Inducible Stem-Cell-Derived Embryos Capture Mouse Morphogenetic Events In Vitro

1Mouse embryo development relies on interactions between the epiblast (EPI), the extra-embryonic ectoderm (ExE), and the visceral endoderm (VE), which are the respective precursors of the embryo proper, the placenta, and the yolk sac. These tissue interactions transform the embryo from the blastocyst into the egg cylinder: the EPI and the ExE polarize and open two luminal cavities, which, eventually, fuse to form the proamniotic cavity, while VE grows to envelop embryonic and extra-embryonic tissues. The anterior-posterior axis is established by a group of VE cells that migrates unilaterally to protect anterior EPI from posteriorizing signals by secreting Dkk1, Cerl, and Lefty1, which antagonize Wnt, Bmp4, and Nodal signaling. The anterior EPI later upregulates the expression of Sox1 and commits to become neuroectoderm and surface ectoderm. Mesodermal identity in the EPI is marked by the upregulation of Brachyury (Bry) at the EPI/ExE boundary, followed by the epithelial-to-mesenchymal transition of cells that egress through the primitive streak (PS) to form all three germ layers.

2Over recent years, stem cell lines derived from mouse embryonic and extra-embryonic tissues have become powerful tools to complement embryological studies. Their ability to capture embryogenesis, however, is limited since each cell type is cultured in isolation and on their own do not acquire the morphology of embryos, hindering the modeling of the tissue-tissue interactions and signaling that are crucial for embryo patterning and morphogenesis in vivo.

3To address this, we have developed a stem-cell-based model of embryonic development by combining ES, TS, and extra-embryo endoderm (XEN) stem cells into structures called ETX embryos, and our results have been independently validated. These ETX embryos closely resemble egg-cylinder-stage embryos at the morphological and transcriptional level. Yet, complex developmental events such as gastrulation occur only to a limited extent in this system, thus suggesting the need to improve it in order to capture the morphogenetic events occurring during natural embryo development.

4To explain the limits of the ETX-embryo system, we reasoned that one or more cell types used for ETXembryo generation do not have the correct developmental potential to recapitulate embryo development. Indeed, XEN cells were reported to be more similar to parietal endoderm than to primitive endoderm (PrEn) or VE, pointing to a need to replace XEN cells as a partner in the ETX embryo.

5As induction of two transcription factors Gata4 or Gata6 in ESCs is sufficient to differentiate them toward endoderm, we hypothesized that such “induced” endodermal cells could functionally replace XEN cells. Here, we test this hypothesis by combining ES and TS cells with ESCs transiently expressing Gata4 in response to Doxycycline (Dox). The resulting induced ETX embryos (iETX embryos), in addition to expressing canonical post-implantation markers, can recapitulate complex morphogenetic events leading to formation and migration of the anterior signaling center and gastrulation.

6To test whether replacing XEN cells with a cell type more similar to PrEn or VE could increase the developmental potential of ETX embryos, we modified our CAG-GFP/tetO-mCherry ES line (showing constitutive membrane GFP and transient mCherry expression following Dox-treatment) to transiently express Gata4 in response to Dox (CAG-GFP/tetO-mCherry/tetO-Gata4 ESCs, CAG-tetOG4 hereafter). We confirmed robust expression of Gata4 mRNA after 6 h of Dox-treatment, independently of the ES culture medium (Figure 1A). Gata4 induction was necessary and sufficient for expression of endodermal proteins Gata4, Sox17, and Gata6 1 day after cell seeding, but Oct4 expression was retained at this time (Figure 1B); also note the CAG-GFP downregulation upon Gata4 induction, Figure S1A).

7To test the effect of Gata4 overexpression on cell fate, we aggregated 8-cell stage embryos with ESCs and found that untreated CAG-tetOG4 ESCs contributed exclusively to the EPI, while Dox-treated CAG-tetOG4 ESCs could contribute to the PrEn (Figures 1C and 1D, 63 embryos). These results indicate that transient Gata4 expression is sufficient to change the potential of ESCs from EPI to PrEn lineage.

8As an additional test to determine whether the CAG-tetOG4 ESCs could function as “building blocks” for ETX embryos, we generated ES aggregates using CAG-tetOG4 ESCs by combining (1) solely ESCs in control conditions (no Dox), (2) solely ESCs treated with Dox, or (3) Dox-treated and untreated ESCs in a 50:50 ratio and assessed cell fate after 3 days in vitro. In untreated aggregates, the ESCs maintained their identity and expressed Oct4 with a few exceptions (2/43 structures scored). We did not observe expression of Gata6 (Figure 1E), indicating that without Dox-treatment, upregulation of endodermal markers was extremely rare. Podxl distribution was scattered, indicating a lack of polarization and lumenogenesis. In contrast, aggregates of Dox-treated ESCs completely downregulated Oct4 and induced Gata6 (55/56 structures scored). Podxl was on the outside of these aggregates in a continuous layer, suggesting that these structures were polarized but they did not undertake lumenogenesis (Figure 1E). Lack of Oct4 indicated that all cells failed to retain an ESC identity. Finally, aggregates generated by combining treated and untreated cells in a 50:50 ratio had an outer layer of Gata6-expressing cells and an inner compartment of Oct4-expressing ESCs. These aggregates underwent polarization and lumenogenesis and their morphology was reminiscent of the post-implantation EPI surrounded by VE (VE-like layer) (60/65 structures scored) (Figure 1E).

9A lineage-tracing experiment with Dox-treated CAG-tetOG4 ESCs and unlabeled ESCs showed that the VE-like layer expressed CAG-GFP, confirming that it was generated by the Dox-treated cells with the Gata4 transgene (Figures S1B and S1C). These results suggest that combining Dox-treated and untreated CAG-tetOG4 ESCs leads to the generation of self-organising aggregates containing VE-like and EPI-like compartments, a crucial building block in ETX embryogenesis.

10To test whether XEN cells could be functionally replaced by CAG-tetOG4 ESCs to generate ETX embryos, we combined wild-type CAG-GFP ESCs, Dox-treated CAG-tetOG4 cells, and TS cells in AggreWells (Figure 2A). After 24 h, cells had aggregated but no clear organization could be discerned. After 48 h, aggregates had fully compacted and increased in size. At 72 h, aggregates began elongation and developed an egg-cylinder-like morphology reminiscent of ETX and natural post-implantation embryos. At 96 and 120 h, the structures outgrew the microwells, requiring transfer to a bigger culture vessel. From 72-h onward, we could observe formation of the VE-like layer, encompassing the ES and the TS cells, which had formed distinct abutting compartments. Lumen formation in ES and TS cells was observed at 72 h, and at 96 h the lumens merged. Because these ETX embryos were generated by replacing XEN cells with Dox-treated “induced” CAG-tetOG4 cells, we termed them iETX embryos.

