Abstract
Mouse embryonic stem cells are derived from in vitro explantation of blastocyst epiblasts1,2 and contribute to both the somatic lineage and germline when returned to the blastocyst3 but are normally excluded from the trophoblast lineage and primitive endoderm4–6. Here, we report that cultures of expanded potential stem cells (EPSCs) can be established from individual blastomeres, by direct conversion of mouse embryonic stem cells (ESCs) and by genetically reprogramming somatic cells. Remarkably, a single EPSC contributes to the embryo proper and placenta trophoblasts in chimeras. Critically, culturing EPSCs in a trophoblast stem cell (TSC) culture condition permits direct establishment of TSC lines without genetic modification. Molecular analyses including single cell RNA-seq reveal that EPSCs share cardinal pluripotency features with ESCs but have an enriched blastomere transcriptomic signature and a dynamic DNA methylome. These proof-of-concept results open up the possibility of establishing cultures of similar stem cells in other mammalian species.
We sought to establish cultures of new stem cells from cleavage stage mouse embryos. Under such a culture condition, we speculated that the self renewing stem cell population might have expanded potential as the cells of 4-cell (4C) or 8-cell (8C) embryos or the individual blastomeres retain the potential to differentiate to both the trophoectoderm (TE) and the inner cell mass (ICM)7–10. In order to prevent blastomeres from further differentiation and to derive stem cell lines from these cells, we speculated that modulation of the key signaling pathways implicated in the earliest stages of embryonic development might be a rate-limiting step. Genetic and developmental studies have revealed the key roles of conserved mitogen-activated protein kinases (MAPKs), Src, and Hippo pathways in the segregation of the TE and ICM lineages; and furthermore, how disrupting them causes developmental arrest11–17. In addition, Wnt signaling is known to be a key orchestrator of the earliest development stages of vertebrates18, and to be involved in the development of mouse preimplantation embryos and trophoblasts19–22. Recent advances have uncovered important functional interactions between Wnt and MAPK pathways via Yap, the key Hippo pathway downstream effector23–25. We therefore selected inhibitors to simultaneously target these pathways or kinases to block development for the derivation of novel stem cell lines.
In order to target MAPKs, we used Mek1 inhibitor PD0325901, JNK Inhibitor VIII (for Jun N-Terminal Kinase) and p38 inhibitor SB203580. We chose A-419259, a potent pyrrolopyrimidine inhibitor, to block activities of Src family kinases26. To modulate Wnt signaling, XAV939 was used to stabilize AXIN1, the concentration-limiting component of the β-catenin and Yap destruction complex27,28. XAV939 may also suppress Yap activities via angiomotin25, and has been shown to improve culturing pluripotent stem cells29,30. Finally, we included a GSK3 inhibitor, CHIR99021, and leukemia inhibitory factor (LIF). Although, the pro-pluripotency role of GSK3 inhibition (higher Wnt activities) appears to be partially redundant when PD0325901 and LIF are present31–34, CHIR99021 improves metabolic and biosynthetic processes and therefore cell culture robustness35. LIF may promote the rare totipotent cells in mouse ESC culture36. We hereafter named the medium that contains the aforementioned inhibitors and LIF as EPSC Medium or EPSCM, which was used in subsequent experiments, unless otherwise stated.
We first investigated whether we could derive cell lines directly from single blastomeres of an 8C embryo in EPSCM, as previous attempts to culture mouse blastomeres only produced standard ESC lines37. We seeded each blastomere into a well of a 96-well plate with feeders (Fig. 1a). In the following days, in 9 out of 32 wells with EPSCM, blastomeres proliferated, and by day 12, formed large outgrowths consisting predominantly of Oct4+/Cdx2− cells (Fig. 1a). Stable EPSC lines could be established in EPSCM from most of the outgrowths (Fig. 1a). By contrast, no outgrowths were obtained from blastomeres in either M15 (15% serum plus LIF, or mouse ESC medium) (n=32) or 2i/LIF (n=32) on feeders, or in feeder-free conditions in any medium (n=96).
