Loss of cell-identity DNA methylation tracks division history, not DNA damage
- 1Tel Aviv University
Abstract
One influential proposal holds that DNA damage drives aging by degrading the epigenome, so that a cell loses the information that specifies what it is. We test that specific proposed driver, not the broader idea that aging involves a loss of epigenetic information. Cell-type-unique CpGs are a direct place to test it. They are the sites where a cell's methylome differs most from every other cell type, so they are where its identity information can be accurately measured. We built such a panel for human skin fibroblasts against twenty primary cell types, and found that most of its 9,150 CpGs behave as a record of proliferation rather than identity. Their methylation state in fibroblasts is shared with continuously proliferating cell lines, consistent with the bulk hypomethylation of late-replicating chromatin that accumulates in culture, and this layer must be separated from identity marks before identity loss can be tested. We then expressed every perturbation as the fraction of the panel's identity information lost beyond baseline-matched background CpGs, using a Bernoulli cross-entropy between the fibroblast code and a generic comparator cell. In 24 matched fibroblast cultures, replicative exhaustion cost 24.3% of the identity information (95% CI 20.2 to 28.7) and RAS activation cost 5.4% (4.6 to 6.1), both at p = 1.2 X 10^-7;. A senescence-inducing dose of X-rays cost nothing: -0.5% (-0.7 to -0.2), with identity CpGs if anything moving slightly further from other cell types than matched background did. The design could detect a loss of 0.4%, about one sixtieth of the exhaustion effect, so the X-ray result is a bounded negative rather than an absence of power. Doxorubicin-induced and irradiation-induced senescence did not move the panel either. Clones that kept dividing after irradiation moved further from other cell types rather than collapsing toward them. In human fibroblasts, loss of identity information tracks prolonged division, not acute DNA damage.
Related
- DNA methylation defines lineage identity yet converges among functionally related cells and is cell‑specifically remodeled during in vitro passaging.
- The physical state of the nucleus tracks its function across eukaryotes
- Parameter-Based Prioritization of Time-Resolved LC–HRMS Ion Tracks for Untargeted Reaction Monitoring
- Sequential division of labor between PAXX and XLF drives NHEJ synaptic complex stability and maturation
- Mitotic assembly of respiratory supercomplexes ensures accurate cell division and fuels tumor growth
BibTeX
@misc{sagy2026loss,
title = {{Loss of cell-identity DNA methylation tracks division history, not DNA damage}},
author = {N. Sagy and N. Benedek and D. Z. Bar},
howpublished = {bioRxiv},
year = {2026},
month = nov,
doi = {10.64898/2026.11.03.763914},
url = {https://www.biorxiv.org/content/10.64898/2026.11.03.763914v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Loss of cell-identity DNA methylation tracks division history, not DNA damage
AU - N. Sagy
AU - N. Benedek
AU - D. Z. Bar
PY - 2026
DA - 2026/11/03
PB - bioRxiv
DO - 10.64898/2026.11.03.763914
UR - https://www.biorxiv.org/content/10.64898/2026.11.03.763914v1
AB - One influential proposal holds that DNA damage drives aging by degrading the epigenome, so that a cell loses the information that specifies what it is. We test that specific proposed driver, not the broader idea that aging involves a loss of epigenetic information. Cell-type-unique CpGs are a direct place to test it. They are the sites where a cell's methylome differs most from every other cell type, so they are where its identity information can be accurately measured. We built such a panel for human skin fibroblasts against twenty primary cell types, and found that most of its 9,150 CpGs behave as a record of proliferation rather than identity. Their methylation state in fibroblasts is shared with continuously proliferating cell lines, consistent with the bulk hypomethylation of late-replicating chromatin that accumulates in culture, and this layer must be separated from identity marks before identity loss can be tested. We then expressed every perturbation as the fraction of the panel's identity information lost beyond baseline-matched background CpGs, using a Bernoulli cross-entropy between the fibroblast code and a generic comparator cell. In 24 matched fibroblast cultures, replicative exhaustion cost 24.3% of the identity information (95% CI 20.2 to 28.7) and RAS activation cost 5.4% (4.6 to 6.1), both at p = 1.2 X 10^-7;. A senescence-inducing dose of X-rays cost nothing: -0.5% (-0.7 to -0.2), with identity CpGs if anything moving slightly further from other cell types than matched background did. The design could detect a loss of 0.4%, about one sixtieth of the exhaustion effect, so the X-ray result is a bounded negative rather than an absence of power. Doxorubicin-induced and irradiation-induced senescence did not move the panel either. Clones that kept dividing after irradiation moved further from other cell types rather than collapsing toward them. In human fibroblasts, loss of identity information tracks prolonged division, not acute DNA damage.
ER -