11To calculate the formation efficiency of iETX embryos, we collected the whole contents of several wells 3 days and 4 days after cell seeding and counted the number of structures containing an epithelial, GFP-expressing ES compartment, segregated from an unlabeled TS compartment, both surrounded by a layer of cells, over the total number of structures (Figures 2B and S2A). At day 3, over 30% of the structures showed correct morphology, but at day 4, this fraction had decreased to 20% (Figure 2C), likely because some structures failed to develop properly between days 3 and 4 (Figure 2B). CAG-GFP expression was downregulated after Dox-treatment of the CAG-tetOG4 ESCs in the VE-like layer (Figure S2B), as in natural embryos. For the remaining experiments, we restricted our analyses to iETX embryos with correct morphology.

12To follow iETX development, we seeded our cells in PEG-hydrogel plates to monitor iETX embryogenesis by live imaging (Figures 2D and S2C). After 24 h, cells compacted and formed an aggregate. Between 24 and 48 h, a CAG-GFP ES compartment became surrounded by a thin layer of cells; at 24 h the TS cells formed a very small layer or clump in the structure. The TS compartment emerged between 48 and 72 h, and the layer surrounding both ES and TS compartments became thicker and more prominent. The ES compartment underwent lumenogenesis between 60 to 67 h. At 72 h and onward the structure elongated until it grew larger than the diameter of the well. Similar to our observations with ETX embryos, iETX embryos developed to strongly resemble mouse post-implantation embryos. The VE-like layer, unlike the ES compartment, was mCherry positive for the first 24 h, indicating that it had been generated by Dox-treated CAG-tetOG4 cells (Figures 2E and S2D). Transient mCherry disappeared by 72 h (Figure S2E).

13To ascertain whether iETX embryos expressed the appropriate post-implantation markers, we analyzed canonical lineage markers 4 days after cell plating. The VE-like layer expressed VE markers Gata6, Gata4, and Sox17 (Figures 2F–2H and S2F); the TS compartment expressed the ExE markers Eomes, Ap2γ and Cdx2 (Figures 2F–2H and S2F); and the ES compartment expressed EPI markers Oct4 and Otx2 (Figures 2F–2H, 80/82 structures; Figure S2F). Finally, the VE-like layer adjacent to the ES compartment expressed Eomes and Otx2 in nearly all the cases examined (Figures 2F and 2I, 68/73; Figure S2F), in agreement with Eomes and Otx2 expression in the embryonic part of the VE. In comparison, the XEN layer of ETX embryos expressed Otx2 and Eomes in 40% of cases. iETX embryos at day 4 had a higher number of cells in each lineage in comparison with ETX embryos at day 4 (Figure 2J).

14Since ETX embryos have a limited ability to form the anterior VE (AVE) signaling center, we wondered whether iETX embryos could better capture this process. We found that iETX embryos at 4 days of development expressed three canonical AVE markers, Cerl, Lefty1, and Dkk1 (Figures 3A, 3B, S3A, and S3B). Cells expressing Cerl, Lefty1, and Dkk1 were either at the distal tip of the iETX embryo, as at the time of AVE formation in E5.5 embryos (Figures 3A and S3A) or were positioned asymmetrically on one side of the iETX embryo, like their expression at E5.75 as the AVE begins its migration toward the future anterior (Figures 3B and S3B).

15Expression of these markers was maintained at day 5 (Figures 3C and 3D). At day 4, 65% of iETX embryos expressed Lefty1, 50% expressed Dkk1, and 36% expressed Cerl (Figures 3E–3G, Cerl, 38/106; Lefty1, 58/89; Dkk1, 46/89). At day 5, 80% of iETX embryos expressed Cerl (Figure 3E, 133/171) and Dkk1 expression increased to 70% (Figure 3F, 52/73). In contrast, the proportion of iETX embryos expressing Lefty1 decreased from 65% to 40% between days 4 and 5 (Figure 3G, 32/87). We hypothesized that these differences at day 4 might relate to differences in the size of the iETX embryos, because Cerl is repressed by BMP4 signaling from the ExE, and Cerl is expressed only when the EPI has extended beyond a specific length. We, therefore, measured the length of the ES and TS compartments and correlated it to Cerl expression. iETX embryos with an ES compartment longer than the TS compartment consistently expressed Cerl; in contrast, when the TS compartment was longer than the ES compartment, there was no Cerl expression (Figure 3H). We did not observe this relationship for Dkk1 (Figure 3I) and it was much weaker for Lefty1 (Figure 3J), suggesting that Cerl expression might be regulated differently than Lefty1 and Dkk1. Analysis of the aspect ratio of Cerl-positive and Cerl-negative cells indicated that Cerl-positive cells were very similar in their morphology to AVE cells in embryos (Figures 3K and 3L).

16In embryos, the distribution of Cerl-, Lefty1-, and Dkk1-expressing cells varies between E5.5 and E6.0 because the AVE migrates toward the EPI/ExE boundary. At day 4, the AVE was either distal or lateral (52% and 41% of the cases, Figures 3N and 3O) and very rarely proximal (6%, Figure 3P). In contrast, at day 5, 22% of iETX embryos had proximal AVE and 33% had distal AVE, but there was no change in lateral AVE (Figures 3N–3P). Thus, at day 4, iETX embryos resemble E5.5/E5.75 embryos in terms of AVE specification and position. At day 5, the increase in the proportion of proximal AVE could either suggest AVE migration or localized de novo upregulation of AVE markers instead of cell migration.

17To distinguish between these possibilities and investigate the dynamics of AVE induction and migration, we generated iETX embryos with a Cerl-GFP reporter/tetO-Gata4 line derived from our transgenic Cerl-GFP mouse line, in which GFP expression is under the control of the Cerl promoter. We began live imaging at day 3, when the reporter was not yet active (Figures 3Q and S3C). Reporter expression commenced in a single cell at the tip, expanded from a single cell to several, and initially resulted in the formation of a GFP-positive domain. Following consolidation of GFP expression at the iETX embryo’s tip, the AVE began its migration until it reached the boundary with the TS compartment after an average of 18 h (Figures 3Q, S3D, and S3E; Videos S1 and S2), which is slower than AVE migration in the embryo.

18Tracking the migration trajectories of individual Cerl-GFP cells showed they migrated directionally toward the proximal end of iETX embryos (Figure S3F; Videos S3 and S4), consistent with AVE migration in embryos. When compared with the speed of AVE cells in embryos (0.23 ± 0.07 μm/min; mean ± SD), AVE in iETX embryos migrated at a similar albeit slightly reduced speed (0.198 ± 0.040 μm/min, 26 cells from 7 iETX embryos). Immunofluorescence confirmed the asymmetric localization of the Cerl-GFP cells in the majority (26/36) of iETX embryos examined (Figures S3G and S3H). This is an important improvement over the ETX embryos, in which we could observe neither the Cerl and Dkk1 proteins nor the AVE migration.

19We next sought to determine whether the formation of the anterior domain was accompanied by the establishment of a posterior one. Embryos at E6.5 express Cerl and mesoderm marker Bry on opposite sides (Figures 4A and S4A). iETX embryos displaying Bry on only one side of the ES compartment on the fifth day of development comprised 55% of the total Bry- and Cerl-expressing structures (Figures 4B and S4B, 38/69 structures). In this subset, 87% of iETX embryos expressed Cerl on the side opposite to Bry (Figures 4C and 4D), whereas in 13% of iETX embryos, Cerl was at the distal tip. Importantly, we could not observe any instance in which Cerl and Bry were expressed on the same side (Figures 4C and 4D). We examined the expression of two other anterior markers, Dkk1 and Lefty1, in relation to Eomes, which is expressed in the PS and in iETX embryos at day 5 (Figure S4C), which confirmed that formation of anterior and posterior domains occur on opposite sides in 80%–90% (Cerl/Dkk1, 28/31; Cerl/Lefty1, 19/21) of the iETX embryos examined (Figures 4E–4H).

