Every Cell Typed Its Own Family History

Researchers have traced the ancestry of 1.28 million cells in a developing mouse back to the fertilized egg's first division, using a DNA recorder they call DNA Typewriter, according to a study in Science.

The paper, published on October 8, 2026, is titled "A DNA Typewriter records the cell lineage history of a mouse, from zygote to late organogenesis." Its co-first authors are Qi Yu, Haedong Kim and Sophie Seidel, and its co-senior authors are Jay Shendure of the University of Washington and Chengxiang Qiu of Dartmouth College ^[1]. The tree follows the embryo through 13.5 days of development, by which point it had formed major body structures and organs ^[1][2][3]. A full mouse pregnancy lasts 19 to 21 days ^[2].

It is the largest cellular family tree yet built for a mammal, according to the Allen Institute, which took part in the work ^[3].

A recorder inside the egg

The researchers did not watch the embryo grow. They planted a recorder in the very first cell, let development run, then read what the cells had written down.

The recorder is a stretch of engineered DNA. A tool called a prime editor writes short genetic marks onto it, one after another. Each new mark closes the active site where it was written and opens the next one, much as a typewriter carriage moves on after every keystroke.

The marks therefore stay in the order they were made, and they are copied into every daughter cell ^[1][3].

Two cells that carry the same early marks must share an ancestor. Cells that diverge in their later marks branched apart later. Read enough of them and the branching history can be inferred ^[1][3]. The paper's abstract says the reconstructed tree is supported by parsimony, a method that favors the simplest lineage history consistent with the marks observed ^[2].

Shendure and Junhong Choi invented DNA Typewriter ^[1]. According to Nature and the University of Washington release, earlier gene-editing lineage recorders could damage cells or leave marks whose order was hard to work out. Ordered marks are what make a long history readable ^[1][2].

One embryo out of 100 eggs

The team tried the procedure in 100 fertilized eggs and examined 10 resulting embryos. It then put its full analysis into embryo No. 3, because the recorder was clearly active in it and it held the richest history ^[1].

The Allen Institute says the team measured genetic marks in 1.58 million individual cell nuclei. The lineage tree covers about 10% of the cells in the two-week-old embryo, and Nature puts the number of cells placed in it at about 1.3 million ^[2][3].

That makes this a demonstration of what the method can record. It is not a complete map of mouse development, and it says nothing yet about human pregnancy.

What the tree shows

The first division of the fertilized egg produced two founding cells, and their marks let the team tell their descendants apart. One lineage contributed more cells in total, about 57% against 42% for the other. Yet the two lineages produced similar proportions of the different cell types ^[1][3]. Why one founder out-multiplied the other is an open question in the data.

The tree also gave timing. Blood cells and retinal cells appeared to commit to their identities relatively early, and the outer layer of skin did so later. The University of Washington release says these results agree with earlier research, though the lineage measurements here came from a single mouse ^[1].

Why a worm still holds the record for clarity

The benchmark for this kind of work is a microscope map. John Sulston followed every cell division in the nematode Caenorhabditis elegans, a small transparent worm, and charted the lineage of its 959 somatic cells, as Nature describes it ^[2].

The worm lets itself be watched. A mammalian embryo does not. It develops out of sight, it is far larger in cell count, and outside cues can influence what a cell becomes, which makes simple inheritance a poor predictor of fate ^[2]. A molecular recorder gets around the problem of watching. Each cell carries its own diary.

What Sulston could do by eye for 959 cells, the Typewriter has now done by sequencing for more than a million. The mouse tree covers only about a tenth of the embryo, and the worm map was complete. The mouse tree still has no precedent in size.

What it may be for

Shendure said the team wants to understand how early ancestry influences later cell fates. He said lineage research could eventually advance understanding of congenital malformations, neurodevelopmental conditions, genetic disorders and cancer ^[1]. The authors and institutional releases also point to stem-cell engineering as a possible use ^[1][3].

Those are prospects. The experiment did not test any of them, and a mouse lineage map is not evidence about what causes human pregnancy loss or birth defects.

What the map offers is a reference. Researchers can ask when a given cell type splits from its neighbours, and later compare normal development with disease.

My reading is that the real advance is the recorder, not the single tree. A tree from one embryo, chosen because its recorder worked best, is a proof of method. The test comes when the same system runs across many embryos, including ones where the recorder is less active. Only then will anyone know which patterns, such as the 57-to-42 split, are general and which belong to embryo No. 3.

The data are open to anyone. The Allen Institute says the lineage tree is publicly available through an interactive browser called NextCell ^[3].