Next, we examined whether a cleavage stage embryo could progress developmentally in EPSCM. Surprisingly, these 4C or 8C embryos in EPSCM initially appeared to progress to blastocyst-like structures (Extended Data Fig. 1a, b). However, in EPSCM blastocysts, the blastocoel was obliterated by day 6 and was filled with cells, generating a structure reminiscent of an enlarged morula, or a “filled” blastocyst (Extended Data Fig. 1a, b). The “filled” blastocysts eventually attached and hatched, and after day 10, primary outgrowths appeared (Extended Data Fig. 1a). EPSC lines were established from the primary outgrowths in EPSCM with an efficiency of approximately 20% under feeder-free condition, and up to 100% on SNL feeder cells. We investigated Oct4 and Cdx2 expression in the developing embryos in EPSCM (Fig. 1b) and in the control medium M15 (Extended Data Fig. 1c). Immunostaining revealed a gradual loss of Oct4 and Cdx2 expression in the mid-“filled” blastocysts, which eventually led to a temporary “blank” state, where most cells had large nuclei, but almost none expressed nuclear Oct4 or Cdx2 (Fig. 1b and Extended Data Fig. 1d, e). This transient “blank” state appears to bear certain similarities with some of the early blastomeres (4C embryo in M15 in Extended Data Fig. 1c), or with the 2C-like ESCs where Oct4 protein expression is absent38. Genetically, Oct4 expression is not essential for the establishment of totipotency39. Further analysis of the “filled” blastocysts indicated that the cells appeared to be undergoing a complex reprogramming process involving cell proliferation and apoptosis, as many were positive for Ki67 (Extended Data Fig. 1d) and cleaved Caspase 3 (Extended Data Fig. 1e). This observation, and particularly the properties and potential of the individual cells within the “filled” blastocysts, warrant further investigation. Notably, in the primary outgrowths, some cells initially expressed both Oct4 and Cdx2 (Fig. 1b), as was the case with the co-expression of the two transcription factors in many 4C-8C blastomeres (Fig. 1b and Extended Data Fig. 1c). Eventually the outgrowth cells and the established EPSC lines expressed only Oct4 (Fig. 1b). In contrast, in 2i/LIF, only 5 out of 54 embryos (8C) reached the blastocyst stage and all died soon after.
The fact that embryo cells in EPSCM went through transient loss of Oct4 and Cdx2 expression, and also that some cells in the primary outgrowths initially expressed both Oct4 and Cdx2, implied that EPSCs might retain some blastomere features, even though they were morphologically similar to standard ESCs. We characterised two cells lines (DR10-EPSCs and DR25-EPSCs) derived de novo in EPSCM from 8C embryos. These cells, which expressed pluripotency genes at levels comparable to standard ESCs, had a normal karyotype, were able to form mature teratomas, and contributed to both the somatic lineage and the germline in chimeras (Extended Data Fig. 1f, g, h, i, j, k). Remarkably, once injected into morulas, both EPSCs (mCherry+) contributed to the ICM and the TE in the blastocysts (Fig. 1c and Extended Data Fig. 1l).
