The emergence of human primordial germ cell-like cells in stem cell-derived gastruloids

Abstract

Most advances in early human postimplantation development depend on animal studies and stem cell-based embryo models. Here, we present self-organized three-dimensional human gastruloids (hGs) derived from embryonic stem cells. The transcriptome profile of day 3 hGs aligned with Carnegie stage 7 human gastrula, with cell types and differentiation trajectories consistent with human gastrulation. Notably, we observed the emergence of nascent primordial germ cell-like cells (PGCLCs), but without exogenous bone morphogenetic protein (BMP) signaling, which is essential for the PGCLC fate. A mutation in the ISL1 gene affects amnion-like cells and leads to a loss of PGCLCs; the addition of exogenous BMP2 rescues the PGCLC fate, indicating that the amnion may provide endogenous BMP signaling. Our model of early human embryogenesis will enable further exploration of the germ line and other early human lineages.

Fig. 1.. Self-organization of hESCs into 3D gastruloids establishes primary germ layer derivatives.

(A) Experimental plan showing the generation of the hG model from hESCs in vitro. Ch, Chiron (CHIR99021); Act, Activin A; E8, Essential 8 Medium; RB27, Advanced RPMI Medium supplemented with B27. H, hours. (B) IF staining of transiently induced hESCs showing pluripotent markers stained for POU5F1 (OCT4), SOX2, and NANOG (top); mesendoderm markers EOMES, SOX17, and BRACHYUARY (BRA) (bottom); and GATA3 and pSMAD1/5/8, which are negative (middle). Scale bars, 50 μm (n = 15 samples from three experiments). (C) Formation of the three germ layer derivatives on hGs from D2 to D8 derived from RUES2-GLR hESCs. Reporter colors represent ectoderm (SOX2-mCitrine), mesoderm (BRA-mCerulean), and endoderm (SOX17-tdTomato). Scale bars, 250 μm (D2, N = 1620; D4, N = 1100, n = 10 experiments). (D and E) Representative confocal micrographs showing the primary germ layer derivatives (endoderm, ectoderm, and mesoderm) at D4 hG sections derived from W15-tdTomato hESCs stained for SOX17, SOX2, and BRA (or BRA), respectively. Scale bars, 200 μm (C) and 100 μm (D) (n = 10 samples from three experiments). (F) Quantification of multicellular lineages detected in D4 hGs [segmented from 39,822 cells (nuclei) from five images]. (G) Projection of D4 hGs representing multicellular lineages in (F). Cyan, ectoderm (SOX2⁺); yellow, endoderm (SOX17⁺); magenta, mesoderm (BRA⁺).

Fig. 2.. Transcriptional characterization of hGs.

(A) Uniform manifold approximation and projection (UMAP) plots showing the clustering and cell type annotation of hGs sampled on D0, D2, D3, D4, and D8 obtained by 10× scRNA-seq. (B) Proportion of cell types detected in hGs at different time points of in vitro development from D0 to D8. (C) Plot showing the prediction score based on the mapping of cell types detected in CS7 human gastrula and hGs at different time points. (D to F) Coembedding in a UMAP plot of the single-cell transcriptomes from a CS7 human gastrula and D3 hGs. Endo, endoderm; Ecto, ectoderm; EM, early mesoderm; PM, paraxial mesoderm; IM, intermediate mesoderm; MM, mixed mesoderm; AxM, axial mesoderm; YSM, yolk sac mesoderm; EmM, emergent mesoderm; NM, nascent mesoderm; Epi, epiblast; Ery, erythrocytes; HEP, hemogenic endothelial progenitors.

Fig. 3.. PGCLCs are detected in the absence of exogenous BMP in hGs.

(A) Microscopic images showing the expression of human PGCLCs in the absence of exogenous BMP, as shown by the NANOS3 reporter in hGs at different time points (n = 150 from five experiments). Scale bars, 50 μm. (B to D) Confocal micrographs showing key hPGCLC markers NANOS3 (RFP), SOX17, and TFAP2C (B); POU5F1 (or OCT4) and BLIMP1 (or PRDM1) (C); and NANOS3 (RFP), SOX17, and TFAP2C (D) by IF in hG sections on D2, D3, and D8, respectively (n = 15 from three experiments). Scale bars, 50 μm (B), 100 μm (C), and 250 μm (D). (E) UMAP showing the detection of hPGCLCs in the PS and AMLC regions during early specification (D2 and D3 hGs). (F) Dot plot showing key hPGCLC markers detected in hGs. BMP4 is specifically enriched in AMLCs. (G to I) Diffusion components (DC) analysis showing the detection of TFAP2A and TFAP2C on (G) D2 hPGCLCs, (H) D3 hPGCLCs, and (I) D2/D3 combined PGCLC dataset. (J) Hierarchical clustering of in vivo PGCs of human and nonhuman primates with in vitro human PGCLCs (D2) from this study. Cy, cynomolgus monkey; hPGCs, human PGCs; w, in vivo developmental stages in weeks; ePGCs, early Cy PGCs; lPGCs, late Cy PGCs.

Fig. 4.. ISL1 KO hGs do not display PGCLC formation.

(A) Confocal micrographs showing the lack of amnion markers ISL1 and GATA3 in the ISL1 KO hGs. Scale bars, 50 μm (n = 10 samples from three experiments). Red arrowheads indicate AMLCs. (B) Live microscopic images showing the detection of the fluorescence reporter NANOS3-tdTomato in WT but absent in the ISL1 KO hGs. NANOS3-tdTomato is rescued in gastruloids treated with exogenous BMP supplementation. Scale bars, 250 μm (WT) and 100 μm (KO and KO + BMP) (n = 300 samples from five experiments). (C) Confocal micrographs showing the maximum intensity projection of PGCLC markers, including SOX17 and TFAP2C in WT, KO, and KO + BMP–supplemented hGs. Scale bars, 50 μm (WT and KO) and 25 μm (KO + BMP) (n = 30 samples from three experiments). Red arrowheads indicate hPGCLCs. (D) Flow cytometry analysis showing the quantification of PGCLCs and AMLCs in the WT, ISL1 KO, and KO + BMP–supplemented hGs (n = 3 samples from three experiments). One-way ANOVA with Tukey’s multiple comparisons test was performed for statistical analysis. P < 0.05 was considered significant. ****P < 0.0001, ***P < 0.0005, and **P < 0.005. (E) Confocal micrographs with maximum intensity projection of SOX17, POU5F1 (OCT4), and ISL1 showing the distribution of PGCLCs and AMLCs in hGs. PGCLCs are indicated by red arrowheads. AMLCs are shown in white circles. Scale bars, 50 μm (n = 15 samples from three experiments). (F) Live microscopic images showing the detection of the fluorescence reporter SMAD1-RFP in hGs from D1 to D4. Scale bars, 100 μm (n = 300 samples from five experiments). (G) Confocal micrographs showing the expression of amniotic ectoderm markers, TFAP2A and GATA3, and pSMAD1/5/8, an intracellular target of BMP4 signaling in D1 hGs. Scale bars, 50 μm (WT). Red arrowheads indicate coexpression of pSMAD1 with either TFAP2A or GATA3 (n = 30 samples from three experiments).