henhaobio
Sign in

CellDevelopmental BiologyPeer-reviewed journal article

Comprehensive Lineage Tracing Maps the Landscape of Cell Fate Decisions in Mouse Embryogenesis

William N. ColganiD, Luke W. KoblaniD, JoAnne VillagranaiD, Tien-Chi Jason HouiD, Minming WangiD, Gokul GowriiD, Whitney Chandler, Leonardo A. SepúlvedaiD, Didar CiftciiD, Karina Smolyar, Kathryn E. YostiD, Alicia Young, Lars WittleriD, Styliani MarkoulakiiD, Kyle M. LohiD, Xiaowei ZhuangiD, Nir YosefiD, Zachary D. SmithiD, Jonathan S. WeissmaniD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

A comprehensive cell fate map of mammalian embryogenesis has remained out of reach given the scale, cellular diversity, and non-deterministic nature of development in utero. Here, we use PEtracer to continuously install heritable genetic marks as development progresses, reconstructing lineage trees that resolve ~75% of cell divisions across >1.4 million cells from 16 replicate embryos. We pair these trees with deep transcriptional profiling to resolve cell fate biases, restriction timing, progenitor pool sizes, and lineage relationships throughout embryogenesis. Using this quantitative reference, we uncover strikingly reproducible lineage architecture between replicate embryos and chart the lineage dynamics driving fate specification across diverse tissues. Specific biological insights include the progressive restriction of neural crest fate, the relative contributions of distinct mesodermal origins to endothelium, and the dynamics of axial elongation. This work provides a foundation for a quantitative and predictive understanding of mammalian development.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (3 Oct 2026) · CC BY 4.0

Prime-editing lineage trees for 16 mouse embryos (E7.5 to E10.0, >1.4 million cells) quantify fate restriction timing, progenitor numbers and lineage relationships, and are reproducible across replicate embryos.

Reconstructing fate restriction in mammalian embryos needs continuous lineage recording, enough capacity to resolve millions of divisions, and cell state measured in the same cells. The authors pair PEtracer with whole-embryo sampling to resolve ~75% of divisions, which they say captures late commitment events that subsampled datasets systematically miss.

Key results

  1. ShownWhole-embryo sampling gave high-quality joint lineage and transcriptome data for a large atlas, with most branches carrying at least one edit.

    1,792,520 high-quality cells from 16 embryos; trees for 1,447,839 donor cells; 94% lineage mark detection, 98% of cells with a unique edit profile, ~75% of branches marked by ≥1 edit; 99 cell types and 142 subtypes resolved. · Abstract; Comprehensive cell state and lineage profiling…

  2. ShownLineage architecture is highly reproducible across replicate embryos, and reconstruction error looks small.

    Pairwise ancestral linkage differed by <6 hours across triplicate E9.5 embryos; cell type abundances correlated between E9.5 replicates (R² = 0.92 and 0.98); germ layer mismatch rate ~1% (0.97% mean at E9.5); clade statistics robust to simulated error and jackknife. · Quantifying germ layer cell fate restriction…; Lineage architecture of cell type…; Quantifying germ layer mismatch; Figure legends

  3. ShownGerm layer fate bias appears early and restriction timing differs by layer, with posterior cell types staying closer to uncommitted progenitors.

    Weak but significant bias from ~E4.0 (permutation test, 100 permutations); fate-restricted clades defined as >90% of descendants in one germ layer; endoderm restricted earliest (~E7.0), ectoderm and mesoderm later and over a broader interval; restriction timing correlated strongly with linkage to uncommitted progenitors. · Quantifying germ layer cell fate restriction…; Calculating cell fate bias; Identifying fate-restricted clades; Ancestral linkage analysis

  4. ShownEndothelial cells arise from several mesodermal sources in parallel, and the cells keep a faint transcriptional trace of their origin.

    20,053 endothelial-restricted clades; origin inferred from sibling clades: lateral plate 34.3%, head 22.3%, paraxial 14.5%, extraembryonic 11.8%, cardiac 10.6%, intermediate 2.2%; proportions consistent across three E9.5 replicates; paraxial-derived cells retained somite identity program (P45) expression. · Multiple parallel developmental origins of…

  5. ShownTrunk neural crest (NC) clades show no deviation from independent DRG/autonomic assortment, consistent with fates not being prespecified at NC specification but resolving over time. The axial progenitor pool, by contrast, is a mixture of fate-biased clades, with bipotent clades less frequent than expected.