We subsequently cultured standard ESCs (AB2.2 and E14Tg2a (ref. 40,41)) or induced pluripotent stem cells (iPSCs) reprogrammed from MEFs in EPSCM, which were previously maintained in M15 or 2i/LIF. After five passages in EPSCM, these ESCs or iPSCs appeared to acquire the expanded potential to contribute to the TE once injected into recipient embryos (Fig. 1c and Extended Data Fig. 1l). Importantly, the donor cells from injected EPSCs expressed Cdx2 (Fig. 1d). Notably, once EPSCs were returned to 2i/LIF for as few as three passages, a much lower contribution to the TE was observed (Fig. 1c), indicating that EPSCM is necessary to sustain the expanded potential, and that EPSCs could re-acquire naïve ESC characteristics in an ESC culture condition. We named ESCs or iPSCs consecutively cultured in EPSCM for five passages as ESC-EPSCs or iPSC-EPSCs. The ESC-EPSCs could be cultured for over 30 passages, and retained robust levels of reporter GFP expression from the Oct4 and Rex1 loci42,43 (Extended Data Fig. 1m, n). They expressed comparable levels of mRNAs of pluripotency and lineage-specific genes as in ESCs (Extended Data Fig. 1o); they preferentially used the Oct4 distal enhancer in lieu of the proximal enhancer44,45 (Extended Data Fig. 1p); they differentiated to all three somatic germ layers in vitro (Extended Data Fig. 1q) and to the germline in chimeras (Extended Data Fig. 1r), and also showed efficient gene targeting (around 50%) at the Rosa26 locus46. Epigenetically, female EPSCs had two active X chromosomes (Extended Data Fig. 1s). EPSCs thus shared the cardinal features associated with pluripotency. Biochemically, targeting the pathways and kinases known to be important in the earliest developmental stages by the inhibitors in EPSCM resulted in the effective modulation of their activities (Extended Data Fig. 1t). Notably, EPSCs had increased Axin by XAV939 (Extended Data Fig. 1t), which led to higher levels of phosphorylated-β-catenin and lower active β-catenin in the nucleus (Extended Data Fig. 1u). Consequently, EPSCs did not have detectable Wnt activity in Topflash assay (Extended Data Fig. 1v), even though EPSCM contained the GSK3 inhibitor CHIR99021, demonstrating XAV as a potent Wnt inhibitor. EPSCs were responsive to LIF or Jak/Stat signaling (Extended Data Fig. 1w) but were resistant to FGFRi and ALK5i, similar to ESCs in 2i/LIF (Extended Data Fig. 1x).
We further determined the in vivo differentiation potency of EPSCs by implanting injected recipient 8C embryos after they had developed to the morula (at CRI) or blastocyst (at the Sanger) stage. The embryos were subsequently collected around 6.5dpc for assessment of contribution. In about half of the chimeras (59/113), donor mCherry+ cells were found in the extraembryonic ectoderm region that was stained positively for Elf5 (Ref. 47) (Fig. 2a, and Extended Data Fig. 2a). In contrast, 2i/LIF ESCs rarely contributed to this region. To ensure reproducibility of EPSC contribution in chimeras, we independently converted another GFP-expressing ESC line in EPSCM at the University of Tokyo. These GFP+ EPSCs, but not the GFP+ 2i/LIF ESCs, once injected into recipient blastocysts, contributed to the extraembryonic ectoderm region in 6.5dpc chimeras (11/68) (Fig. 2a. Top panel, live images).
In order to examine the contribution of EPSCs in the placenta, we analysed 14.5dpc chimeras. Whole-mount fluorescence examination of the chimeras indicated mCherry+ donor cell contribution to the embryo proper and possibly the extraembryonic tissues (33 of 71 embryos for DR10-EPSCs, and 23 of 41 for DR25-EPSCs, 21 out of 38 for AB2.2-EPSCs and 16 of 28 for AB2.2-2i/LIF ESCs) (Extended Data Fig. 2b). Flow cytometry analysis of dissociated single cells from the placenta confirmed the presence of descendants of mCherry+ EPSCs (Fig. 2b). We sorted mCherry+ (donor) and mCherry− (host) placenta cells for gene expression, which revealed that both groups of cells expressed comparable levels of trophoblast genes, such as Ascl2 (Mash2), Gcm1, Pl-1, and Hand1 (ref. 48) (Fig. 2c). By contrast, flow cytometry failed to detect a substantial number of descendants of AB2.2-2i/LIF ESCs in the placenta (Fig. 2b).
Many mCherry+ placenta cells from EPSC chimeras were polyploid in DNA content analysis (Extended Data Fig. 2c), similar to mCherry− cells, as trophoblasts could be detected as 2N, 4N, 8N and so on, due to endoreplication and cell fusion in placenta development49. At the cellular level, immunostaining of placenta sections identified mCherry+ cells in EPSC chimeras, but not in the placenta of 2i/LIF ESCs (Extended Data Fig. 2d). Notably, mCherry+ placenta cells were stained positively for transcription factor Tfap2c (Extended Data Fig. 2d), which is expressed in spongiotrophoblasts and other types of trophoblasts50, and is one of the factors for the generation of induced trophoblast stem cells51,52. We further FACS-purified mCherry+ placenta cells from the EPSC-chimeras, cytospun and immunostained them for additional trophoblast markers. Besides Tfap2c, trophoblast markers Gcm1, Ezrin and Cytokeratin 7 (CK7) could clearly be detected in the sorted mCherry+ placenta cells (Fig. 2d, and Extended Data Fig. 2e).
