Epiblast inducers capture mouse trophectoderm stem cells in vitro and pattern blastoids for implantation in utero

The embryo instructs the allocation of cell states to spatially regulate functions. In the blastocyst, patterning of trophoblast (TR) cells ensures successful implantation and placental development. Here, we deﬁned an optimal set of molecules secreted by the epiblast (inducers) that captures in vitro stable, highly self-renewing mouse trophectoderm stem cells (TESCs) resembling the blastocyst stage. When exposed to suboptimal inducers, these stem cells ﬂuctuate to form interconvertible subpopulations with reduced self-renewal and facilitated differentiation, resembling peri-implantation cells, known as TR stem cells (TSCs). TESCs have enhanced capacity to form blastoids that implant more efﬁciently in utero due to inducers maintaining not only local TR proliferation andself-renewal,but also WNT6/7Bsecretion thatstimulatesuterinedecidualization. Overall, theepiblastmain-tains sustained growth and decidualization potential of abutting TR cells, while, as known, distancing imposed by the blastocyst cavity differentiates TR cells for uterus adhesion, thus patterning the essential functions of implantation. sequence S7D), as well as a Cas9, and BFP sequence. 10 m g of plasmid were used to transfect 1x10 6 cells. BFP positive cells were FACS sorted the next day. After their expansion, BFP negative cells were FACS sorted individually onto inactivated MEFs for clonal expansion and cultured in TXV medium. MEFs were depleted and cells were genotyped by Sanger sequencing. The effect of genomic editing on gene expression was conﬁrmed by qRT-PCR using primers that amplify the 3 0 UTR and Western Blot (Figure S7E and S7F). Sequence information for all primers and sgRNAs used in this study are available in Methods S3.

Correspondence nicolas.rivron@imba.oeaw.ac.at In brief Here, Seong et al. identify an optimal set of epiblast inducers that captures mouse trophectoderm stem cells (TESCs) as a stable and highly self-renewing state reflecting the blastocyst stage. TESCs have enhanced capacity to form blastoids that induce deciduae formation more efficiently in utero due to WNT6/7B secretion.