20Formation of anterior and posterior domains on opposite sides was also conserved in iETX embryos expressing Bry around the ES/TS boundary (45%, 31/69)(Figure 4I). In most cases, one posterior domain was more expanded than the other, hence we considered the more expanded domain as the “true” posterior and the less expanded one as an ectopic posterior. In 50% of cases, the Cerl domain was on the side of the less expanded, ectopic posterior domain and opposite the large domain (Figures S4D and 4J); in only 8% of cases was the Cerl domain on the same side of the more expanded domain; in 7% of cases the posterior domains were equal and, thus, the position of the Cerl domain could not be assigned; and in 33% of the cases, the Cerl domain was distal (Figures S4D and 4J).

21To determine whether the position and the size of the AVE could influence the likelihood that an iETX embryo would develop an ectopic posterior domain, we measured the angle of displacement of the Cerl domain away from the distal tip and toward the ES/TS boundary and the angle of extension of the Cerl domain. Structures without an ectopic Bry domain had a more extended Cerl domain, which was closer to the ES/TS boundary than structures with an ectopic Bry domain (Figures S4E and S4F). Location of the Lefty1 domain was not significantly different, but its extension was (Figures S4G and S4H). On the other hand, the Dkk1 domain was significantly closer to the ES/TS boundary in structures without ectopic Bry but the extension of the domain itself was no different (Figures S4I and S4J). These results suggest that the position and extension of the Cerl domain are important factors in preventing the expression of ectopic Bry in iETX embryos, whereas Lefty1 and Dkk1 may contribute to a lesser extent.

22Out of the iETX embryos without Cerl, 65% displayed ectopic expression of Bry, suggesting that a subset of iETX embryos can break symmetry without Cerl (Figures S4K and S4L). Since a decrease in Nodal levels can rescue ectopic Bry expression in the absence of AVE, we tested whether altering Nodal signaling levels can affect mesoderm formation. Culturing iETX embryos from day 4 to day 5 in Nodal inhibitor completely abrogated expression of Bry and Eomes (Figures S4M and S4N) in agreement with what was reported in natural embryos. These results indicate that establishment of anterior and posterior domains occurs on opposite sides of the ES compartment in iETX embryos and Nodal is involved in this process as in natural embryos.

23Having observed robust formation of anterior and posterior domains in the iETX embryo, we wished to determine whether the posterior side could also establish a PS to undertake gastrulation. Gastrulation is evident in embryos at E6.75 when Bry-positive cells near the EPI/ExE boundary undergo the epithelial-to-mesenchymal transition (EMT) and egress to form a mesoderm layer between the EPI and the VE. We found that iETX embryos also developed Bry-expressing cells between the ES compartment and the VE-like layer between days 5 and 6 (Figures 5A–5C) but not on the opposite side of the structure where Cerl was expressed.

24To understand the dynamics of this process, we performed time-lapse imaging of iETX embryos from day 4 (Figures 5D and S5A; Video S5). At the onset of imaging, cells on all sides of the GFP-expressing ES compartment were epithelial and columnar, but at 13 h, the cells on one side of the ES and TS compartment boundary began to change their shape from columnar to rounded and positioned themselves between the ES compartment and the VE-like layer (Figure 5D). Between 13 and 16 h, the layer formed by the egressing cells had almost extended to the distal tip, similar to PS formation in embryos. Throughout this process, the opposite, putative anterior side remained epithelial and did not undergo EMT. PS extension continued until 27 h, with the streak reaching the distal tip of the iETX embryo (Figure 5D). During gastrulation, the thickness of the posterior side progressively increased relative to the anterior side, which remained constant and the streak steadily increased in length (Figures 5E and 5F).

25On day 6, iETX embryos showed a large, Bry-positive domain on the side of the ES compartment undertaking gastrulation (Figure 5G). In most of the structures examined, 40% of Bry-positive cells expressed N-cadherin (Figure S5B), and there was downregulation of laminin in the basement membrane at the point of EMT, in contrast to other parts of the iETX embryo not undergoing EMT (Figures 5G and S5C). Egressed cells were Bry-positive and expressed laminin; formation of a laminin layer could be detected between the mesoderm and the VE-like layer (Figure 5G). The PS of iETX embryos expressed higher levels of Bry (Figure 5H), which was accompanied by downregulation of E-cadherin and upregulation of N-cadherin (Figure 5H). Finally, egressed cells co-expressed Bry and the EMT marker Snail (Figure 5I), suggesting that gastrulation in iETX embryos follows the same process as the natural embryo. Based on Snail-YFP expression, 64% of the structures were undergoing EMT at day 5 (27/42, see also Figure S5D).

26Gastrulation sees not only the formation of embryonic mesoderm but also the proximal migration of mesodermal cells to form extra-embryonic mesoderm. In 45% of day 5 iETX embryos undergoing EMT, Bry/Oct4-expressing cells were observed between the VE-like layer and the TS compartment (Figures S5E and S5F) and either formed patches above the ES/TS boundary or were arranged as a line of cells between the TS compartment and the VE-like layer overlaying the TS compartment (Figure S5F, orange asterisks). In 20% of the examined iETX embryos, Bry was co-expressed with Runx1, which marks extra-embryonic mesoderm cells committing to the hematopoietic lineage (Figure S5G). This represents an important improvement over ETX embryos, in which we did not observe formation of extra-embryonic mesoderm.

27During gastrulation, the developing streak generates multiple mesodermal and endodermal cell types required for subsequent development and secretes Cerl, Lefty1, and Dkk1 leading us to ask whether similar changes take place in gastrulating iETX embryos. Analysis of gastrulating iETX embryos at day 5 indicated that, similar to natural embryos (Figures 6A, 6C, and 6E), we identified expression of Cerl, Dkk1, and Lefty1 in the AVE and also in the developing PS (Figures 6B, 6D, and 6F). Posterior expression of Cerl, Lefty1, and Dkk1, respectively, was present in 51% (17/33), 79% (27/34), and 37% (21/56) of the examined iETX embryos, indicating gastrulation processes further than in ETX embryos.

28We also observed co-expression of Bry and FoxA2 at day 6 in the distal part of the PS (71%, 37/52) similar to E7.5 natural embryos (Figures 6G and 6H), which is indicative of axial mesoderm formation, and FoxA2 and Sox17 expression, indicative of definitive endoderm (Figures 6I and 6J). Bry and Sox17 were co-expressed in the ES compartment, suggesting that a subset of cells express mesodermal markers before acquiring definitive endoderm identity (Figure S6A), similar to the embryo.

