We characterized the expression pattern of Rreb1 during early mouse development using wholemount preparations of embryos harboring a LacZ-tagged transcriptional reporter (Figure 1—figure supplement 1A, European Conditional Mouse Mutagenesis Program) (Bradley et al., 2012). At pre-implantation stages (embryonic day (E) 4.5), Rreb1^LacZ was expressed within the inner cell mass (ICM), comprising epiblast cells that will generate the fetus and primitive endoderm (PrE) cells that will give rise to the endoderm of the yolk sac, and the trophectoderm that will form the placenta (Figure 1—figure supplement 1B). In the early post-implantation embryo (E5.5), before gastrulation, Rreb1^LacZ was expressed within the PrE-derived visceral endoderm (VE) and trophectoderm-derived extraembryonic ectoderm (ExE), but not the epiblast (Figure 1A). Subsequently, during gastrulation (E6.5–7.5), expression was observed within the VE, primitive streak, a region where cells undergo an EMT and start to specify and pattern the mesoderm and endoderm germ layers, embryonic and extraembryonic mesoderm (derived from the primitive streak), and distal anterior epiblast (Figure 1A, Figure 1—figure supplement 1C). Around midgestation (E8.0–10.5), Rreb1^LacZ was expressed within the yolk sac endoderm, node, notochord, primitive streak, blood, allantois, head mesenchyme, and pharyngeal arches (Figure 1A, Figure 1—figure supplement 1D–G). We noted that at E10.5 Rreb1^LacZ was expressed in regions of high FGF signaling activity (Morgani et al., 2018b), including the limb buds, frontonasal processes, and isthmus (Figure 1—figure supplement 1F). The domain of Rreb1^LacZ expression within the tailbud varied between embryos, suggesting that Rreb1 transcription may be regulated by the segmentation clock (Figure 1—figure supplement 1F). Data were further validated by comparison to available single-cell transcriptomic (scRNA-seq) datasets of equivalent embryonic stages (Figure 1—figure supplement 1H; Nowotschin et al., 2019; Pijuan-Sala et al., 2019).

In vitro, the Rreb1^LacZ reporter marked a subpopulation of pluripotent embryonic stem cells (ESCs) and epiblast stem cells under self-renewing conditions and became more widely expressed as cells were differentiated by removal of the cytokine LIF or addition of FGF (Figure 1B). Thus, Rreb1 is initially expressed by all lineages of the pre-implantation blastocyst and is downregulated within the epiblast as it transitions from a naïve to a primed state of pluripotency. During post-implantation development, Rreb1 continues to be expressed in extraembryonic tissues and is re-expressed in the embryonic lineages as primed pluripotency is exited and the germ layers are specified.

Previously, we generated chimeric embryos by injecting Rreb1^-/- ESCs into wild-type host embryos. While Rreb1^-/- cells could undergo the gastrulation EMT, migrate within the wings of mesoderm, and differentiate into germ layer derivatives, cells accumulated at the primitive streak over time suggesting that later EMT events are perturbed (Su et al., 2020). To further interrogate the developmental function of Rreb1, we proceeded to generate a Rreb1 knockout mouse using CRISPR-Cas9 technology (Figure 2A). Rreb1^+/- mice were viable and fertile, but heterozygous intercrosses yielded no homozygous mutant offspring. From E7.5 onwards, mutant embryos were smaller than wild-type littermates (Figure 2B,C, Figure 2—figure supplement 1A,B) and, based on morphology and somite number, were approximately 9.5 hr retarded (Figure 2—figure supplement 1B,C). At E9.0–9.5, Rreb1^-/- embryos showed various defects, including microcephaly (2/7 at E9.5, Figure 2—figure supplement 1D), an open foregut (Figure 2—figure supplement 1E), and an open neural tube at the forebrain, midbrain, and posterior neuropore level (8/10 Rreb1^-/- at E9.5, Figure 2D–E, Figure 2—figure supplement 1F). The proportion of Rreb1^-/- embryos with open neural tubes was reduced at E10.5 compared to E9.5 (2/8 [25%] versus 8/10 [80%] Rreb1^-/-, Figures 2B and 3F), suggesting that some mutants close their neural tube upon further development.