INTRODUCTION
To implant into the uterus, mammals form a blastocyst comprising an inner epiblast (EPI) that later forms the body, surrounded by a fluid-filled, epithelial cyst called trophectoderm (TE) composed of trophoblasts (TRs) that will form the placenta. The EPI is placed asymmetrically in the TE cyst, a positioning that defines the first developmental axis (polar-mural, a.k.a. embryonic-abembryonic). Along this axis, the TE abutting the EPI (polar TE) proliferates and self-renews to progressively build the placenta into a composite organ that fulfills crucial functions (e.g., gas exchange, excretion of waste products, and functions that are immunological). The separation from the EPI, on the other hand, is associated with TE differentiation, decreased proliferation, and the capacity to attach to the uterus (Copp, 1978;Das et al., 1994;Gardner, 2000;Klaffky et al., 2001). Thus, one function of the blastocyst cavity is to separate the pools of TRs that will form the placenta or mediate the initial uterus attachment. In response to blastocyst implantation, uterine cells proliferate and undergo functional changes to create the decidua, a cocoon required for a successful pregnancy. Faulty decidualization is associated with infertility and miscarriages (Cha et al., 2012).
The TR lineage develops in a different way as compared to the EPI lineage: while the EPI lineage progress through pluripotency via a sequence of irreversible and epigenetically engraved lock steps with different signaling requirements (Nichols and Smith, 2012), TR progenitors self-renew in vivo thanks to a niche environment generated at the EPI/TR interface. This environment is currently defined by a number of soluble (Guzman-Ayala et al., 2004) and extracellular matrix (Kiyozumi et al., 2020) molecules that promote proliferation while inhibiting precocious differentiation, thus maintaining self-renewal. FGF4 is a niche factor produced by the EPI starting in mid-stage blastocysts (E3.25; Guo et al., 2010;Ohnishi et al., 2014) whose importance is evidenced by Oct4 À/À mutant blastocysts that fail to express FGF4 but respond to exogenous FGF4 by increasing TR proliferation (Nichols et al., 1998). FGF4 induces the phosphorylation of ERK in some of the polar TE cells (E3.25-4.5) (Azami et al., 2019), which promotes proliferation and prevents differentiation and apoptosis (E4.5-5.0) through FGF receptors (Chai et al., 1998), FRS2 (Gotoh et al., 2005), SHP2 (Yang et al., 2006), and the signaling effector kinase ERK2 (Hatano et al., 2003;Saba-El-Leil et al., 2003). During implantation, a tissue descending from the polar TE, the extraembryonic ectoderm (ExE), maintains the TR progenitors and has a continuous requirement for ERK signals between at least E4 and E8 (Corson et al., 2003;Azami et al., 2019). TGFb superfamily members also contribute to the post-implantation maintenance of TR progenitors through activity of the Nodal receptor ACVR1B (Gu et al., 1998), the Nodal pro-protein convertases PCSK6 (PACE4) (Constam and Robertson, 2000) and Furin (Roebroek et al., 1998), and the effectors SMAD1 (Aubin et al., 2004) and SMAD2 (Weinstein et al., 1998). Finally, the Hippo and Notch pathways are important regulators of the initial TE specification (Nishioka et al., 2009;Rayon et al., 2014), and IGF2 is expressed at high levels in the polar TE , ExE, ectoplacental cone (EPC), allantois, and chorion (Lee et al., 1990) and contributes to TR proliferation (Gardner et al., 1999;Constâ ncia et al., 2002;Zechner et al., 2002).
The EPI/TR interface is a dynamic and regulative environment: the EPI sustains FGF4 production that stimulates BMP4 secretion from the ExE (Lawson et al., 1999;Murohashi et al., 2010) that might act as a direct niche factor (; Graham et al., 2014;Rivron et al., 2018) but also indirectly induces the EPI to express Nodal and Wnts (E6) (Ben-Haim et al., 2006;Miura et al., 2010). Nodal and Wnt ligands act in autocrine ways on the EPI but also feed back onto the TR progenitors to maintain self-renewal in part by regulating FGF4 expression (Guzman-Ayala et al., 2004;Ben-Haim et al., 2006). Altogether, this interface is a hotspot of regulative interactions maintaining the TR progenitors that are thought to persist until the mid-gestation placenta (Guzman-Ayala et al., 2004;Ueno et al., 2013;Natale et al., 2017).
In the presence of FGF4, TR progenitors isolated either from the blastocyst (E4.5) or ExE (E6.5) can proliferate and self-renew in vitro as TR stem cells (TSCs) (Tanaka et al., 1998). This shows that these progenitors possess an intrinsic, long-term selfrenewal potential maintained by an EPI inducer. Here, we show that TSCs, which are notoriously heterogeneous (Sebastiano et al., 2010;Wu et al., 2011;Ohinata and Tsukiyama, 2014;Kuales et al., 2015;Motomura et al., 2016;Perez-Garcia et al., 2021), comprise fluctuating, interconvertible subpopulations resembling peri-implantation TRs. However, an optimal set of EPI inducers captures stable, highly self-renewing TRs resembling the blastocyst TE. We term these lines trophectoderm stem cells (TESCs). We show that TESCs not only maintain a high self-renewing capacity while inhibiting precocious differentiation but also secrete Wnt ligands that contribute to the decidual reaction in utero. Observing the behavior of TESCs and TSCs, we argue that optimal exposure to inducers sustains high self-renewal after implantation, while suboptimal exposure enables the concomitance of interconvertible subpopulations facilitating the exit from self-renewal and differentiation.
Altogether, this analysis highlights specific EPI inducers, TRs signaling pathway activities, and markers that define TR progression around implantation time.
By analyzing single TSCs and 6-day-differentiated TSCs, we obtained a different resolution of the heterogeneity profile. Unsupervised clustering analysis separated TSCs into four subpopulations and 6-day-differentiated TSCs into two subpopulations ( Figure 2G). Cells aligned along a Monocle-predicted pseudotime trajectory ( Figures 2H and S2E) (Trapnell et al., 2014) in which low pseudotime values corresponded to a TSC subpopulation with higher Cdx2 and Esrrb expression levels, while the next cells along pseudotime values showed higher expression of Elf5, consistent with ExE identity ( Figure 2H). Following in the trajectory were cells abundant for Tfap2c and Ascl2 expression and TFs marking the ExE/EPC (Guillemot et al., 1994;Auman et al., 2002;Werling and Schorle, 2002;Latos et al., 2015), while the highest pseudotime values marked cells expressing the TR differentiation markers Flt1 or Gcm1. Importantly, we identified differentiated TRs within TSCs (9%, Figure S2E), showing that FGF4/TGFb1 consent to spontaneous differentiation. We confirmed the presence of subpopulations using smFISH with cells rich in either Cdx2, Gsto1, and Esrrb or Ascl2, Gcm1, and Krt18 transcripts ( Figure S2F). This was further confirmed at the protein level (CDX2/KRT18, Figure S2G). Taken together, these analyses show that FGF4/TGFb1 consent to the maintenance of concomitant subpopulations reflecting the blastocyst TE, post-implantation ExE, and more differentiated TRs.