29We could not determine whether expression of neuroectoderm markers would occur, because after day 6, iETX embryos became dark and lost their cylindrical appearance, suggesting that 6 days of culture is the current limit for our system. Developmental milestones of iETX embryos are presented in Figure S6B and Table S1.

30To understand further the developmental potential of iETX embryos, we performed inDrop single-cell RNA sequencing on iETX embryos and ETX embryos at day 4 and compared them with embryos at E4.5(GEO: GSE134240, ), E5.5, and E6.5 (Figure 7A). Pearson correlation coefficients between all samples indicated that iETX embryos were most similar to E5.5 natural embryos (Figure 7B). We subdivided the natural and ETX/iETX embryos into their constitutive lineages and observed that all TS cells clustered with ExE derivatives and all ESCs clustered with EPI derivatives, irrespective of whether they originated from ETX or iETX embryos. On the other hand, VE, XEN layer, and VE-like layer correlated less. The XEN layer showed some similarity with the PrEn of the E4.5 embryo but correlated little with E5.5 and E6.5 VE; the VE-like layer of iETX embryos showed similarities to E4.5 PrEn and to the VE at E6.5 but was most similar to E5.5 VE. This suggests that tetOG4 ESCs may generate an endodermal layer more similar to VE (Figure 7C).

31To identify specific cell subpopulations in this dataset, gene signatures (Figure S7A) and Leiden clustering were utilized for further classifications. The ExE cluster cells were divided into ExE or differentiating ExE based on reciprocal expression of the TS marker Cdx2 and differentiation marker Gata2, and we also observed these signatures in the TS compartment. The EPI cluster expressed the primed pluripotency marker Otx2; the E6.5 EPI could be subdivided into nascent mesoderm, based on Mesp1 expression, or into a PS subpopulation, based on Eomes and Nanog. Nascent mesoderm and PS signatures were already present in the ES compartment, suggesting that it is primed for gastrulation and may be at a more advanced stage than the E5.5 EPI. Nevertheless, expression of the EMT marker Snail1 mostly in the E6.5 EPI and not in the ES compartment confirmed that EMT in iETX embryos does not occur until day 5. The VE cluster was initially identified by Gata4 expression and then subdivided into parietal endoderm based on LamB1 expression, in embryonic VE based on Otx2/Eomes expression, and cells that were Gata4-positive, but LamB1/Otx2/Eomes-negative were classified as extra-embryonic VE. Within the embryonic VE cluster, we identified cells expressing Cerl, Lefty1, Hhex, and Dkk1, indicating the AVE. In iETX embryos, AVE-like cells expressed Dkk1, Lefty1, and Hhex1 but low Cerl, suggesting an earlier stage of AVE development, more similar to E5.25. We could detect an additional, small subpopulation of cells expressing Hhex, Lefty1, and low Otx2, similar to the expression of these markers in the E4.5 PrEn, we termed these “early VE-like layer” (Figure 7D).

32With these classifications, TS cells from ETX and iETX embryos clustered together with ExE as a group, although differentiating TS cells were most different from ExE at E5.5 and E6.5. ESCs from the ETX and iETX embryos clustered together with EPI. Subdivision of the endodermal lineages into these subpopulations shed further light on the relationship between the XEN layer, the VE-like layer, and the VE. The XEN layer samples were most similar to the parietal endoderm of embryos. By contrast, the VE-like layer correlated with the VE and differed from E5.5 and E6.5 parietal endoderm. This suggests that while induced ESCs are not identical to the natural VE, they are a marked improvement over standard XEN cells in recapitulating the natural gene-expression signature (Figure 7E).

33Differential gene-expression analysis using Wilcoxon rank sum test between the VE-like layer and the VE indicated that 25 genes were downregulated and 71 genes were expressed at higher levels in the VE-like layer of iETX embryos in comparison with the VE (Figure 7F). Genes with higher expression in iETX embryos included Smad2, Wnt11, and Id3, suggesting differences in Nodal, Wnt, and Bmp signaling, and Igfbp2, which may indicate differences in metabolic pathways. Comparison of the VE-like layer to the XEN layer (Figure 7G) revealed that the former expressed several regulators of AVE and embryonic VE formation, including Hhex, Hexs1, Gsc, Lhx1, and Afp. Consistent with our previous observations, we also detected Eomes, Lefty1, and Dkk1. The VE-like layer expressed a higher level of Nodal than the XEN layer. Gene ontology analysis of the genes upregulated in the VE-like layer of iETX embryos also indicated the presence of regulators of the Wnt pathway (Figures S7B–S7E; Table S2). Comparison of the VE with the XEN layer of ETX embryos showed differences in the same genes identified when we compared the VE-like layer with the XEN layer (Figures S7F–S7H), once again indicating the high degree of similarity between VE and VE-like layer. In summary, the VE-like layer of iETX embryos expresses a number of crucial regulators of VE identity, AVE specification, as well as factors regulating mesoderm formation and gastrulation, which could increase the developmental potential of iETX embryos (Figure 7H).

34Stem cell models of embryo development hold the promise to streamline the study of uncharacterized genes, novel drugs, and developmental pathways because of their modular design and ease of separately manipulating each compartment. Here, we have developed a model of mouse embryo development using iETX embryos, which builds upon, and significantly improves, our previously reported ETX embryos.

35Morphologically, iETX embryos are very similar at days 4 and 5 to natural embryos at E5.5 and E6.5 and express the canonical markers of post-implantation development at these stages. Importantly, iETX embryos can induce formation of AVE, as indicated by the expression of Cerl, Lefty1, and Dkk1, and this AVE is able to migrate asymmetrically toward the boundary of ES- and TS-derived compartments. Anterior and posterior domains are specified on opposite sides and the embryo-like structure then undertakes gastrulation with formation of embryonic and extra-embryonic mesoderm and definitive endoderm. These processes occur reproducibly in the great majority of the structures examined and make this a valuable system to study AVE induction and gastrulation in vitro.

36At a technical level, iETX embryos are easier to generate because they grow in standard, nonconditioned tissue-culture media and do not require the establishment of novel XEN lines, but existing ES lines can be modified for this assay. Our single-cell sequencing data show that replacing XEN cells with tetOG4 ESCs is beneficial because they make a VE-like layer, which is more similar to the natural VE and expresses several regulators of AVE development, potentially accounting for the broader developmental potential of iETX embryos.

37One current limitation is the observed downregulation of Lefty1 at day 5. Since Lefty1 is a direct Nodal target, this may mean that Nodal signaling is not sustained enough at that point. Our data also suggest that expression of Cerl, Lefty1, and Dkk1 in the AVE may be differentially regulated, but this needs to be tested in the future. Second, when the AVE is not positioned correctly, posteriorizing signals from the TS compartment induce mesoderm formation on both sides of the ES compartment. One of the two sides is consistently more advanced than the other, suggesting that even incomplete migration of the AVE is sufficient to set up an asymmetry that favors one posterior domain over the other. Finally, iETX embryos cannot be cultured beyond 6 days; thus, improved culture conditions will be needed to test their developmental potential further.