Additionally, mutant embryos displayed aberrant notochord formation. In wild-type embryos, the axial mesoderm, marked by BRACHYURY expression in cells anterior to the gut tube, gives rise to the prechordal plate rostrally (Figure 2—figure supplement 1G i) and to the tube-like notochord caudally (Figure 2—figure supplement 1G ii-iv) (Balmer et al., 2016). However, in Rreb1^-/-, BRACHYURY-expressing cells did not establish a tube, instead, intercalating into the foregut (Figure 2—figure supplement 1v), protruding into the foregut lumen (Figure 2—figure supplement 1G vi), or generating multiple distinct clusters (Figure 2—figure supplement 1G vii). Thus, loss of Rreb1 results in a range of phenotypic abnormalities initiating at gastrulation and resulting in midgestation lethality.

Homozygous mutants began to be resorbed at E11.5, as marked by the disintegration of embryonic tissues (Figure 2C), and were not recovered at E12.5 (Figure 2F). Thus, Rreb1 is an essential factor regulating numerous processes during early mouse development.

Rreb1 is a context-dependent transcriptional repressor or activator (Deng et al., 2020). To define the gene expression changes associated with a developmental loss of Rreb1 we performed RNA-sequencing of Rreb1^-/- embryos and compared them to wild-type (Rreb1^+/+) transcriptomes. Embryos were isolated and analyzed at E7.5 (Figure 3A), coinciding with the onset of morphological defects in mutant embryos (Figure 2—figure supplement 1A,B). We identified 65 downregulated and 200 upregulated genes in Rreb1^-/- vs. Rreb1^+/+ embryos (fold-change >log2(2), p<0.05, Figure 3—source data 1). To assess the function of these genes, we implemented Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Downregulated genes were enriched for GO terms associated with blood, including ‘blood microparticle’, ‘fibrinogen complex’, and ‘platelet alpha granule’ (Figure 3—source data 2), and the ‘complement and coagulation cascades’ (Figure 3—source data 3) that play a role in vasculogenesis (Girardi et al., 2006; Moser and Patterson, 2003). Key genes within these groups included the complement inhibitors Cd59a and complement component factor I (Cfi), and the secreted proteins fibrinogen alpha and gamma (Fga, Fgg), complement factor B (Cfb), protein C (Proc), and Alpha fetoprotein (Afp). We also observed a downregulation of Jag2 and Slit1 (Figure 3—source data 1), components of the Notch and Slit-Robo signaling pathways that regulate hematopoiesis and vasculogenesis (Blockus and Chédotal, 2016; Kofler et al., 2011).

The majority of downregulated factors were specifically expressed or enriched within the VE (84% of the 55 differentially expressed genes that were also detected by scRNA-sequencing of gastrulating mouse embryos Nowotschin et al., 2019; Pijuan-Sala et al., 2019; Figure 3B, Figure 3—figure supplement 1A). As these data were generated by whole embryo bulk RNA-sequencing, the downregulation of VE-associated genes could represent a reduction in the expression of specific factors or a relative decrease in the size of the VE. To distinguish between these possibilities, we assessed the expression levels of a number of other critical VE lineage determinants, including the transcription factors Gata6, Gata4, Sox17, and Hnf4a. We found that these genes were not significantly altered in Rreb1^-/- (Figure 3—figure supplement 1B) suggesting that the downregulation, almost solely, of VE markers did not represent a global reduction in VE-associated transcription.

Afp, a plasma glycoprotein secreted by the yolk sac and fetal liver, which regulates angiogenesis (Liang et al., 2004; Takahashi et al., 2004), was one of the VE-associated genes significantly downregulated in mutant embryos. Thus, to validate our RNA-sequencing, we crossed Rreb1^+/- mice to a transgenic reporter whereby the Afp cis-regulatory elements drive GFP expression (Kwon et al., 2006). We then analyzed Rreb1^+/+ and Rreb1^-/-; Afp-GFP^Tg/+ embryos (Figure 3—figure supplement 1C). At E7.5 and 8.5, Afp-GFP is expressed by the embryonic and extraembryonic VE in wild-type embryos (Kwon et al., 2006; Kwon et al., 2008; Figure 3—figure supplement 1D,E). However, while Rreb1^-/- mutant embryos expressed Afp-GFP within the embryonic VE and at the embryonic-extraembryonic boundary, they showed little to no Afp-GFP within the proximal extraembryonic VE (Figure 3—figure supplement 1D,E), consistent with the transcriptional downregulation of this gene.