Trophoblast stem cells' subpopulations reflect functionally different and interconvertible stem cell states Next, we examined whether CDX2 high and CDX2 low TSCs represent functionally different states. Using flow cytometry analysis, we observed that CDX2 high cells were present in all cell-cycle phases, while CDX2 low cells were predominantly in the G0/G1 phases ( Figure 2I). Consistently, CDX2 high cells had a 30% clonogenicity rate, three times higher than that of CDX2 low cells (Figure 2J), which were more prone to differentiate upon removal of FGF4/TGFb1 (Pl1 and Synb mark TR giant cells and syncytiotrophoblasts, respectively; Figure 2K).
Protein fluctuation can occur stochastically but synchronously for members of the same biological pathway, which generates heterogeneity (Sigal et al., 2006). When we conducted live imaging of CDX2-eGFP TSCs, we observed transitions between the CDX2 high and CDX2 low states ( Figure S2I and Video S1) reminiscent of fluctuations in ESCs (Chambers et al., 2007;Hayashi et al., 2008;Hastreiter et al., 2018). Thus, we investigated whether subpopulations were capable of interconversion. Within 5 days after sorting of CDX2 high or CDX2 low cells, these subpopulations reestablished the initial heterogeneity ( Figure 2L), and analysis of multiple sortings over 50 days showed that the initial transcriptome was restored (Figures S2I and S2J). We concluded that these states are reversible and that heterogeneity is an intrinsic, possibly regulated property of TSCs under minimal conditions. However, the CDX2 low cells proliferated more slowly than the CDX2 high cells ( Figure S2K), suggesting a priming mechanism.
In ExE and TSCs, the Cdx2 locus is marked with high levels of the activating H3K4me3 mark, while these levels decrease upon differentiation (Rugg-Gunn et al., 2010). Accordingly, H3K4me3 levels at the Cdx2 promoter were 50% lower in CDX2 low TSCs, suggesting either reduced promoter activity or fewer active promoters (Figure S2L). In contrast, levels of the active transcription marker H3K9Ac were comparable between populations. We concluded that the Cdx2 locus remains accessible and can be reactivated, for example in response to an environmental factor such as a growth factor, but is subjected to reversible epigenetic regulations that prime CDX2 low TRs for differentiation. This transcriptomic and epigenetic reversibility is consistent with the reversibility of the mural TE, as shown by blastocyst microdissection/recombination (Gardner et al., 1973;Gardner, 1983). Overall, we concluded that FGF4/TGFb1 allow fluctuating, interconvertible, peri-implantation-like TR states with different proliferation and self-renewing potentials, a phenomenon that facilitates differentiation.
Exposure to optimal embryonic inducers generates stable CDX2 high trophoblast stem cells Next, we investigated the impact of combined EPI inducers. We tested individual molecules acting on pathways found active in the TE ( Figure S1D) for their capacity to induce CDX2 in TSCs (Figures 3A and S3A and Table S4). Nine molecules acted in a dose-dependent manner including IL11, Activin, BMP4/7, 1-Oleoyl Lysophosphatidic Acid (LPA) (Yu et al., 2012(Yu et al., , 2021Goto et al., 2015), 8-Br cAMP, XAV939, and the PPAR (H) Pseudotime heatmap for visualization of expression patterns along with marker genes for various differentiation states. For clusters, see also Figure S2E. (I) Cell-cycle analysis of TSCs shows a correlation between CDX2 expression (higher CDX2 content in red, lower in blue) and the cell-cycle state. (J) Colony formation potential of single CDX2 high and CDX2 low cells based on three independent experiments. Data are means ± SEM, analyzed by Student's t test.
(K) mRNA expression of differentiated marker genes in CDX2 high andCDX2 low TSCs. (L) Percentage of the different subpopulations upon pure subpopulation sorting and further independent culture, based on three independent experiments and a total of 96 wells of cells for each condition. Dark green for CDX2 high , gray for CDX2 intermediate , light green for CDX2 low . For each panel, *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S2. (F) GSEA for polar TE genes with TSCs exposed to optimal inducers for 15 days. (G) Pseudotime heatmap for visualization of expression patterns along with self-renewal, polar, differentiated, and classical TSCs marker genes. Most selfrenewal and polar genes are enriched in TXV TSCs compared to TX or differentiated TSCs. See also Figure S3.
Epiblast inducers enhance the epithelial phenotype and morphogenetic potential of trophoblast stem cells The transcriptome of TXV TSCs was enriched in epithelia transcripts including those related to extracellular matrix organization, cell adhesion, pathways related to ECM-receptor interaction, focal adhesion, cytoskeleton, and tight junctions (e.g., Cldn4, Cldn6, Tjp2, and Jam2) that contribute to TE morphogenesis (Moriwak et al., 2007) ( Figure S3O). By quantifying phenotypic changes of single-cell morphologies using E-CADHERIN/ Hoechst-staining segmentation, we extracted 161 morphometric features from both TX TCSs (502 cells) and TXV TSCs (297 cells) ( Figure S5A). After being ranked based on the p value scores (Mann-Whitney; Methods S1), the top 20% morphometric features separated the two populations (Figures S5B and S5C). TXV TSCs had significantly larger cell size and nuclei areas and were more circular and less lobulated ( Figure S5D). Cells reflecting the TE should efficiently undergo epithelial morphogenesis. As compared to TX TSCs, TXV TSCs more efficiently formed blastoids that were larger and more circular ( Figures 5A-5D and S5E). This was partly due to an enhanced autonomous potential measured by a higher capacity to Article form trophospheres (cavitation efficiency and circularity) ( Figures S5F-S5H). This was also due to an enhanced response to ESCs as TXV TSCs formed blastoids that had a larger diameter, a feature that did not occur in trophospheres (Figures 5B and S5G). Because TX and TXV TSCs proliferate at similar rates in blastoids ( Figure S5I), we concluded that the increased diameter was due to enhanced swelling in response to inducers. Blastoids correctly localized the basal adherens junctions (E-CAD-HERIN [E-CAD]) and apical cytoskeletal protein (KRT8/18) ( Figure 5E). Taken together, these observations show that TXV TSCs have an enhanced epithelial phenotype and morphogenetic functions, consistent with the TE.
Cumulatively, we showed that TSCs cultured in TXV (1) enhance their transcriptomic similarity to the TE, (2) enhance their self-renewal, (3) repress gene expression associated with differentiation, (4) maintain their potential to rapidly differentiate and chimerize the ExE, and (5) have an enhanced potential to recapitulate features of TE epithelial morphogenesis. Therefore, we call these cells TESCs.