38We conclude that the state of the VE precursor cells plays a considerable part in defining the extent of development of synthetic embryo structures. Here, by expressing Gata4 in ESCs, we generate VE-like cells that are able to undertake considerably more extensive AVE development and PS formation at gastrulation in iETX embryos and, therefore, provide a better building block for these post-implantation structures. We show that this is a consequence of the greater similarity of these cells to natural VE than XEN cells, since they express important regulators of VE as well as crucial regulators of mesoderm formation and gastrulation. These different genetic signatures can account for the different developmental potential of these two stem-cell-based systems.

39Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Magdalena Zernicka-Goetz (mz205@cam.ac.uk).

40All unique/stable reagents generated in this study are available from and will be provided by the Lead Contact with a completed Materials Transfer Agreement.

41The accession number for the single-cell sequencing data reported in this paper is Gene Expression Omnibus: GSE134240 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE161947).The code used in these analyses is available at https://github.com/fhlab/scRNAseq_inducedETX.

42Cell lines used in this study include:CAG-GFP/tetO-mCherry mouse ESCs (constitutive GFP expression in the membrane; transient mCherry expression upon Dox treatment). The parent CAG-GFP/tetO-mCherry ESC line was derived from an existing mouse line with constitutive CAG-GFP expression and Dox-induced transient mCherry expression. This line was generated by breeding CAG-GFP reporter mice and tetO-mCherry Histone mice. For the purpose of this study, an independent Dox-inducible Gata4-expressing cassette was introduced into the CAG-GFP/tetO-mCherry ES lines by piggyBac-based transposition, as described below, thus mCherry and Gata4 are regulated by two, independent Dox-responsive promoters.CAG-GFP/tetO-mCherry/tetO-Gata4 ESCs generated in-house.Cerl-GFP mouse ESCs (GFP expression under the control of the Cerl-promoter) were derived from a published Cerl-GFP mouse line.Cerl-GFP/tetO-Gata4 ESCs generated in-house.Bry-GFP ESCs.Snail-YFP ESCs were derived from Snail-YFP transgenic mice, which were a generous gift of Dr. Robert Weinberg (Whitehead Institute for Biomedical Research USA) and Dr. Michaela Frye (University of Cambridge, UK).Mouse Confetti TS cells were a generous gift of Prof. Jenny Nichols (Stem Cell Institute, Cambridge, UK). Because they were not treated with Tamoxifen, they did not express any reporter.Wildtype TS cells were generated in house from CD1 mice.

43CAG-GFP/tetO-mCherry mouse ESCs (constitutive GFP expression in the membrane; transient mCherry expression upon Dox treatment). The parent CAG-GFP/tetO-mCherry ESC line was derived from an existing mouse line with constitutive CAG-GFP expression and Dox-induced transient mCherry expression. This line was generated by breeding CAG-GFP reporter mice and tetO-mCherry Histone mice. For the purpose of this study, an independent Dox-inducible Gata4-expressing cassette was introduced into the CAG-GFP/tetO-mCherry ES lines by piggyBac-based transposition, as described below, thus mCherry and Gata4 are regulated by two, independent Dox-responsive promoters.

44CAG-GFP/tetO-mCherry/tetO-Gata4 ESCs generated in-house.

45Cerl-GFP mouse ESCs (GFP expression under the control of the Cerl-promoter) were derived from a published Cerl-GFP mouse line.

46Cerl-GFP/tetO-Gata4 ESCs generated in-house.

47Bry-GFP ESCs.

48Snail-YFP ESCs were derived from Snail-YFP transgenic mice, which were a generous gift of Dr. Robert Weinberg (Whitehead Institute for Biomedical Research USA) and Dr. Michaela Frye (University of Cambridge, UK).

49Mouse Confetti TS cells were a generous gift of Prof. Jenny Nichols (Stem Cell Institute, Cambridge, UK). Because they were not treated with Tamoxifen, they did not express any reporter.

50Wildtype TS cells were generated in house from CD1 mice.

51The sex of the cell lines is not known because we did not genotype them to determine it.

52All cell lines were routinely tested every two weeks to ensure that they were not contaminated with mycoplasma.

53Mouse embryonic stem cells were cultured on gelatinised plates at 37°C, 5% CO2, 21% O2 in N2B27 which is comprised of 50% Neurobasal-A (Gibco 10888022), 50% DMEM/F-12 (Gibco 21331020), 0.5% N2 (in-house), 1% B27 (Gibco 10889038), 2mM GlutaMAX (Gibco 35050038), 0.1mM 2-mercaptoethanol (Gibco 31350010) and 1% penicillin/streptomycin (Gibco 15140122). N2B27 was supplemented with 3μM CHIR99021 (Cambridge Stem Cell Institute), 1μM PD0325901 (Cambridge Stem Cell Institute) and 10 ng ml-1 leukaemia inhibitory factor (Cambridge Stem Cell Institute). Mouse trophoblast stem (TS) cells were cultured on mitotically inactivated mouse embryonic fibroblasts (MEFs, Insight Biotechnology, ASF-1201) in feeder cell (FC) medium which contained Dulbecco’s modified essential medium (Gibco 41966052), 15% foetal bovine serum (Cambridge Stem Cell Institute), 1mM sodium pyruvate (Gibco 11360039), 2mM GlutaMAX (Gibco 35050038), 1% MEM non-essential amino acids (Gibco 11140035), 0.1mM 2-mercaptoethanol (Gibco 31350010) and 1% penicillin/streptomycin (Gibco 15140122). FC medium was supplemented with 1 μg ml-1 heparin (Sigma-Aldrich H3149-25KU), 25 ng ml-1 FGF2 (Cambridge Stem Cell Institute) and 25 ng ml-1 FGF4 (R&D Systems 7486-F4-025) (FC F42H). Passaging of ES and TS cells was performed when they were at 70% confluency as follows: cells were washed once in 1x PBS (Life Technologies 10010056) and trypsinised (Trypsin-EDTA 0.05% Life Technologies 25300054) for 4 minutes at 37°C. The reaction was stopped by adding 2 mL of FC, cells were dissociated by pipetting gently 4-5 times and centrifuged for 4 minutes at 200 x g. TS cells were then resuspended in FC F42H culture media and plated onto MEF-coated plates in 1:20 dilution. ESCs were washed once with 1 mL of 1x PBS, centrifuged again, resuspended in N2B27 2iLIF and plated at 1:10 or 1:20 onto gelatine-coated plates.

54Mice were handled following national and international guidelines. All experiments performed were under the regulation of the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 and were reviewed by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB). Experiments were approved by the Home Office. Animals were inspected daily and those showing signs of any health concern or condition were promptly culled by cervical dislocation. All experimental mice were free of pathogens and were on a 12-12 hour light-dark cycle, with unlimited access to water and food. Temperature in the facility was controlled and maintained at 21°C. Mice for post-implantation embryo recovery (CD-1 females and males from Charles River, acclimatised for 1 week prior to use) were utilised from 6 weeks of age. Female and males were naturally mated and kept together for up to five days or until a plug was found; females were inspected daily for plugs. Females were culled by cervical dislocation 5.5, 5.75, 6.5, 6.75 or 7.5 days after a plug was found. Embryos were dissected out of the deciduae in M2 medium (Sigma M7167). For chimera experiments, F1 females (Charles River, 1 week of acclimatisation prior to use) at 5-6 weeks of age were super-ovulated by injection of 7.5 IU of pregnant mares’ serum gonadotropin (Intervet) and 7.5 IU of human chorionic gonadotropin (Intervet) after 48 hours and were mated with F1 males (Charles River, 1 week of acclimatisation prior to use). Pregnant F1 females were culled at E2.5 by cervical dislocation to recover embryos by uterine and oviduct flushing in M2 medium.

