henhaobio
Sign in

bioRxivDevelopmental BiologyPreprint

An integrated single cell and spatial omics atlas of human prenatal development

Simone WebbiD, Antony RoseiD, Chuan XuiD, Lloyd SteeleiD, Masami A Kuri, Emily StephensoniD, Kemal İnecikiD, Daniyal J. JafreeiD, April Rose FosteriD, Daniela Basurto-LozadaiD, Nana-Jane ChipampeiD, Anna Vathrakokoili PournaraiD, Marc‐Antoine JacquesiD, Ken ToiD, Chloe AdmaneiD, Efpraxia KritikakiiD, Marta A. ChroscikiD, David HorsfalliD, Julia ForemaniD, Koen RademakeriD, Julia Karjalainen, Anna LaddachiD, Shaista MadadiD, John E. G. LawrenceiD, Vitalii KleshchevnikoviD, , Muzlifah A. HaniffaiD

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

Abstract

Single cell genomics has enabled analysis of human prenatal development at unprecedented resolution. However, most studies have relied on dissociated tissues during restricted windows of development, limiting insights into how spatially distributed networks of cells, and multicellular niches emerge and adapt to distinct organ microenvironments in situ. Moreover, existing human developmental atlases have not yet been harmonised, and we thus lack a comprehensive catalogue of known cell types in the developing human body. Here, we introduce the Human Developmental Cell Atlas (HDCA), a unified structural, cellular and molecular resource for prenatal human development. The HDCA integrates published and unpublished single cell/nucleus RNAseq atlases across prenatal organs, and includes a newly generated, spatially resolved, multimodal cell atlas of intact human embryos. Spanning 4-22 post conceptional weeks, capturing embryonic and early to mid fetal stages, the HDCA contains [~]4.6 million cells/nuclei which resolve into [~]450 cell types, explorable with a bespoke web portal. For a global overview of the human embryos multicellular communities, we applied unsupervised deep learning to our intact human embryo spatial data, charting 114 tissue niches that are structural and signalling hubs for the cellular interactions of the embryo. Guided by these niches, we profiled cellular networks over space and time, not examinable using single-organ atlases. In so doing, we revealed tissue-specific fibroblast patterning from previously undescribed mesenchyme progenitors, early diversification of organ-specific blood capillaries and lymphatic vasculature, emergence of neural crest cell fates, the formation of placode-and neural crest-derived peripheral sensory neurons, and how tissue niches guide peripheral neuron maturation and axonal migration. The HDCA thus serves as a comprehensive step towards a comprehensive understanding of human prenatal development, and a template towards unravelling the biology of congenital disorders.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (21 Apr 2026) · CC BY 4.0

Human Developmental Cell Atlas integrates ~4.6 million cells/nuclei from 4-22 post-conception weeks into 454 refined cell types and maps 114 spatial niches across 6PCW whole-embryo sections.

Earlier human developmental atlases mostly profiled single organs from different donors with lab-specific protocols, so distributed networks such as stroma, vasculature and the peripheral nervous system, and the in situ niches they form, stayed under-explored and cell type naming was inconsistent. The Human Developmental Cell Atlas harmonises community atlases with new whole-embryo single-cell and spatial data, adding reusable reference models and a disorder-gene portal.

Key results

  1. ShownA harmonised atlas of 186 samples spanning 4-22 PCW resolves ~4.6 million cells/nuclei into a 4-tier annotation hierarchy and serves as a mapping reference for external developmental data.

    186 samples: 183 community samples across 21 organs plus 3 new 6PCW embryos dissected into 12 sections; tiers of germ layers, 12 systems, 69 broad and 454 refined types (NMI 0.85 vs author labels). scArches mapping of Cao et al (~4.1 million cells) shared 72 of 77 broad groups and refined annotation for 44 cell types. · Integrated single cell and spatial omics atlas…; A comprehensive, adaptable human developmental…; HDCA: Integration and annotation harmonisation…

  2. ShownUnsupervised NicheCompass niche detection on whole-embryo Xenium data charted 114 niches. These include structures that H&E cannot distinguish, such as thyroid, parathyroid and ultimopharyngeal body progenitor niches.