Epiblast inducers spatially pattern the polar-mural axis Next, we asked whether inducers spatially pattern the polarand mural-like states that confer specific functions during implantation. Single TRs isolated from blastoids formed a pseudotime trajectory ( Figure S5J) that included three transitioning clusters ( Figure S5K). Cells with low pseudotime values more abundantly expressed transcripts for Esrrb, Cdx2, and Ly6a (Figures 5F and 5G) and numerous polar genes (Ly6a, Gsto1, Ddah1, Utf1, and Duox2). In contrast, the cluster with the highest pseudotime value showed enhanced expression levels of mural markers (Krt8/18, Ndrg1, Basp1, Ctsb, Flt1, and Slc5a5) and of Lgals1, which mediates endometrial interaction (Sood et al., 2006;Shi et al., 2013;Barrientos et al., 2014) (Figure 5F). Using smFISH, we confirmed that the polar marker Ly6a was more prominently expressed in the blastoid polar cells (7/10) (Figures 5H and S5L). At the protein level, the axis defined by CDX2 and KRT8/18 formed at a low efficiency similar to that in previously reported blastoids formed either with ESCs (Rivron et al., 2018) or with EPSCs (Sozen et al., 2019) ( Figure S5M). However, this frequency increased when FGF4/TGFb1 were removed from the medium (45.5% versus 28.6% for CDX2 and 73.2% versus 50.8% for KRT8/18, see method details) ( Figures 5I, 5J, S5M, and S5N). We concluded that the absence of these molecules leaves the blastoid EPIlike cells as the main source of positional information, which facilitates mural differentiation. Accordingly, when exposed to TXV molecules, blastocysts maintained ELF5 and LY6A expression in the mural cells, which were less able to become KRT8/ 18 high (Figures 5K and S5O). Collectively, these data indicate that EPI inducers act locally to pattern the TE axis.
Proximity to inducers maintains the trophectoderm decidualization capacity After the initial attachment ($E4.5-5.0), the blastocyst instructs the uterus to form an enveloping decidual tissue ($E5.0-7.5) but the contribution from different pools of TRs is unknown. Previous experiments showed that trophospheres are composed of differentiated TRs and have a diminished potential to induce decidualization (Gardner and Johnson, 1972;Rossant and Tamura-Lis, 1981;Rivron et al., 2018). Here, we examined whether the polar TE might be critical for decidua formation. To minimize signaling between conceptus and uterus, we first used fixed blastoids. They were incapable of inducing decidua formation ( Figures 6A and 6B, p < 0.0001). Along with the diminished potential of trophospheres to decidualize the uterus (Rivron et al., 2018), this suggests that blastoids actively instruct decidualization. Blastoids formed with TESCs or with TSCs including a CDX2 inducible transgene (CDX2i-TSCs, Figure S6A) had an enhanced capacity for decidualization as compared to TSCs blastoids ( Figures 6C and 6D, 18.7% versus 7.6%, p = 0.0002; Figure 6F, p = 0.0128). They formed larger deciduae similar in size to that of the blastocyst ( Figure 6E) and achieved a higher receptivity rate ( Figure S6B, 96.7% versus 64.9%). We concluded that inducers contribute in making the TE competent for decidualization. Blastoids formed from ESCs also had a higher potential to regulate decidualization as compared to blastoids formed from EPSCs ( Figure 6E). Finally, a GW501516 treatment of blastoids reduced CDX2 expression and diminished their potential for decidualization ( Figures 6G, 6H, and S6C). Overall, we concluded that EPI inducers regulate CDX2 expression, which endows TRs with the capacity to decidualize the uterine tissues, consistent with CDX2 genetic loss-of-function experiments in which null blastocysts fail to implant (Strumpf et al., 2005). (K) Immunostaining against GFP and ELF5 in E7.5 chimeric embryos obtained upon the injection of blastocysts with serum-cultured TSCs, TX-cultured TSCs, and TXV-cultured TSCs. Non-injected Ctrl represents E7.5 conceptus, which developed from blastocysts that were not injected with TSCs but that, similar to injected blastocysts, underwent laser incision of the Zona Pellucida. For each graph, data are means ± SEM, analyzed by Student's t test: **p < 0.01, ***p < 0.001. Scale bars: 40 mm in (D); 150 mm in (F) and (G); 300 mm in (K). See also Figure S4. Article WNT6/7B are downstream effectors of CDX2 contributing to decidualization We sought to identify molecules (1) whose secretion is regulated by CDX2 and (2) that could contribute to decidualization. Using the computational framework SCENIC (Aibar et al., 2017), we identified multiple Wnt ligands (Wnt6 and Wnt7b) and receptors (Fzd2/7/10) with promoter regions predicted to be bound by CDX2 (two interaction sites for Wnt6, one for Wnt7b promoter region; Figure 7A). In blastocysts, Wnt7b transcripts are the most abundant, followed by Wnt6. In addition, Wnt7b transcripts are enriched in CDX2 high relative to CDX2 low TSCs and upon Cdx2 overexpression ( Figure S7A). Consistent with a role for Wnt in decidualization (Mohamed et al., 2005), WNT7B is highly expressed in the TE and its derivatives ( Figure 7B), and Wnt6/7b expression is maintained in the ExE ( Figure S7B). This expression pattern is conserved in humans with Wnt6/7b transcripts being also abundant in the TE ( Figure S7C) (). Consistent with a role for TEsecreted Wnt ligands in decidualization, 8-cell embryos cultured with porcupine inhibitor (IWP2, 2.5 mM, 48 h) formed blastocysts that had a significantly decreased potential for decidualization (I) Immunostaining for CDX2 and KRT18 as marker proteins for axis formation in the blastoid. Shown is a representative 3D projection of a TESCs blastoid formed without addition of TGFb1. Scale bar represents 50 mm.
(J) Percentage of blastoids with an axis for CDX2 or KRT18. Plotted are the mean percentages of five individual experiments (total 40-60 blastoids per group).
(K) Immunostaining of blastocysts cultured with TXV factors for 6 or 20 h. Scale bar, 40 mm.
For each graph, data are means ± SEM, analyzed by Student's t test and one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S5. . Consistent with a role for WNT activity in TE formation (Rivron et al., 2018), these trophospheres had a decreased cavitation capacity ( Figure S7G). This suggests an autocrine function for WNT/7B in TE morphogenesis. Of note, Wnt controls Hippo pathway activity in several other systems. Finally, we observed that blastoids formed with Wnt6 or Wnt7b KO TESCs initially attached to the uterus comparably to wild type (E5.5, Figures S7H and S7I). Thus we did not detect an early attachment defect, often associated with mural TE functions. However, later (E7.5), the diameter of the deciduae was significantly decreased (Figures 7E and S7J). Note that a 20% decrease in sphere diameter correspond to a 2-fold decrease in volume. Taken together, we concluded that the EPI inducers not only locally maintain the TR progenitors, but also support the secretion of Wnt ligands for decidua formation.