55To prepare the AggreWell plate (STEMCELL Technologies 34415), 500μl of anti-adherence rinsing solution (STEMCELL Technologies 07010) was added to each well. The plate was then centrifuged at 2,000 x g for 5 minutes and was incubated for 20 minutes at room temperature. Rinsing solution was then aspirated from the well and 1ml of PBS was added to wash each well. 500μl of FC medium was added to each well after aspirating the PBS.

56To prepare ESCs for generating ES aggregates, Doxycycline (1 μg/mL) (Sigma-Aldrich D9891-5G) was added to CAG-GFP tetO-Gata4 ESCs 6 hours prior to plating in AggreWell. ESCs were washed once with 1x PBS, and trypsinised with 0.05% trypsin-EDTA (ThermoFisher Scientific) for 4 minutes at 37°C. The reaction was stopped by adding 2 mL of FC. Cells were dissociated gently by pipetting for 4-5 times and centrifuged at 200 x g for 4 minutes. The cell pellet was washed once with 1x PBS, centrifuged again and resuspended in 1-2 mL of FC. Cell suspensions containing 1) 12,000 Doxycycline-treated CAG-GFP tetO-Gata4 ESCs, or 2) 12,000 untreated CAG-GFP tetO-Gata4 ESCs, or 3) a mixture of 6,000 Doxycycline-treated and 6,000 untreated CAG-GFP tetO-Gata4 ESCs were pelleted again by centrifugation. After resuspending in 1ml of FC medium, the cell suspension was added dropwise to the AggreWell and the plate was centrifuged at 100 x g for 3 minutes. 1ml of fresh FC medium was added each day after removing 1ml of medium from the well. ES aggregates were collected and fixed after 1 day or 3 days.

57To generate iETX embryos, Doxycycline was added to CAG-GFP tetO-Gata4 ESCs for 6 hours. TS cells were trypsinised and were added to a gelatinised plate to deplete the MEFs for 20 minutes at 37°C. CAG-GFP WT ESCs and CAG-GFP tetO-Gata4 ESCs were subsequently trypsinised. Cell suspensions with 19,200 TS cells, 6,000 CAG-GFP WT ESCs and 6,000 CAG-GFP tetO-Gata4 ESCs were mixed and pelleted by centrifugation. The cell pellet was resuspended in 1ml of FC medium with 7.5 nM ROCK inhibitor (Y27632, STEMCELL Technologies 72304). After adding the cell mixture dropwise to the AggreWell, the plate was centrifuged at 100 x g for 3 minutes. On the next day, media change was performed twice by removing 1ml of medium from each well and adding 1ml of fresh FC medium without ROCK inhibitor. On day 2, media change was performed once to replace 1ml of medium with 1ml of fresh FC medium. On day 3, 1.2 ml of medium was removed from each well and 1.5ml of IVC1 was added, after equilibrating for 20 minutes in the incubator. IVC1 is made of advanced DMEM/F12 (Gibco, 21331-020) supplemented with 20% (v/v) FBS, 2 mM GlutaMax, 1% v/v penicillin–streptomycin, 1X ITS-X Thermo Fisher Scientific, 51500-056), 8 nM β-estradiol, 200 ng/ml progesterone and 25 μM N-acetyl-L-cysteine. On day 4, iETX embryos in the AggreWell were transferred to CELLSTAR 6 well multiwell plate for suspension culture (Greiner Bio-One 657185) with 5ml of IVC1 (with FBS at 30% v/v) per well. On day 5 IVC1 was replaced with fresh IVC2 (30% Knockout Serum Replacement instead of FBS, Thermo Fisher 10828010).

58To generate chimeras using CAG-GFP tetO-Gata4 ESCs and mouse embryos, Doxycycline was first added to CAG-GFP tetO-Gata4 ESCs 6 hours prior to the experiment. Mouse embryos at E2.5 before compaction were recovered from F1 females that were super-ovulated by injection of 7.5 IU of pregnant mares’ serum gonadotropin (Intervet) and 7.5 IU of human chorionic gonadotropin (Intervet) after 48 hours and were mated with F1 males. Embryos were recovered in M2 medium by flushing the oviducts. After transferring to KSOM (Millipore MR-020P-5F), the embryos were cultured in the incubator at 37°C in 5% CO2. Meanwhile, CAG-GFP tetO-Gata4 ESCs were dissociated by 2 minutes of trypsinisation at 37°C and the resulting cell clumps were aggregated with the recovered embryos in KSOM. The chimeras were cultured for 48 hours until E4.5.

59iETX embryos were collected at day 4 and incubated in SB431542 (STEMCELL Technologies 72234) for 24 hours at a concentration of 10 μM. Following Nodal inhibitor treatment, they were fixed and processed for immunofluorescence (below).

60Live imaging was performed using an SP8 scanning confocal microscope with a 25X objective. iETX embryos were imaged on a glass-bottom dish and were kept in a humidified chamber with 5.6% CO2 and 21% O2 during the imaging. Images were captured every 20 minutes (AVE migration and gastrulation movies) with a z-step of 2 μm. For self-organization movies, cells were imaged after seeding on Gri3D PEG-hydrogel dishes provided by SunBioscience (Geneva, Switzerland), set up according to the manufacturer’s guidelines and samples were imaged every 60 minutes. Samples were imaged on a Zeiss Axiovert 200M connected to a 3i CSU-W Spinning Disk Confocal system with an OBIS 488nm and an OBIS 561-nm LS laser and exported using Slidebook.

61Gata4 cDNA was PCR-amplified from pSAM2-mCherry-Gata4 using the Gata4/AttB primers (see Key Resources Table). The primers were designed as outlined in the Gateway cloning manual. Because the plasmid already contained attB sites, there was no need to incorporate parts of the Gata4 open reading frame in the primer design. pSAM2-mCherry-Gata4 was a gift from Timothy Kamp (Stem Cell and Regenerative Medicine Centre, University of Wisconsin – Madison, USA; Addgene plasmid # 72690; http://n2t.net/addgene:72690; RRID:Addgene_72690,). It was subsequently cloned into PB-tetO-hygromycin by Gateway technology (Thermo Fisher Scientific), according to the manufacturer’s instructions. Clones were verified by sequencing. Transformations were performed using 5α-competent E.coli following the manufacturer’s guidelines (New England Biolabs C2987I). To generate ESCs with Doxycycline inducible Gata4, PB-tetO-hygro-Gata4, pBAse and rtTA-zeocyin (0.25 μg/each/reaction) were transfected into 12,000 CAG-GFP ESCs using Lipofectamine 3000 Transfection Reagent (Invitrogen L3000001), followed by antibiotics selection for 7 days with hygromycin (1:250; Gibco 10687010) and zeocyin (1:1000; InvivoGen ant-zn-1). The PB-tetO-hygro, pBAse and rtTA-zeocyin were generously gifted by Dr. José Silva (Stem Cell Institute, Cambridge, UK).