We wished to determine and exclude the possibility that mCherry+ placenta cells had gone through cell fusion between the donor EPSCs and host trophoblasts at any stage of development. Therefore, we injected H2B-mCherry+ AB2.2 EPSCs into host embryos that were genetically marked (The SleepingBeauty Transposase-SB10 was targeted to the Rosa26 locus)46. Genotyping the sorted mCherry+ placenta cells (14.5dpc) for the presence of mCherry or SB10 genomic DNA confirmed robust mCherry DNA amplification, but no SB10, whereas in the mCherry− cells SB10 DNA could be readily detected (Extended Data Fig. 2f), thus genetically excluding cell fusion events that could have produced mCherry+ placenta cells. In the mCherry− placenta cells, minor mCherry DNA amplification was noticed (Extended Data Fig. 2f), indicating a small number of mCherry+ donor cells becoming mCherry−, due to the silencing of the CAG promoter that controlled mCherry expression.
The yolk sac of 14.5dpc chimeras consists of cells derived from the extraembryonic mesoderm, which come from the epiblast, and extraembryonic endoderm cells, which are differentiated from the primitive endoderm (hypoblast). Standard ESCs only contributed to yolk sac extraembryonic mesoderm cells (endothelial cells and mesothelial cells)6 (Extended Data Fig. 2g). Remarkably, descendants of EPSCs could clearly be found in both the extraembryonic endoderm and the mesoderm cell layers (Extended Data Fig. 2g). These in vivo results demonstrated EPSCs’ expanded developmental potential for both the embryo proper and all the major extraembryonic lineages.
To address the possibility of distinct EPSC subpopulations independently contributing to either embryonic or extraembryonic lineages, and to further demonstrate the potency of EPSCs, we tested chimera contribution following the injection of a single EPSC. To this end, each 8C host embryo was injected with a single EPSC (DR10, DR25 and AB2.2) or with a control ESC (AB2.2-2i/LIF) for chimera production. Under the fluorescence microscope, mCherry+ cells were found in 7 out of 85 14.5dpc embryos for DR10 EPSCs, 8 out of 77 for DR25 EPSCs, 7 out of 51 for AB2.2-EPSCs, and 10 of 82 for AB2.2-2i/LIF ESCs (Extended Data Fig. 2h). Flow cytometry analysis of the mCherry+ chimeras showed EPSC descendants in the embryo proper, and in the extraembryonic tissues in almost every chimera examined (10/11) (Extended Data Fig. 2i). FACS-purified mCherry+ placenta cells from EPSC chimeras expressed comparable levels of trophoblast lineage markers (Extended Data Fig. 2j), and contained an 8N polyploid population (Extended Data Fig. 2k). By contrast, injected single AB2.2-ESCs cultured in 2i/LIF contributed poorly to the embryo proper, and little to the placenta in chimeras, as revealed by flow cytometry (Extended Data Fig. 2i).
We formulated EPSCM based on genetic and developmental studies. To dissect the effect of individual inhibitors or their redundancy in EPSCM, we cultured AB2.2 ESCs in various combinations of inhibitors, based on the 2i/LIF medium, for 5 passages before being injected into 8C host embryos. The chimeric blastocysts were co-stained to detect mCherry for the donor cells and Cdx2 for TE contribution. Adding A419259 or XAV939 to the 2i/LIF medium substantially increased TE contributions whereas JNKi and p38i had moderate effects (Extended Data Fig. 2l). The functions of these inhibitors on TE contribution, in particular XAV939 and A419259, were further confirmed by the removal of individual inhibitors from EPSCM (Extended Data Fig. 2l). Although removing CHIR99021 only had a minor effect on the TE contribution (Extended Data Fig. 2l), EPSCs required it for robust proliferation, as revealed in the colony formation assay (Extended Data Fig. 2m). These results thus revealed that the acquisition of TE contribution is possibly an additive effect of modulating more than one kinase or pathway. Importantly, although EPSCM appears to have some redundancy in inhibitor requirement, the percentage of extra-embryonic contribution was highest when all inhibitors were present. On the other hand, the fact that 3-inhibitor combinations, including (2i+A419259, 2i+XAV939, or CHIR99021+A419259+XAV939), were able to confer ESCs with some expanded potential (Extended Data Fig. 2l, 2n) suggests that in the future, EPSCM can potentially be further simplified or optimised. We subsequently examined ESCs cultured in 2i+A419259 medium (2i+A) for their contribution in 14.5dpc chimeras, and detected the descendants of donor cells in the placenta (Extended Data Fig. 2o), which expressed trophoblast markers (Extended Data Fig. 2p), and donor cells in the yolk sac extraembryonic endoderm (Extended Data Fig. 2q).