DISCUSSION
For development to occur, embryos must maintain progenitors that fuel growth while allowing a subset of cells to differentiate in order to fulfill functions. Here, we show that a specific combination of EPI inductions increases the optimality of the TF network (CDX2, EOMES, and ESRRB), enhances self-renewal, and prevents differentiation. On the contrary, suboptimal exposure to inducers favors a fluctuation of the progenitor state, creating reversible subpopulations with facilitated differentiation. We propose that the dynamic regulation of this EPI/TR interface endows the progenitor pool with a flexible strategy for either maintaining more progenitors or generating differentiated cell types. We surmise that cellular heterogeneity arises as the embryo exploits a suboptimal environment to control an equilibrium between interconvertible TR populations. This EPI/ TR interface and the reversibility of cellular states would endow the progenitor pool with adaptive and regulative properties to synchronize tissue development, thus acting as a checkpoint. (B) Immunostaining of an E4.5 blastocyst and an E6.5 conceptus for WNT7B and E-CAD. WNT7B was detected in the TE rather than the EPI at E4.5. In the E6.5 conceptus, E-CAD + cells surrounded WNT7B + cells, reminiscent of the CDX2 and KRT18 expression pattern ( Figure 1E). The data indicate that WNT7B is enriched in polar TE derivatives. (C and D) E7.5 deciduae induced by IWP2-treated blastocysts (C) and quantification of the implantation efficiency (D). Arrows indicate resolving deciduae. Each dot represents an individual mouse.
Conceivably, the rapid geometric changes (size and shape) of the developing embryo influence the exposure to EPI inducers, thus linking morphogenesis with the maintenance of TR progenitors. The origin of the observed heterogeneity and plasticity of TSCs is unknown. Upstream CDX2 regulators (e.g., Hippo/ Notch signaling components) can vary stochastically (Sigal et al., 2006) due to the properties of intracellular biochemical loops or to fluctuations in cell-cell membrane interactions. CDX2 expression might also be regulated by geometrical or mechanical cues as observed during the morula-toblastocyst transition. Beyond inductions, both the intrinsic properties of biochemical networks and extrinsic geometric cues might thus contribute to distributing cell states. Signaling pathways act as CDX2 regulators but also as inducers of histone remodeling. For example, Notch activates target genes with trimethylation of H3K4 by inhibiting dimethyl-transferase (KDM5A) activity (Rayon et al., 2014;(Liefke et al., 2010)). Hippo signaling also controls histone remodeling (Hillmer and Link, 2019). Our data showed higher trimethylation of H3K4 (H3K4me3) in CDX2 high cells. Thus, in the CDX2 low TRs that are prone to differentiation, epigenetic mechanisms might act as feedforward loops engraving the differentiation path.
Blastoids formed from TESCs spontaneously generate a gene expression pattern along the axis. We conclude from these experiments of reconstruction that, following subtle early patterning events (Graham and Zernicka-Goetz, 2016;Zhang and Hiiragi, 2018), the EPI produces inductive signals that significantly contribute to axis formation. This process ultimately ensures the blastocyst/uterus interaction of implantation. We propose that inducers, including LPA, FGF4, Nodal, BMP4, BMP7, and IL6/11, contribute toward regulating CDX2 expression in the polar TE, which impacts proliferation, self-renewal, and epithelial morphogenesis but also the expression of WNT ligands that contribute to decidualization. The regulation of WNT ligands might be a mechanism with potential translational applicability to human implantation (Koler et al., 2009).
Altogether, this study provides a framework to explain how the conceptus leverages inductions and TR state fluctuation to maintain progenitors, facilitate differentiation, or allocate and balance the functions necessary for implantation to occur.