    114 niche clusters from 8 Xenium sections (3.1 million cells): 37 focal (32 H&E-matched, 5 novel), 30 CNS, 47 distributed. Ultimopharyngeal niche showed ISL1, ASCL1, NKX2-1 with high CALCB but low CALCA. Reference-query mapping of 3 further sections gave 90 niches, including a retinal niche (ATOH7, PRDM13) absent from the reference. · An unsupervised systematic decomposition of the…; Histologically discernible and novel tissue…

  3. ShownThe authors introduce scTRAM, a benchmark of trajectory fidelity that scores topology, cell ordering and lineage-manifold dominance against curated lineage graphs. Used together with scIB, it selected scVI for atlas integration.

    Combined scIB+scTRAM score: scVI 0.494, scPoli 0.437, TarDis 0.421; PCA, Harmony and Scanorama lower. scVI led mean scTRAM by trajectory group and beat scPoli on most informative edges (Benjamini-Hochberg 5% FDR) across haematopoietic (57 types, 58 edges), germline (8, 7) and peripheral nervous system (12, 11) graphs. · Benchmarking atlas integration with trajectory…; scTRAM: Trajectory-specific fidelity evaluation; Edge-resolved and lineage-stratified analysis

  4. ShownOrgan-specific capillary endothelial programs (brain, heart, liver) are already present in the first trimester, and lymphatic endothelial heterogeneity is established early, including an early FOXO1+CALCRL+ lymphatic endothelial cell (LEC) state predicted to express PIEZO1.

    94,681 endothelial cells (ECs) in 37 types; adult organ gene-program scores detectable by 6PCW, supported by label transfer to 163,973 Xenium ECs from 11 sections (SLC7A5, POSTN, OIT3). 3,932 LECs in 7 subsets; capillary and early LECs in 9 of 10 head niches; RNAscope at 5-6PCW showed PIEZO1 in PROX1+FLT4+ cells near the cardinal vein. · Organ-specific blood endothelial specialisation…; Lymphatic endothelial heterogeneity during human…; Spatiotemporal expression of genes associated…

  5. SuggestedEarly embryos contain 10 FRZB+ fibroblast progenitor populations (some also HIC1+); apart from a broadly distributed PDGFRA+ subset they are regionally patterned, including ZIC2-high scalp, meningeal and spinal subsets that reappear in iPS-derived skin organoids.

    Joint scRNA-seq/Xenium integration of ~3.6 million predominantly mesenchymal cells yielded 108 types in 12 groups; progenitors were focal (RUNX2+ skeletal, IRX3+ISL1+ limb, FOXF1+ gut/anterior body, 4 ZIC-high in scalp, primary meninx, spine). Organoid fibroblasts mapped mainly to progenitor, skin and skeletal groups, recovering all 4 ZIC-high subsets. · Mesenchymal progenitors in the human embryo; Decoding cranial embryo niches; Novel mesenchymal progenitors are recapitulated…

Methods, data and reproducibility

Community 10x sc/snRNA-seq atlases were harmonised (IDTrack gene mapping, batch-aware variable genes) and integrated with new grid-dissected 6PCW whole-embryo scRNA-seq via scVI, chosen by scIB plus scTRAM benchmarking; Xenium, Visium and multiome embryo data supported NicheCompass niche detection and analyses of mesenchyme, endothelium and peripheral nervous system.