To understand the molecular characteristics of EPSCs, we profiled the transcriptomes of EPSCs and ESCs in different culture conditions by RNA-seq. In hierarchical clustering, the transcriptomes of the cells segregated by their maintenance conditions, irrespective of their original derivation methods or culture history (Fig. 3a). Using single cell RNA-seq (scRNA-seq), we profiled 84 EPSCs (DR10) to study the molecular heterogeneity of the EPSC culture (Extended Data Fig. 3a). We compared our data with the scRNA-seq dataset of 2i/LIF and M15 ESCs generated on the same platform with an ESC line (G4) (ref. 53). In principal component analysis (PCA), individual cells were segregated by their culture conditions (Extended Data Fig. 3b), demonstrating the global differences between these cells (Extended Data Table 1). Gene ontology term enrichment analysis revealed that 2i/LIF ESCs were enriched in genes of metabolic processes such as oxidative reduction and the electron transport chain (Extended Data Fig. 3c), concordant with a previous report54, whereas biological terms related to transcriptional regulation and embryonic development, particularly placental development, were preferentially featured in EPSCs (Extended Data Fig. 3c).
Expression variability of key pluripotency genes (quantified by the coefficient of variation) at the single cell level was compared between EPSCs and 2i/LIF or M15 ESCs, which showed that both EPSCs and 2i/LIF ESCs had similar low variability of these genes (highly homogeneous), unlike the high variability observed in M15 ESCs (Extended Data Fig. 3d). Similarly, EPSCs and 2i/LIF ESCs also had comparable transcriptional variability of protein-coding genes, genes of putative transcription factors as well as global gene expression, which were different from M15 ESCs (Fig. 3b). These scRNA-seq data confirmed that EPSCs had similar transcriptomic homogeneity to 2i/LIF ESCs, and importantly, did not appear to have subpopulations contributing to distinct lineages.
To assess the molecular similarities of EPSCs to in vivo preimplantation blastomeres, we retrieved the mouse embryonic development time course single-cell data from Deng et al55 for comparison. EPSCs were separated from the native developmental trajectory of blastomeres in the first three principal components (Extended Data Fig. 3e). Yet, the scores of EPSCs were at the range of 4C blastomeres in PC1 (Figure 3c), the component representing the major embryonic development axis. To test whether EPSCs had enriched transcriptomic features of blastomeres compared with standard ESCs, we compiled the top 500 stage-specific genes of each blastomeric stage from Deng et al and compared the expression of these signature gene sets in EPSCs and 2i/LIF ESCs by GSEA. The result showed significantly higher enrichment of early pre-implantation (zygote, 2C and 4C) signatures in EPSCs (Fig. 3d). Nr5a2 (Lrh1) and Rarg are specifically highly expressed in 2C-8C blastomeres55,56. A recent genome-wide epigenomic profiling study in blastomeres identified them as key determining factors in the segregation of the TE and ICM where Nr5a2 promotes ICM gene expression56. Interestingly, from the scRNA-seq data, EPSCs expressed both genes at high levels whereas only Nr5a2 was highly expressed in 2i/LIF ESCs (Extended Data Fig. 3d). Similarly, EPSCs, but not 2i/LIF ESCs, had high levels of Aire, Thap11 (Ronin) and Lin28 (Extended Data Fig. 3d), which are also highly expressed in 2C or cleavage stage blastomeres, respectively57,58. Collectively, these results indicate that the transcriptome of EPSCs is enriched with features of early blastomeres. Despite some transcriptomic similarities, it is important to note that EPSCs are in vitro cultured cells and are different in many aspects from in vivo blastomeres. Finally, the recently reported rare totipotent-like ESC subpopulations (2C-like or MERV-TdTomato+, and Hhex-Venus+)36,38 or in vivo reprogrammed iPSCs (iviPSCs)59 appear to have trophoblast potential. We showed previously that the 2C-like cells had limited gene expression differences compared to standard ESCs53. EPSCs, on the other hand, were distinct from all these cells (Extended Data Fig. 3f). Unlike in 2C-like cells, EPSCs had only limited up-regulation of endogenous retroviral transcripts (Extended Data Fig. 3g).