    Among 3,404 posterior NC-restricted clades, those with differentiated outputs at E9.5 were 42.9% dorsal root ganglion (DRG)-biased, 30.7% autonomic-biased and 26.3% bipotent. After size-stratified permutation the bipotent fraction matched independent assortment. Axial progenitor sibling clades were 37.0% ectoderm-biased, 41.9% mesoderm-biased and 21.1% bipotent, less than expected under independence. · Fate restriction in the neural crest; Dynamics of posterior axis elongation; Quantifying neural crest fate restriction…; Quantifying axial progenitor bias

Methods, data and reproducibility

Engineered mouse embryonic stem cells carrying the PEtracer prime-editing lineage recorder (>100 editable sites) were used to make chimeric embryos, which were fully dissociated and profiled by scRNA-seq at half-day intervals. Trees were built with a hybrid greedy plus FastTree neighbor-joining method, benchmarked on simulations, and analysed with fate bias, clade and ancestral linkage metrics.

  • Whole-embryo scRNA-seq with lineage readout from chimeric embryos, E7.5 to E10.0 at half-day intervals (triplicate except E10.0) · mouse · n = 16 embryos; 1,792,520 high-quality cells (1,447,839 donor cells in trees) · GEO GSE342607
  • Published reference atlases for transcriptional staging and label transfer (Pijuan-Sala et al., 2019; Qiu et al., 2024) · mouse
  • Published Stereo-seq spatial transcriptomics, sagittal section of an E13.5 embryo, used to train anterior-posterior and dorsal-ventral position predictors · mouse · n = one section
  • Long-read whole-genome sequencing of the engineered mESC line to map lineage tracing cassette integrations · mouse · n = one line with 36 integrated cassettes
Methods & tools
PEtracer, Cell Ranger v9.0.1, FastTree 2 v2.1.11, ConvexML v1.0.0, Cassiopeia v3.0.0, LAML-Pro v0.1.0, scVI (scvi-tools v1.4.0), Scanpy v1.11.5, Hotspot v1.1.3, geomloss, minimap2 v2.28, Claude Opus (annotation aid)
Reagents & environment
Versions stated: Cell Ranger v9.0.1, Scanpy v1.11.5, scvi-tools v1.4.0, Hotspot v1.1.3, Cassiopeia v3.0.0, FastTree v2.1.11, ConvexML v1.0.0, minimap2 v2.28. No container or requirements file mentioned.
Reproducibility gaps
Engineered mESC lines are available only from the lead contact under a materials transfer agreement; further reanalysis information is available on request. Plasmids are deposited at Addgene.
Limitations (6)
  • Trees are an imperfect record: unmarked divisions and incomplete sampling can obscure progenitor-progeny links, and branch lengths assume edit accumulation is roughly proportional to elapsed time. · stated by the authors
  • Cell type annotations are limited by current references and possible residual doublets; spatial coordinates were imputed, not measured. · stated by the authors
  • Trees record observed fates, not developmental potential; cells may be able to adopt other fates in a different environment. · stated by the authors
  • Embryos are chimeras seeded by 2-6 donor mESCs, so the earliest fate decisions should be read in this engineered context; subtle effects of tracing cannot be fully excluded. · stated by the authors
  • All trees come from one engineered donor line at one editing kinetics, and the spatial position models were trained on an E13.5 section, a later stage than the profiled embryos. · noted in this summary
  • Cell type and gene program annotations used a large language model alongside manual curation; the program annotations are described as provisional and not experimentally validated. · noted in this summary

Read this if you work on lineage tracing, fate mapping or tree-based analysis of single-cell data and want a quantitative reference of mouse fate restriction with open code and data.

Citations
7
FWCI
14.47
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISLiu '26Chan '19Wang '26Guo '26Qiu '21Sendra '24
77% semantic matchDevelopmental Biology
Yunzheng Liu, Jihye Park, Laura Amaya+17 more·bioRxiv·Oct 12, 2026