  • Published and shared community 10x Chromium sc/snRNA-seq atlases of 21 developing organs, 4-22 PCW · human · n = 21 datasets; 183 samples; 3,652,263 cells
  • New 6PCW whole-embryo scRNA-seq, each embryo grid-dissected into 12 anatomical regions (10x 5' v1/v2 chemistry) · human · n = 3 embryos; 1,029,326 cells after QC · E-MTAB-11504, E-MTAB-11520, E-MTAB-11911 (private until publication)
  • Single-nucleus multiome (snRNA-seq plus snATAC-seq) from leftover frozen cells of the same whole embryos · human · n = 467,575 snRNA cells (raw); 521,423 snATAC cells (filtered) · E-MTAB-16757 (private until publication)
  • Xenium Prime 5,000-gene panel sagittal sections of 6PCW embryos SOB25 (Carnegie stage 16) and SOB26 (Carnegie stage 17) · human · n = 11 sections (8 core, 3 query); 3.1 million cells in core set · BioImage Archive S-BIAD2878 (private until publication)
  • Visium CytAssist FFPE sagittal sections of embryo SOB26 with HDBR H&E region annotations · human · n = 2 sections; 7,778 spots analysed (section_121) · E-MTAB-14785 (private until publication)
  • Cao et al. sci-RNA-seq3 mid-gestation atlas (10-18PCW) mapped as an external query · human · n = ~4.1 million cells and nuclei; 77 broad groups
  • Published iPS-derived skin organoid scRNA-seq mapped onto the HDCA reference · human (iPS-derived organoid)
  • Published single-cell data of lymphatic and associated endothelial cells at onset of lymphatic emergence · mouse; zebrafish
Methods & tools
scVI (scvi-tools v0.19.0), scTRAM (trajectory-fidelity benchmark introduced here), scIB metrics; baselines scPoli, TarDis, Harmony, Scanorama, PCA, NicheCompass, CellTypist v1.4.0, scArches, IDTrack, cell2fate v1.0.0 with Velocyto v0.17.17, LIANA+ v1.5.0 (CellPhoneDB-derived p-values), SnapATAC2 v2.8.0 with MACS3, CellRanger v6.0.1, Scrublet v0.2.3, Souporcell v2.4, CellBender v0.2.0, 10x Xenium Prime, Visium CytAssist, RNAscope
Software & environment
Scanpy v1.7.2 and v1.11.2, scvi-tools v0.19.0, CellTypist v1.4.0, CellRanger v6.0.1, cellranger-arc v2.0.2, SpaceRanger v2.1.0, SnapATAC2 v2.8.0, LIANA+ v1.5.0, decoupleR v2.1.1, cell2fate v1.0.0; GRCh38-2020-A; Ensembl release 107 gene symbols
Assay conditions & compute
cell2fate trained on an NVIDIA A100-SXM4-80GB GPU for 200 epochs; work supported by de.NBI Cloud infrastructure
Reproducibility gaps
The haniffalab/hdca-paper code repository, all ArrayExpress and BioImage Archive accessions, and the Cherita portal are private until publication; 3 of 21 community datasets were obtained through internal communications.
Limitations (5)
  • Spatial profiling covered only 6PCW embryos; the authors say sequential timepoints (e.g., 8PCW and 10PCW) and broader sampling are needed to capture transient microenvironments and full niche composition across the embryo. · stated by the authors
  • Endocrine tissues and the period between implantation and 4PCW remain undersampled; mapping a pre-implantation dataset aligned only along the pluripotent-to-endoderm axis, primordial germ cells and gonads. · stated by the authors
  • The conclusion that the human trigeminal ganglion is purely neural crest-derived cannot rule out a transient placodal contribution before the sampled stages; lineage-tracing and functional studies are needed. · stated by the authors
  • The benchmark may favour autoencoders: a 4,000-configuration scVI sweep guided scVI, scPoli and TarDis settings while Harmony and Scanorama ran at defaults, and scTRAM ground truth covers only 3 curated lineages. · noted in this summary
  • New spatial data come from embryos SOB25 and SOB26 (Carnegie stages 16 and 17), and lineage origins (placode vs neural crest, neural crest-derived mesenchyme) rest on expression, velocity and co-location rather than human lineage tracing. · noted in this summary

Read this if you need a harmonised human prenatal single-cell reference with label- and niche-transfer models, or study embryonic stroma, vasculature or peripheral nervous system development in situ.

Citations
6

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
THISXu '23Braun '23Ding '21Eze '21He '22Suo '22
76% semantic matchDevelopmental Biology
Yichi Xu, Tengjiao Zhang, Qin Zhou+7 more·Nature Cell Biology·Mar 16, 2023· 160 citations