Blastomeres of 4C-8C embryos have high expression of genes encoding Tet proteins and DNA methyltransferases55, and have high levels of 5hmC owing to active demethylation60. Further investigation of scRNA-seq data revealed that EPSCs had similar high expression of genes of both methyltransferases (Dnmt1, Dnmt3a & Dnmt3b) and components involved in active DNA demethylation (Tet1, Tet2 and Tdg) (Extended Data Fig. 3d), whereas 2i/LIF ESCs showed Prdm14-mediated down-regulation of Dnmt3a and Dnmt3b due to ERK signaling inhibition61 (Extended Data Fig. 3d and Extended Data Table 1). The expression patterns of the genes encoding DNA methylation/demethylation proteins in EPSCs were reflected on the global cytosine methylation and hydroxymethylation levels. Consistent with previous reports61,62, substantially lower levels of DNA methylation were found in 2i/LIF ESCs, compared to that in M15, whereas EPSCs showed an intermediate level (Fig. 3e). Strikingly, much higher levels of hydroxymethyl-cytosine were found in EPSCs (Fig. 3f). The DNA methylome of EPSCs is therefore highly dynamic with active DNA methylation and demethylation.
Besides the dynamic DNA methylome, EPSCs had more genes (6224 vs. 3968) associated with both H3K4me3 and H3K27me3 (bivalent) than 2i/LIF ESCs54 (Fig. 3g, Extended Data Fig. 3h, i. and Extended Data Table 2). In EPSCs, bivalency of H3K4me3 and H3K27me3 signals at key pluripotency loci such as Pou5f1, Sox2 and Nanog were highly similar to 2i/LIF ESCs (Extended Data Fig. 3j), in line with the fact that these pluripotency genes were expressed at similar levels in both cell types (Extended Data Fig. 3d). Gene ontology term enrichment analysis showed that EPSC-specific bivalent genes (Extended Data Table 2) were enriched in biological processes of somatic lineage and placental development (Extended Data Fig. 3k), including genetic loci such as Cdx2, Eomes, Esx1, Ascl2, Pax6, Sox17, Gata6 and Brachyury (T) (Fig. 3h), providing an epigenetic basis for the observed co-expression of lineage markers in the Hhexhigh cells in standard ESC culture that show some totipotency features36. Remarkably, in line with EPSCs having enriched transcriptomic features of blastomeres, the H3K4me3 patterns of those key developmental loci in EPSCs resembled that of 8C embryos63 (Fig. 3h). Long and high intensity H3K27ac domains, i.e. super enhancers, have been associated with master regulators and cellular identity64. To further study the epigenetic characteristics of EPSCs, we profiled the genome-wide distribution of H3K27ac in EPSCs (DR10) and identified genes that are associated with super enhancers. As expected, pluripotency master regulators such as Pou5f1 and Nanog were associated with super enhancers in EPSCs (Fig. 3i and Extended Data Fig. 3l), similar to 2i/LIF ESCs. Nevertheless, EPSCs had acquired a unique super enhancer at the Gata3 locus, an essential regulator for trophectoderm development65, which is absent in 2i/LIF ESCs (Fig. 3i and Extended Data Fig. 3l).