Limitation of study
Although blastoids formed with ESCs and TESCs implant better into the uterus and form decidua more efficiently than blastoids formed with ESCs and TSCs, we did not observe the formation of a fetus. The minimal requirements for development to occur are not yet met.
We selected WNT6 and WNT7b as downstream effectors of CDX2 based on RNA-seq, ChIP, TESC and CDX2 overexpression analysis in TSCs, and the expression pattern of these genes in the TE of human blastocysts. We propose that these WNT ligands follow the dynamical expression pattern of CDX2. However, decidualization is a complex process that involves multiple players. Beyond WNT6/7b, other molecules secreted by the conceptus might affect decidualization.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

INCLUSION AND DIVERSITY
We worked to ensure sex balance in the selection of non-human subjects. We worked to ensure diversity in experimental samples through the selection of the cell lines. We worked to ensure diversity in experimental samples through the selection of the genomic datasets. One or more of the authors of this paper self-identifies as a member of the LGBTQ+ community. The author list of this paper includes contributors from the location where the research was conducted who participated in the data collection, design, analysis, and/or interpretation of the work.

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Nicolas C. Rivron, Ph.D. (nicolas.rivron@imba.oeaw.ac.at).

Material availability
This study did not generate new unique reagents.
Data and code availability d The raw data of the scRNA-seq, bulk RNA-seq, and ATAC-seq reported in this paper have been deposited to the GeneExpression Omnibus (GEO) database and are publicly available as of the date of publication. Accession numbers are also listed in the key resources table. Microscopy data reported in this paper will be shared by the lead contact upon request. d All original code has been deposited at Zenodo (https://doi.org/10.5281/zenodo.6602725) and is publicly available as of the date of publication. Link is listed in the key resources d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Mouse lines and embryos
All animal experiments (e.g., blastocyst flushing and uterus transfer) were conducted using 8-20-week-old female mice on a B6CBAF1 background, unless noted otherwise (e.g., C57BL/6J, FVB/N, and 129/Sv background for the TSCs derivation experiment). Mice were maintained at the IMP/IMBA animal house. All animal experiments were approved by the IMP/IMBA animal house and performed in accordance with the guidelines of the institution.
Six days after checking the plug (for E6.5 embryos), the uterine wall was gently removed with forceps. The deciduae were fixed with 4% formaldehyde overnight at 4 C and washed with PBS (three to four x for 1 h each). After EtOH-xylene-paraffin dehydration processing, the deciduae were embedded in paraffin and sliced to 2-mm thickness.

METHOD DETAILS
Generation of inducible Caudal Type Homeobox 2 overexpression cell lines Inducible CDX2 TSCs were generated in the F4 TSCs line. pCAG-PBase (5 mg) and PB-TAC-Cdx2-ERP (5 mg) were transfected by NEPA21 electroporation (Nepa Gene Co. Ltd) into 1x10 6 cells in single-cell suspension. One day after transfection puromycin (1 mg/ml) was added for 7 days and the selected cells were maintained in a lower concentration of puromycin (0.1 mg/ml).
Generation of WNT6 and WNT7B knockout trophoblast stem cells WNT6 and WNT7B KO TSCs were generated in the F4 TSCs line using a CRISPR-Cas9 system. Cells were electroporated (NEPA21 Super Electroporator, Nepa Gene Co., Ltd.) to transfect plasmids containing a specific guide RNA sequence ( Figure S7D), as well as a Cas9, and BFP sequence. 10 mg of plasmid were used to transfect 1x10 6 cells. BFP positive cells were FACS sorted the next day. After their expansion, BFP negative cells were FACS sorted individually onto inactivated MEFs for clonal expansion and cultured in TXV medium. MEFs were depleted and cells were genotyped by Sanger sequencing. The effect of genomic editing on gene expression was confirmed by qRT-PCR using primers that amplify the 3 0 UTR and Western Blot ( Figure S7E and S7F). Sequence information for all primers and sgRNAs used in this study are available in Methods S3.

Cell cycle analysis
After trypsinization, 1x10 5 TSCs and TESCs were incubated in 0.5 mL of TX medium with 10 ug/ml Hoechst 34580 for 30 min at 37 C. After the incubation time, tubes with cells were placed on ice and analyzed on a FACS Canto II.

Colony formation assay
Single cells were sorted into MEF coated plates with either TX or TXV medium. Medium was changed every 48 h and the number of wells containing colonies was assessed 7 days after sorting.

Combinatorial screen
A library of all compounds tested in the combinatorial screen can be found in Table S4. To identify positive modulators for CDX2 induction, CDX2-eGFP TSCs were cultured in TX medium for 24 h and then exposed to new TX media containing the different concentrations of individual compounds. After 48 h, CDX2 expressions of TSCs in each condition were analyzed by flow cytometry (FACS Fortessa, and FlowJo). For the experiment in Figure S3B, IGF2 (50 ng/mL) and ZSTK474 (PI3K inhibitor, 200 nM) were added to the medium.

Immunofluorescence
Samples were fixed using 4% formaldehyde in PBS for 20-30 min at room temperature (RT) followed by three washing steps with PBS. A 0.3% triton solution in PBS (PBS-T) was used for permeabilization for 30-60 min at RT, followed by a 1-2 h blocking step with 0.1% PBS-T + 2% BSA +3% serum (goat or donkey serum complementary to the host of the secondary antibody). Samples were then incubated overnight with the primary antibody diluted in blocking solution at 4 C. A detailed summary of all primary antibodies used in this study is provided in Methods S4. The next day samples were washed three times in 0.1% PBS-T and incubated with the corresponding secondary antibodies for 1 h at RT. Hoechst was used for counterstaining with or without WGA. Images were taken with one of the following microscopes: PerkinElmer Ultraview VoX spinning disk microscope, confocal Axio Observer inverted microscope equipped with a Yokogawa CSU X1 Spinning disk, and Olympus IX3 Series (IX83) inverted microscope equipped with a dual-camera Yokogawa W1 spinning disk. The images were analyzed with Fiji and photoshop.

