A single-cell atlas of pig gastrulation as a resource for comparative embryology

Abstract

Cell-fate decisions during mammalian gastrulation are poorly understood outside of rodent embryos. The embryonic disc of pig embryos mirrors humans, making them a useful proxy for studying gastrulation. Here we present a single-cell transcriptomic atlas of pig gastrulation, revealing cell-fate emergence dynamics, as well as conserved and divergent gene programs governing early porcine, primate, and murine development. We highlight heterochronicity in extraembryonic cell-types, despite the broad conservation of cell-type-specific transcriptional programs. We apply these findings in combination with functional investigations, to outline conserved spatial, molecular, and temporal events during definitive endoderm specification. We find early FOXA2 + /TBXT- embryonic disc cells directly form definitive endoderm, contrasting later-emerging FOXA2/TBXT+ node/notochord progenitors. Unlike mesoderm, none of these progenitors undergo epithelial-to-mesenchymal transition. Endoderm/Node fate hinges on balanced WNT and hypoblast-derived NODAL, which is extinguished upon endodermal differentiation. These findings emphasise the interplay between temporal and topological signalling in fate determination during gastrulation.

Fig. 1. Overview of the pig single-cell atlas.

a Onset and duration of primary and secondary gastrulation in humans, mice, pigs and monkeys. Timepoints where high-resolution single-cell datasets are available, are marked for each species^2,6,28 as well as the time points covered in this atlas. Numbers indicate embryonic day. b Diagrammatic representations of the earliest and latest embryo samples in this dataset with visible embryonic structures/cell types labelled. Epi Epiblast, PS Primitive streak, NM Nascent mesoderm, Hyp Hypoblast, TB Trophoblast, NT Neural tube, Phar Pharyngeal arches, Som Somites, PosNP Posterior neuropore, AntNP Anterior neuropore, YS yolk sac. c Uniform manifold approximation and projection (UMAP) plot showing atlas cells (91,232 cells, 23 biologically independent samples). Cells are coloured by their cell-type annotation and numbered according to the same legend as d below. d Stacked area plot showing the fraction of each cell type at each time point, a progressive increase in cell-type complexity can be seen across time points with mesodermal cell type diversification preceding that of ectoderm. APS Anterior Primitive Streak, ExE extra-embryonic, PS Primitive streak, NM Nascent mesoderm, HE Hematoendothelial, PGC Primordial germ cells.

Fig. 2. Alignment of Pig, Mouse and Monkey datasets.

a UMAPs showing E6.5–8.5 mouse embryo cell types² and Pig E11.5 to E15 with mouse annotations after reciprocal PCA-based projection onto the mouse dataset. b Heat map showing the percentage of pig cells in each stage allocated to a particular mouse stage after label transfer. E Embryonic day. c UMAPs showing E20-29 monkey embryo cell types⁶ and Pig E11.5 to E15 with mouse annotations after reciprocal PCA-based projection onto the monkey dataset. d Heat map showing the percentage of pig cells in each stage allocated to a particular monkey stage after label transfer. A percentage of 100 would indicate that all cells of a given cell type were predicted to be analogous to the cell identity in the queried organism. e UMAPs showing the aligned monkey, mouse and pig datasets with pig cell type and subtype annotations.

Fig. 3. Heterochrony across Pig, Mouse and Monkey differentiation.

a Heat maps showing the percentage of human mesodermal cells⁷ allocated to a pig or mouse² cell identity after label transfer. b As with a, except with endodermal cell types. 100% would indicate that all cells of a given cell type were predicted to be analogous to the cell identity in the queried organism. E Embryonic day.

Fig. 4. Endodermal progenitors do not undergo classical EMT.

a UMAP plot showing epiblast, PS, APS/node, nascent mesoderm and DE clusters (24,874 cells; 23 biologically independent samples; E11.5-E15) coloured by global cell-type annotation and developmental time points. b As with a, coloured by cell subtypes. c Stacked bar graphs showing the frequency of each subcluster within the subset shown in a & b at specific time points in development. E Embryonic day. d Heat map illustrating the scaled expression of genes within individual cells. Expression of selected markers was used to identify cell subclusters as well as epithelial and mesenchymal marker genes.

Fig. 5. Endoderm forms from Epiblast-like low TBXT progenitors.

a UMAP plot with reversed graph embedding trajectories projected on top using Monocle3. Black nodes mark trajectory branching points. n = 16757 cells across 11 biologically independent samples. b UMAP plot showing predicted cell fates inferred from Monocle3. c Bar graphs showing the NANOG, TBXT and FOXA2 expression in lineage-fated cells from a&b. Data are presented as mean values +/- SEM. d Box plots showing epiblast, nascent mesodermal and endodermal lineage scores in selected clusters from a. n = 2340, 263, 1536 and 1140 cells respectively across 11 biologically independent samples. Centre line represents median, minima and maxima hinges represent the 25^th and 75^th percentiles respectively. Whiskers extend from the quartiles to the last data point that is within 1.5 times the interquartile range. Points beyond this range are shown and are considered outliers. P-values indicate the results of a two-sided Mann-Whitney U test. e Volcano plots showing differential expression between differently fated cells. Primitive streak (mesoderm fated) vs APS fates and Endodermal vs node-fated cells. Cut-off criteria for significant DEGs was a Log2 fold change ≥0.5 and an adjusted p value ≤ 0.01. f Heat map illustrating the scaled average expression of selected genes in each of the cell fates identified in b. g UMAP plots showing cells categorised by FOXA2, NANOG, TBXT and SOX17 expression at selected time points. F FOXA2, N NANOG, T TBXT, S SOX17 cells. Cells are coloured by their F/N/T/S category. APS Anterior primitive streak, E Embryonic day.

Fig. 6. FOXA2 and TBXT domains are spatially separated.

a Maximum intensity projection (Dorsal view) of E10.5 (n = 1) to E11.5 (n = 4) porcine embryos showing TBXT and SOX2 expression. E11.5 embryos are ordered left to right by age. E Embryonic day. b Single z-slice of the embryos shown in a showing FOXA2 and TBXT expression. Scale bar: 50 µm. c In Silico representations of embryos following 3D segmentation of embryos from a and b. d Axial and lateral reconstructed sections of embryos stained for FOXA2 and TBXT. Epiblast layer is oriented above the hypoblast/DE layer. White arrowheads indicate FOXA2 + TBXT- cells that are spatially separated from the TBXT domain. Scale bar: 50 µm. e Lateral sections of E11.5 (n = 1) and E12 (n = 1) embryos, showing expression of NANOG, FOXA2 and TBXT. Epiblast layer is oriented above the hypoblast/DE layer. White arrowheads indicate NANOG+ cells. Scale bar: 50 µm. Please refer to Supplementary Fig. 21 for a colour blind friendly version of the figure.

Fig. 7. Effect of Activin A and WNT signaling in pig EpiSC and hESC.

a Representative images depicting differentiation conditions for EDSCL4. Images were captured using an Operetta CLS high-throughput microplate imager. Scale bar: 200 µm. b Box plots summarizing 2D differentiation experiments. Data is normalised to well background signal. n = 3 independent experiments. Centre line represents median, minima and maxima hinges represent the 25^th and 75^th percentiles, respectively. c Proposed model of epiblast-DE differentiation in pig embryos. Please refer to Supplementary Fig. 22 for a colour-blind-friendly version of the figure.