and sialyltransferases. Predicted activity of YAP, TAZ, and TEAD transcription factors rose in aged cells across tissues, most strongly in skin. MERFISH spatial transcriptomics of human skin (1,878 endothelial cells) confirmed the expression pattern. An independent skin dataset spanning ages 33 to 77 showed the strongest YAP–age correlation in endothelium. Immunostaining showed a modest but significant rise in cells with nuclear YAP in aged skin (P=0.0199; n=17 young, 13 aged).

For senescence, CDKN1A (p21) was widely expressed in endothelial cells, whereas CDKN2A (p16) was nearly absent. Neither marker's frequency differed by age. The authors intersected genes enriched in CDKN1A-positive atlas cells with genes from senescent HUVECs (late-passage or doxorubicin-treated). This yielded MAFF, ADAMTS1, ADAMTS4, JUN, SRGN, KLF2, TNFRSF10D, TNFRSF10B, and UGCG. The signature correlated with SenMayo, CoreScence, and CellAge scores. MAFF, KLF2, and STX12 rose in late-passage iPSC-derived brain-like endothelial cells (adjusted P<0.05), and candidates were enriched in progeria mouse models. Overexpressing MAFF in early-passage HUVECs raised p21 and SA-β-gal activity.

The work is descriptive and cross-sectional. Participants 80 and older were underrepresented, the datasets lacked sex, menopausal, medication, and comorbidity metadata, and CDKN1A may not be specific to senescence. CDKN1A-positive cells also showed angiogenesis pathways, suggesting some signals reflect transient remodeling rather than permanent arrest. Causality for YAP activation and venous angiogenic decline remains untested.