Single-molecule FISH
TSCs plated on glass coverslips were allowed to grow and subsequently fixed using RNase free 4% PFA in PBS +1% Acetic Acid for 20 min. After fixation, all samples were processed as described in the Quantigene ViewRNA kit instructions (Affymetrix, QVC0001). Briefly, after three washes with RNase free PBS, samples were incubated for 10 min in a detergent solution. This was followed by three washes with RNase free PBS after which samples were incubated for 5 min at RT with Q protease. Samples were again washed three times with RNase free PBS, and incubated with the probes of interest diluted in Probe set diluent at 40 C for 3 h (in a humidified chamber). After three washes with wash buffer, samples were incubated at 40 C for 30 min with a preamplifier diluted in amplifier diluent and washed again for three times. Samples were then incubated at 40 C for 30 min with amplifier diluted in amplifier diluent. Samples were washed again three times with wash buffer and incubated at 40 C for 30 min with label diluted in label probe diluent. After two washes, they were washed once more for 10 min. Samples were then incubated for 15 min in RNase free PBS with Hoechst and WGA as counterstains followed by three washes with RNase free PBS. Blastocysts or blastoids were carefully placed in mounting media in glass bottom 3.5 mm plates. All samples were imaged with a 63x oil immersion objective on a PerkinElmer Ultraview VoX spinning disk microscope.

Single-molecule FISH polarity quantification
Single molecule fluorescence in situ hybridization (smFISH) confocal images were taken with a z-step of 0.3 mm. Given the complexity of an analysis performed in 3D that would require an algorithm capable of segmenting cells and quantifying the number of transcripts in 3D, we decided to quantify a 2D projection of the slices that included the ICM and blastocoel. Those z stack projections were oriented with the polar side on the left and the mural side on the right and then analyzed for average column pixel intensity, allowing us to plot an average pixel intensity histogram. An intensity profile was plotted for each embryo and gene. Each blastocyst is structurally different showing distinct cavity sizes, which implies that a different percentage of the TE is in contact with the ICM for each embryo.
In order to compare the expression of polar and mural TE, we divided the length of the embryo in three segments of equal distance, irrespective of the total diameter. The intermediate segment was considered a transition stage between the polar and mural regions and therefore was not included in the next analyses. The polar and mural segments of the profile were analyzed by comparing the average pixel intensities of each pixel column included in the segment.
High content imaging Each colony was imaged for E-CADHERIN, CDX2 and Nuclei stainings. E-CADHERIN staining was used for manual cell segmentation in ImageJ. Cell profiler was used for analysis of cell segmentation and the other stainings. Measurements obtained in the Cell profiler were used for further analysis using a Python pipeline. After discarding dividing cells based on the nuclear staining, a total of 502 control cells and 297 TXV TSCs cells were analyzed.

Live cell imaging
For the live cell imaging, CDX2-eGFP TSCs were seeded in glass bottom 12 well plates coated with Matrigel in TX medium at 25,000/ cm2 density and they were incubated for 24 h at 37 C, 5% CO2. At the end of the 24h the medium was replaced with fresh medium and then the well plate was transferred to UltraVIEW spinning disk confocal microscope (PerkinElmer) for live imaging analysis. The stage area was set up to 37 C, 5% CO2 prior to the experiment. Images were collected every 12 min (5 timepoints per hour) for a total of 60 h in the GFP channels. The analysis of the live imaging data was completed with ImageJ.

Flow cytometry
Dissociated TSCs/TESCs with 0.05% trypsin were stained for 30 min on ice with the 100 ng of LY6A antibody, and were incubated on ice with anti-rat Alexa 647 (See Methods S4). Each antibody incubation was followed by a wash with FACS buffer (PBS plus 2% FBS). After resuspension of cells with FACS buffer, flow cytometry was conducted using FACS LSR Fortessa (BD). We used at least 10,000 cells for gating. Data was analyzed by FlowJo software. With CDX2-eGFP TSCs/TESCs, we directly used the cells for the flow cytometry after dissociation.
Blastoids made by H2B-RFP ESCs and CDX2-GFP TSCs (see 'Blastoids formation' section below) were dissociated with 0.05% trypsin. By plotting them with GFP and RFP, GFP + TR cells were sorted out for further analysis.
For CDX2-high, -low cell sorting, we dissociated CDX2-GFP TSCs and sorted the cells with FACS aria III based on naive GFP signal. For CDX2-high and -low groups (H and L), we use top and bottom 10% of GFP + cells, respectively.
qRT-PCR RNA was harvested using either the RNeasy Mini Kit (Quiagen, 74,104) or the innuPREP RNA Mini Kit 2.0 (Analytik Jena, 845-KS-2040050) according to the manufacturer's instructions. For cDNA synthesis RNA was incubated with 2.5 mM OligodT primer (New England Biolabs, S1316S) and 0.5 mM dNTPs (in house, Molecular Biology Service IMBA Vienna) at 65 C for 5 min. Reverse transcription was then performed using the SuperScriptÔ III Reverse Transcriptase (Invitrogen, 18,080,044) together with RNaseOUTÔ Recombinant RNase Inhibitor (Invitrogen, 10,777,019) according to the protocol provided by the manufacturer. For the qPCR reactions GoTaq qPCR Master Mix (Promega, A6001) was used. All qPCRs were performed using a Bio-Rad CFX Connect Real-Time PCR System. Relative expression levels of target genes were calculated with the DDCT method using Hprt as an endogenous reference gene for internal normalization. Sequence information for all primers can be found in Methods S3.