62Total RNA from cell pellet was extracted using Trizol Reagent (Invitrogen 15596-026) and reverse transcribed into cDNA using QuantiTect Reverse Transcription Kit (Qiagen 205310) according to the manufacturer’s instructions. qRT-PCR was performed using SYBR Green PCR Master Mix (Applied Biosystems 4368708) and StepOnePlus Real-Time PCR System (Applied Biosystems). Fold change in Gata4 mRNA expression was determined by ΔΔCt method using Gapdh as endogenous control. See Table for Gapdh and Gata4 primer sequences.

63iETX embryos and natural mouse embryos were fixed with 4% paraformaldehyde at room temperature for 20 minutes and washed with PBST (PBS with 0.1% Tween 20) for three times for 5 minutes each. Samples were then permeabilised in permeabilization buffer (0.1 M glycine and 0.3% Triton X-100 in PBS) for 30 minutes at room temperature, followed by three washes with PBST for 5 minutes. Samples were incubated with primary antibodies diluted in blocking buffer (10% FBS and 0.1% Tween 20 in PBS) at 4°C overnight. After washing with PBST for three times for 5 minutes, samples were incubated with secondary antibodies at 4°C overnight or for 2 hours at room temperature followed by another three washes with PBST for 5 minutes before imaging.

64After recovery, natural embryos and iETX embryos were cut to pieces, transferred to a Falcon tube, centrifuged, washed in PBS and incubated in Tryple Express (Gibco 12604013) for 15 minutes at 37°C, with vigorous pipetting every 5 minutes to dissociate to single cells. If there were clumps left, the incubation was extended for an additional 5 minutes at 37°C and the sample was pipetted further. Samples were filtered to remove large clumps, centrifuged at 200 x g for 5 minutes and resuspended in PBST (PBS with 0.02% Tween20) and then processed for incapsulation (see below). For E5.5 and E6.5, 1 full litter was dissociated (12 embryos). For ETX embryos and iETX embryos, 15 samples each were dissociated. Single cell sequencing data from 20 E4.5 blastocysts was obtained from (, GEO: GSE134240).

65Libraries were prepared according to the inDrops workflow with v3 barcoding scheme. Briefly, polyacrylamide beads were generated and barcoded to obtain a diversity of 147,456 barcodes. Single-cell suspensions were diluted to a concentration of 120,000 cells per ml and co-encapsulated with the barcoded beads and reverse transcriptase and lysis mix. Fractions of ∼2,900 cells were collected in 1.5 ml Eppendorf tubes pre-filled with 200 μl of mineral oil and incubated at 50°C for 2 hours and 70°C for 20 minutes. The droplets were then de-emulsified and further amplified using second-strand synthesis and in vitro transcription. The libraries were then fragmented and reverse transcribed. The final libraries were amplified using limited-cycle PCR and quantified using a Qubit High sensitivity and Bioanalyzer High sensitivity DNA kits. Libraries were pooled at equi-molar ratios and purified using a 1.5x volumetric ratio of AmpureXP beads. The libraries were sequenced on a Nextseq 75 cycle 400M read High Output kit with 5% PhiX spike-in as an internal control. The read cycle distribution was the following: Read1: 61 cycles; Index1: 8 cycles; Index2: 8 cycles; Read2: 14 cycles.

66All iETX embryos were collected from AggreWell for analysis at 3, 4 or 5 days of development and analysed under a stereo microscope. In all instances, we selected iETX embryos with cylindrical morphology, an epithelialized ES compartment with a lumen and two clearly defined cellular compartments surrounded by an outer cell layer. The TS compartment is more variable in appearance and therefore, even though one would also want an epithelial-looking TS compartment similar to the extra-embryonic ectoderm of natural embryos, we select a wider range of appearances for the TS compartment. After this initial selection, structures containing the appropriate fluorophores were quickly checked under a microscope to ensure the presence of an epithelialized CAG-GFP-positive ES compartment. iETX embryos with the correct body plan of ES and TS compartments surrounded by a visceral endoderm-like layer are then transferred to equilibrated media to continue their culture. When selecting at day 5, however, we expect the lumen of the ES and TS compartment to be merged. We provide visual examples of what we consider good iETX embryos at day 4 in Figure 2B.

67Cell lineage quantifications were performed using the “Spot” function in Imaris (Bitplane). For the VE-like layer, the spots were set at a diameter of 9 μm. For ESCs and TS cells the spots were set at a diameter of 6 μm. Following automatic detection, the spots were manually curated to remove erroneous cell calls and to include cells that were missed. Because of the loss of resolution past the midpoint of the iETX embryos, cells were only quantified up to the midpoint of each iETX embryo and then the number of cells obtained was multiplied by 2 to obtain the total number of cells.

68Images were acquired using Leica SP5 and SP8 confocal microscopes (Leica Microsystems) with 40× oil objective and 25× water objective, respectively. A 405 nm diode laser (DAPI), a 488 nm argon laser (Alexa Fluor 488), a 543 nm HeNe laser (Alexa Fluor 568) and a 633 nm HeNe laser (Alexa Fluor 647) were used to excite the fluorophores. Images were taken with a z-step of 1.2-5μm. Fiji and NDSAFIR 3.0, Photoshop and the Smart Denoise (Gurdon Institute) were used to process and analyse the images. In Figure 4A, the Bry antibody in the natural embryo shown had strong non-specific membrane signal, which has been reported elsewhere. To reduce the background signal, we used the Oct4 signal in that image to create a mask of the epiblast and subtract the Bry background from the nuclear signal. This is the only case in which we have edited an image this way.

69Images from time lapse video were processed with Fiji. To correct for the drifting of iETX-embryo during imaging, images were first aligned using StackReg plugin, MultiStackReg (Brad Busse, Division of Program Coordination, Planning and Strategic Initiatives, NIH - USA) and Template Matching and Slice Alignment plugin (Qingzong Tseng, Aix-Marseille Université, France). The movements of individual Cerl-GFP-positive cells were then tracked with Manual Tracking plugin which generated the migration trajectories. The plugin also produced a table of the coordinates of the tracked cells in each time frame which was later imported to Chemotaxis and Migration Tool (iBidi) to calculate the migration directionality and migration speed. The polar histogram of migration directionality was generated using R.

70In Fiji, a rectangle encompassing the length of the ES compartment and wide enough to contain all the Cerl-GFP positive cells was drawn over a day 5 iETX embryo. Using the plot function, signal intensity as a function of distance was obtained and exported. The same area was sequentially used to export Cerl-GFP, Eomes, Phalloidin and DAPI intensities for each iETX embryo. If a single Cerl-GFP peak was observed on one side of the iETX embryo, the Cerl-GFP signal was considered asymmetric, otherwise it was considered symmetric. To normalize the length of the structure, each point of the length was divided by the total length of the ES compartment. To normalize the fluorescence, each fluorescence value was divided by the highest fluorescence value. Normalised samples were combined (asymmetric with asymmetric and symmetric with symmetric) and displayed as a Lowess curve generated in Prism.

