Writing Developmental History into DNA: 1.28 Million Mouse Cells Linked in an Embryonic Family Tree

How a fertilized egg grows into a body with different organs is a story of countless rounds of cell division and differentiation. Examining cells at a single moment in development can reveal their current identities, but makes it difficult to trace where they came from. Researchers have now enabled a mouse embryo to write this history into its own DNA, then read those records to reconstruct a family tree encompassing approximately 1.28 million cells, connecting the fertilized egg to cells in the later stages of organ formation.

The study, published in *Science* on October 8, 2026, was jointly led by Jay Shendure of the University of Washington and Chengxiang Qiu of Dartmouth College. The team used a genomic recording system called DNA Typewriter, allowing a fertilized egg to develop to day 13.5 before reading recording marks from approximately 1.58 million cell nuclei. The final lineage tree included 1,281,141 cells spanning 138 cell types. According to the Allen Institute and the research resource NextCell, this represents approximately one-tenth of the embryo’s total cells at that stage.

This “typewriter” uses prime editing tools to sequentially write short DNA symbols onto engineered recording tapes. Each cell carries 11 recording tapes, each with 6 positions. Existing marks are passed to daughter cells during division, while new marks are added in sequence. By comparing the symbols cells share and the order in which they appear, researchers can infer common ancestors and branching times, then combine these with gene expression data to connect cells’ origins with their final identities.

One key recording event occurred during the fertilized egg’s first division, allowing researchers to identify descendants of the first two daughter cells. The Allen Institute noted that the two branches contributed approximately 57% and 42% of the descendant cells included in the analysis, but produced similar proportions of cell types. At least in this embryo, a larger number of descendants was not accompanied by an equally pronounced bias toward particular tissue types.

Tracing backward along the branches, the team also estimated when different cell types entered more restricted developmental paths. Blood- and retina-related cells established their fates earlier, while cells forming the outer layer of skin did so later. These findings echo previous developmental research and show, within the lineage of a single embryo, that tissues do not all complete their division of labor at the same time, but instead form progressively along different timelines.

This picture still has clear limits. The team began with 100 fertilized eggs and ultimately obtained 10 embryos for examination. The detailed reconstruction focused on the one with the richest records, so the results cannot yet establish variation between embryos. NextCell also explicitly distinguishes lineage relationships from ancestral identities: relationships within the tree were reconstructed from the records, while ancestral cell types were estimated by a model, rather than directly measured at each time point in the past.

The researchers have made an interactive lineage tree, downloadable data, and analysis code available through NextCell. Sequencing data are deposited in the GEO database under accession number GSE341627. The value of this tool lies in establishing a traceable reference for normal development, which could in the future be used to study congenital defects, cancer cell expansion, or how engineered cells acquire their intended identities. Turning these applications into reliable methods for disease research will still require validation in more embryos and under more experimental conditions.