RNA sequencing
For bulk sequencing 1000 control and TXV cultured cells were used for Trizol RNA extraction. Both bulk and single cell sequencing was performed following the Cel Seq 2 protocol (Hashimshony et al., 2016). In-bulk samples were first normalized and then analyzed using the DESeq2 package in Rstudio. Triplicates for each group (F4 GFP in TX TXV, TX differentiated, and TXV differentiated) were analyzed. Genes were considered differentially expressed when showing a 1.5-fold expression change with a p value < 0.05. The DAVID gene ontology online tool was used for gene enrichment analysis.
Mapping and processing of single-cell mRNA sequencing data Read one contains the cell or section barcode and the unique molecular identifier (UMI). Read two contains the biological information. Reads 2 with a valid cell barcode were selected and mapped using STAR-2.5.3a with default parameters to the mouse mm10 genome, and only reads mapping to gene bodies (exons or introns) were used for downstream analysis. Reads mapping simultaneously to an exon and to an intron were assigned to the exon. For each cell or section, the number of transcripts was obtained as previously described . We refer to transcripts as unique molecules based on UMI correction.
Analysis of single-cell mRNA sequencing data To analyze CEL-Seq single cell sequencing experiments, mouse genomic sequence and annotation from NCBI GRCm38.p6 were used. Reads were trimmed using trim_galore v0.6.4 and subsequently aligned to the mouse genome (GRCm38.p6) using STAR v2.7.6a. UMI quantification and raw count matrixes generation were performed using umi_tools v1.0.1.
To culture EPSCs and to make blastoids formed from the association of EPSC and TESCs, we followed the variation on our initial protocol as described in the reference (Sozen et al., 2019), at the exception of the initial TSCs culture conditions that were replaced by TESCs.

Uterus transfer and decidua analysis
Four hours before uterus transfer, the blastoid medium was replaced with DMEM high glucose medium. Picked blastoids from the microwells were placed on four ring-well plate and briefly washed with DMEM high glucose medium. With the few medium, 10-12 blastoids were transferred into the only one of the uterine horns of E3.5 pseudopregnant females, unless noted otherwise. E7.5 deciduae were explanted 4 days after uterus transfer. The bulb which has a clearly bigger diameter than the width of a normal uterus was considered as a decidua, and the number of deciduae was confirmed by three different scientists by performing blind test. We took decidua pictures with the ruler, and measured the length of deciduae from mesometrium side to anti-mesometrium side (perpendicular to the direction of the cervix from the ovary) with Fiji.
For the fixed blastoids transfer, blastoids were fixed with 4% formaldehyde in PBS for 30 min at RT. The blastoids in the control group were in the PBS for 30 min to be fair. After enough washing with PBS, the blastoids were transferred. For the Cdx2i blastoids, 225 nM (100 ng/mL) of doxycycline was added when the TSCs were seeded on the microwells with a blastoid medium. For the blastoids with GW501516, 3 mM of GW501516 was added 1 day before the uterus transfer (around 40-45 h after TSCs seeding).
For the comparison experiment of blastocysts with TSCs and TESCs blastoids, freshly isolated E3.5 blastocysts from pregnant females were directly transferred into the E3.5 pseudopregnant recipients. For IWP2 treatment before uterus transfer, we cultured eight cell-embryos in KSOM either with or without IWP2. After 2 days, they developed late blastocysts (E4.0-E4.5) and were transferred to the uterus.
Blue band assay Two days after uterus transfer, 0.4% trypan blue (Thermofisher Scientific, T10282, 10 mL per 1g of mouse weight) was injected through intravenous (i.v.) injection. After 30 min, the mice were sacrificed and E5.5 uteri were analyzed. For Figure 7C, only four to six blastocysts were transferred to prevent overlap of individual blue bands.

Western Blot
For protein isolation, TSCs were lysed using RIPA buffer (Thermo Fisher Scientific,89,900). For western blotting, proteins were transferred to 0.2mm Nitrocellulose membranes (Bio-Rad, 1,620,112), and membranes were then blocked with 0.1% PBS-T (PBS plus 0.1% Triton X-100) containing 5% skim milk at RT for 1 h. To detect specific proteins of interest, WNT7B and a-TUBULIN antibodies were used (See Methods S4). Horseradish peroxidase (HRP)-conjugated secondary antibodies were then used. Each antibody incubation was followed by washing with 0.1% PBS-T. Luminescence was detected with a Bio-Rad ChemiDoc MP Imaging System.

SCOPE and PTUI
Wnt family expression levels in the mouse embryo were analyzed with SCOPE (https://scope.aertslab.org/) using data resources from (Posfai et al., 2021) (GSE145609) for the analysis of transcriptomic differences between the TE and ExE. Wnt family expression levels in the human embryo were analyzed with PTUI (https://bird2cluster.univ-nantes.fr/demo/PseudoTimeUI/human/PTUI.html) by using data resources from ().

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software) and Excel (Microsoft). All error bars represent the SEM(SEM). Data were analyzed using a one-way ANOVA, a two-tailed t-test (for a difference in means), Mann-Whitney analysis, or Wilcox likelihood-ratio test. The statistical analysis used for each dataset is indicated in the figure legend. A p value < 0.05 was considered statistically significant at the 95% confidence level. The number of biological (non-technical) replicates for each experiment is indicated in the figure legends. All representative images shown are from experiments that have been performed in triplicate at least, except Figure S4H (two independent experiments with pooled 36-40 embryos).