Trophoblast stem cell (TSC) lines are derived from the TE or extraembryonic ectoderm of early implantation embryos66,67. Since ESCs normally originate from the epiblasts in the ICM, which are already separated from the TE in the blastocyst, it has been challenging to derive stable TSC lines from ESCs. On the other hand, genetic or epigenetic manipulation of mouse ESCs, including transcription factor overexpression or inactivation, or modulation of signal transduction pathways, can make ESCs acquire some potential to differentiate to placental trophoblasts11,68–70. However, these TSC-like cells are fundamentally different from TSCs that are derived from mouse embryos or that are reprogrammed from fibroblasts51,52,70. In particular, the trophoblast lineage gatekeeper Elf5 is not robustly expressed in these TSC-like cells.
Distinct from standard ESCs, EPSCs possessed enriched transcriptomic features of blastomeres and contributed to the extraembryonic ectoderm and placental trophoblasts in chimeras. Therefore, we endeavoured to establish a stable TSC line from EPSCs, using a direct culture condition switch. We initially cultured AB2.2 EPSCs for 6 days on MEFs in TX medium71 containing Fgf4 and TGF-β1, which produced cells of various lineages, including some Cdx2+ cell patches morphologically resembling TSC colonies (Extended Data Fig. 4a). To facilitate characterisation of these TSC-like cells from EPSCs, we converted the Cdx2-GFP reporter ESCs72 to EPSCs, where a GFP expression cassette is inserted into the Cdx2 locus to allow tracking Cdx2-expressing cells.
As early as three days after the Cdx2-GFP EPSCs were cultured in TX medium, patches of GFP+ cells were visible (Fig. 4a), which were subsequently FACS-purified and re-plated in TX medium. TSC-like colonies from these sorted single cells were picked and expanded to establish stable lines for continuous passaging over 20 times in TX medium (Fig. 4b). Under the same condition, no GFP+ cells were detected from 2i/LIF ESCs. The TSC-like cells proliferated similar to the control TSCs, and expressed high levels of TSC genes, including Cdx2, Elf5, Eomes, and Tfap2c, similar to that in control TSCs (Fig. 4c and Extended Data Fig. 4b). Notably, they did not express the pluripotency gene Oct4 or the three embryo germ layer markers, Fgf5, Brachyury (T) or Gata6 (Fig. 4c). These EPSC-derived TSC-like cells were hence named EPSC-TSCs. Once Fgf4 was withdrawn from TX medium, EPSC-TSCs terminally differentiated into trophoblasts including some polyploid trophoblast giant cells (Extended Data Fig. 4c), and expressed high levels of the differentiated trophoblast markers Tpbpa, Pl-1, Pl-2, Ctsq and Prl2rc2 (ref. 71) (Extended Data Fig. 4d, e). TSCs have the peculiar property to form haemorrhagic tumours when implanted subcutaneously owing to the invasive properties of trophoblast giant cells during implantation73,74. EPSC-TSCs formed obvious haemorrhagic lesions in the immunocompromised recipients 7 days after subcutaneous injection (Extended Data Fig. 4f). Histological examination of the lesion sections revealed differentiated trophoblastic giant cells and blood-filled lacunaes, demonstrating an invasive capacity of trophoblasts into host vessels (Fig. 4d and Extended Data Fig. 4g). Finally, immunostaining confirmed Tfap2c+ trophoblasts in the sections (Fig. 4d).
To determine whether EPSC-TSCs can function properly in their native environment, we injected the H2B-mCherry-expressing EPSC-TSCs into recipient 8C embryos, which were further cultured for 48 hours prior to staining for mCherry and Cdx2 expression to facilitate localization of the injected cells. Similar to TSCs (expressing a cytoplasmic GFP from a constitutive promoter)66, the mCherry+ EPSC-TSCs were found in the TE or the outer layer of the blastocysts (TSCs: 31 out of 40 blastocysts; EPSC-TSCs: 127 out of 163 blastocysts) with some cells expressing Cdx2 (Fig. 4e). We implanted the injected embryos into recipient mice, and collected E6.5 embryos for analysis. Both the control TSCs (GFP+) and the mCherry+ EPSC-TSCs contributed to the Elf5+ extraembryonic ectoderm region (chimeras and total embryos: TSCs, 19 out of 45; EPSC-TSCs, 33 out of 94) (Fig. 4f). These results proved that bona fide TSCs were directly derived from EPSCs by a simple culture condition switch, and further validated the expanded potential of EPSCs.