71The BCL files were converted to Fastq files using Illumina’s bcl2fastq software. The sequenced libraries were quality-inspected using the FastQC tool and de-multiplexed using the Pheniqs tool from biosails. The fastq files were further filtered, mapped to a mouse GRCm38 reference genome with GRCm38.99 gtf annotation and deduplicated using the zUMIs pipeline. The count matrices were then imported in Scanpy using the scanpy.read function. The Scrublet module was used to predict doublet scores. Cells with predicted doublet scores lower than 0.2 and with gene counts higher than 1,200 were then selected. The filtered matrices were then concatenated after being converted to a sparse format using the csr_matrix function from SciPy. Cells were further filtered on ribosomal RNA (percentage of reads mapping to ribosomal RNA between 2.5% and 10%) and mitochondrial RNA content (percentage of reads mapping to mitochondrial RNA between 2% and 12%). The matrix was then normalized, regressed out for number of UMI counts, ribosomal RNA and mitochondrial RNA content using the scanpy.pp.regress_out function, scaled and a UMAP dimensional reduction was computed. To plot the correlation between ETX, iETX, E4.5, E5.5 and E6.5 and to match the sample size and number of cells per lineage of E4.5 (E4.5 dataset was from), each sample was downsampled to randomly contain 41 cells from the endoderm lineage, 33 cells from the embryo lineage and 22 cells from the trophectoderm lineage. Correlation matrices were computed with scanpy.pl.correlation_matrix function using Pearson correlation. For cell type annotation, a combination of leiden clustering (obtained using scanpy.tl.leiden) and a marker-based approach based on Pijuan-Sala et al.’s annotations was utilised with the following normalized marker expression levels: 1) parietal endoderm annotation for iETX cells of the endoderm compartment with Lamb1 expression higher than 3.5, 2) primitive streak annotation for cells of the embryo compartment with Nanog and Eomes expression level higher than 1.5, 3) nascent mesoderm annotation for cells of the embryo compartment with Mesp1 expression level higher than 1.5, 4) Visceral endoderm cells from iETX embryos with Otx2 expression level higher than 1.5, 5) ExE and differentiating ExE annotations were given according to the resulting leiden clustering and also using Cdx2 for the trophoblast stem cells/extraembryonic ectoderm cells and Gata2 for the differentiated trophoblast cell types (see Figure S7A).

72To perform pairwise analysis of differentially expressed genes, the matrices and annotations were loaded into Seurat v3 and pairwise marker comparisons were obtained with a Wilcoxon Rank Sum test using the FindMarkers function. Differentially expressed genes were then plotted in Prism GraphPad to generate volcano plots. Single-cell sequencing data are available on GEO (Gene Expression Omnibus) under accession GSE161947 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE161947) and the code used in these analyses is available at https://github.com/fhlab/scRNAseq_inducedETX.

73Gene Ontology was performed using the online platform DAVID. Differentially expressed genes from pairwise comparisons were selected by choosing genes with an adjusted p value < 0.05 and enriched in one sample or the other of the pairwise comparison. The list was then uploaded in the DAVID user interface and analysed with the Gene Functional Annotation Clustering tool and the Gene Functional Annotation Table. The first 10 clusters with the highest Enrichment Score (-log p value) were graphed. For the Wnt signalling category identified as enriched in the VE-like layer in comparison with the XEN layer, Wnt-related function of the genes in the list was found using the annotations of the and provided in Table S2.

74All statistical analyses were performed with GraphPad Prism 8 software. Quantitative data were presented as mean ± SD or SEM as indicated in figure legends or as violin plots with median and quartiles. Prior to statistical significance testing, data were tested for normal distribution with the Shapiro-Wilk test. For normally distributed data, the unpaired or paired Student’s t test was used for 2 groups and One-Way ANOVA with Tukey’s multiple comparison post-hoc test for more than 2 groups. For data that did not follow a normal distribution, Mann-Whitney non-parametric test was used for 2 groups and One-Way ANOVA followed by Kruskal-Wellis non-parametric post-hoc test for more than 2 groups. Wilcoxon matched pairs signed rank test was used for non-parametric paired analysis. A p value < 0.05 was considered significant. Sample size and the number of experimental replicates (n) is indicated in the relevant figure legend. Sample size was not predetermined. For the supplemental figure, the statistical test utilised and the exact p values are in the relevant figure legend. For the main figures, the statistical test utilised and the exact p values are shown below:

75Figure 1: A. Unpaired Student’s t test: N2B27 2iLIF: p = 0.0218, FC 2iLIF: p < 0.0001, IDG 2iLIF: p = 0.0020. B. One-way ANOVA, Kruskal-Wallis post-hoc test, p < 0.0001. D. Unpaired Student’s t test, p < 0.0001

76Figure 2: C. Unpaired Student’s t test p=0.0052. J. One-Way ANOVA; ES compartment, Epiblast (ES/EPI): Tukey post-hoc. ∗∗∗∗ p < 0.0001. TS compartment, Extraembryonic ectoderm (TS/ExE): Tukey post-hoc.∗ p = 0.0236. XEN layer, VE-like layer and visceral endoderm (XEN/VE): Kruskal-Wallis post-hoc.∗∗ p = 0.0077, ∗∗∗∗ p < 0.0001.

77Figure 3: E. Unpaired Student’s t test p = 0.0071. F. Unpaired Student’s t test p = 0.0478. G. Unpaired Student’s t test p = 0.0079. H. Mann-Whitney test p = 0.0003. I. Mann-Whitney test p = ns. J. Student’s t test p = 0.0407. L. One-Way ANOVA p < 0.0001, Kruskal-Wallis post-hoc test. ∗∗∗∗ p < 0.0001, ∗∗ p = 0.0033. N. Unpaired Student’s t test p = 0.033. O. Unpaired Student’s t test p = ns. P. Unpaired Student’s t test p = 0.0194.

78Figure 4: D. One-way ANOVA, p < 0.0001; Tukey’s multiple comparison post-hoc test: opposite vs Same and Distal, p < 0.0001; Distal vs Same p = ns. F. One-Way ANOVA p<0.0001; Tukey’s multiple comparison post-hoc test: opposite vs Same and Opposite vs Distal p < 0.0001, Same vs Distal p = ns. H. One-Way ANOVA p < 0.0001; Tukey’s multiple comparison post-hoc test: opposite vs Same and Opposite vs Distal p < 0.0001, Same vs Distal p = ns. J. One-Way ANOVA p = 0.0002; Tukey’s multiple comparison post-hoc test: ectopic vs non-ectopic p = 0.0011; ectopic vs. distal p = 0.0007.

79Figure 5: E. Multiple t tests, p < 0.000001. H. Unpaired t test: Bry p < 0.0001, N-Cad p = 0.0049, E-Cad p = 0.0207. I. (left) One-Way ANOVA and Tukey multiple comparison, ∗∗∗∗ p < 0.0001, ∗∗ p = 0.0022. (right) Wilcoxon matched pairs signed rank test, ∗ p = 0.0178.