These results thus provide clear proof that deriving new stem cells with expanded potential can be achieved. It is envisaged that the successful establishment of mouse EPSCs from cleavage stage embryos and by conversion from ESCs/iPSCs will offer a unique opportunity for the study of the earliest stages of embryo development. Furthermore, the insights gained herein may accelerate the establishment of cultures of similar stem cells from other mammalian species for which embryonic stem cells or iPSCs are still not available.
Author contributions
D. J. R. and W. W. developed EPSCM, derived mouse EPSC lines, performed the in vitro and in vivo differentiation assays, edited the manuscript and made figures. J. Y. produced mouse iPSCs, performed Western blots, immunostaining of 5.5-7.5dpc embryos, placenta sections, yolk sac, and sorted placenta cells, analysed preimplantation embryos in culture and immunostaining, investigated minimal requirement of inhibitors, deriving and characterising EPSC-TSCs, made the final figures and edited the manuscript. J. C. T. carried out the single-cell RNA-seq experiment, analyzed the RNA-seq and ChIP-seq data, and produced the genomics figures and wrote the manuscript. X. G., L. A., Y. Y., J. W. and C. W. performed experiments. L. C. interpreted the placental and teratoma slides. F. Y. and B. F. karyotyped the EPSC and ESC lines. G. L. performed most of the EPSC and ESC injections. H. M. performed independent mouse EPSC conversion and injection experiments at Nakauchi lab. M.W. assisted embryo production. Z. Z. performed confocal imaging and interpretation. J. B., R. R. and X. Z. provided microinjection resources. A. K. and S. T. provided single-cell RNA-seq data from standard ESCs. A.W. performed EPSC differentiation, Y.T. carried yolk sac immunostaining and embryo dissection, and B. G. provided supports for A. W and Y. T. M. E-M. and W. R. provided whole genome DNA methylation data. L. L. contributed intellectually and experimentally. P. L. conceived the concept, designed the studies, wrote the manuscript, and supervised the overall research project and manuscript preparation.
Acknowledgments
We thank colleagues of RSF (Brendan Doe, Stuart Newman and Evelyn Grau and others), Yvette Hooks, Sequencing (Natalie Smerdon) and FACS core facilities (Bee Ling Ng and Jennifer Graham) at the Sanger Institute, the animal facility at CRUK-CI for excellent technical support, Dr. Myriam Hemberger for providing TSCs and Dr. Sebastian Gerety for the fluorescence stereo microscope. We thank Dr. Jong Kyoung Kim for informatics advice, Stephen Rice for helps on DNA bisulfite sequencing analysis and Dr. David Goulding for confocal imaging facility. We thank Dr. Jennifer Nichols for their comments and for female NOD ESCs and TSCs. We thank Professor Jamie Thomson for comments. We appreciate critical comments from Professors Jamie Thomson and Elizabeth Robertson. D.R. is a recipient of the Wellcome Trust Clinical PhD Fellowship for Academic Clinicians. L.A. is a recipient of a Ph.D. fellowship from the Portuguese Foundation for Sciences and Technology, FCT (SFRH/BD/84964/2012). Y. T. was supported by a Japan Society for the Promotion of Science fellowship. A.C.W. was supported by the National Institute of Health Research (RP-PG-0310-10002). M. E-M is supported by an EMBO Fellowship (ALTF938-2014) and Marie Sklodowska-Curie Individual Fellowship. W. R. acknowledges funding from BBSRC (BB/K010867/1) and Wellcome Trust (095645/Z/11/Z). L. Lu is supported by the National Natural Science Foundation of China (31370904 and 30972691). P. L. thanks Professors Mike Stratton, Allan Bradley, Neal Copeland, Nancy Jenkins and James Lupski for their encouragement in these experiments. P. L. lab is supported by the Wellcome Trust (grant number 098051).