As downregulated genes were functionally related to blood and vasculogenesis, we then asked whether vascular development was perturbed in the absence of Rreb1 by performing immunofluorescence staining for FLK-1, a marker of hematopoietic and endothelial progenitors, and PECAM-1 (CD31), marking endothelial cell junctions. At E8.0, when a rudimentary circulatory system is first established, we discerned no difference in the level or pattern of FLK-1 or PECAM-1 expression. As previously described (Yamaguchi et al., 1993), and in both Rreb1^+/+ and Rreb1^-/- embryos, FLK-1 +PECAM-1+ cells were present in the extraembryonic mesoderm, including the allantois and yolk sac blood islands, the cephalic mesenchyme, and heart (Figure 3—figure supplement 1F,G). At E9.5–10.5, while wild-type embryos formed a hierarchical branched network of blood vessels within the yolk sac and embryo-proper (Figure 3D i-iii), Rreb1^-/- embryos had dysmorphic yolk sac capillaries that instead resembled a primitive capillary plexus (5/6 Rreb1^-/- at E9.5, 4/6 Rreb1^-/- at E10.5, Figure 3D iv, Figure 3—figure supplement 1H i, I) and blood within the extravascular space (Figure 3D v, Figure 3—figure supplement 1H ii). Cardiovascular defects were also observed within the Rreb1^-/- embryo-proper including little to no blood within the fetus (2/6 Rreb1^-/- at E9.5, 1/6 Rreb1^-/- at E10.5, Figure 3D vi, Figure 3—figure supplement 1I iii), pooling of blood (2/6 Rreb1^-/- at E9.5, 4/6 Rreb1^-/- at E10.5), an enlarged heart (2/6 Rreb1^-/- at E9.5, Figure 3D vi), and pericardial edema (2/8 Rreb1^-/- at E10.5, Figure 3E). At E9.5, a number of Rreb1^-/- embryos displayed extravascular FLK-1 +PECAM-1+ cells throughout the fetus, including accumulating within the heart (2/6 Rreb1^-/- at E9.5, Figure 3D vii, F). These defects were also associated with an abnormal dorsal aorta morphology. In wild-type embryos, the dorsal aorta comprised flattened endothelial cells surrounding a lumen containing blood cells (Figure 3F,G). Elongated endothelial cells also bordered the neural tube (Figure 3F,G). In contrast, in Rreb1^-/- embryos, the dorsal aorta lumen was collapsed and FLK-1 +PECAM-1+ cells, both of the aorta and surrounding the neural tube, appeared rounded compared to those in wild-type embryos (Figure 3G).

While the dorsal midbrain of E9.5 and 10.5 wild-type embryos contained defined networks of veins and branched blood vessels (Figure 3D iii, H, Figure 3—figure supplement 1J), Rreb1^-/- exhibited displayed reduced vascular remodeling with fewer, wider blood vessels (3/6 Rreb1^-/- at E10.5, Figure 3H, Figure 3—figure supplement 1J). ENDOGLIN (CD105) is also a marker of endothelial cells. In E9.5 wild-type embryos, a thin layer of PECAM-1 +ENDOGLIN + endothelial cells formed between the neuroepithelium and outer surface ectoderm layer. However, in Rreb1^-/-, PECAM-1 +ENDOGLIN + cells were arranged into discrete, rounded clusters (Figure 3—figure supplement 1K). Thus, loss of Rreb1 results in the downregulation of VE-expressed vasculogenesis-associated genes and cardiovascular defects that culminate in embryonic lethality at midgestation.

We then analyzed genes that were upregulated in E7.5 Rreb1^-/- vs. Rreb1^+/+ embryos. The upregulated genes, Bmper and Cxcl10 (Figure 3—source data 1), are associated with ‘endothelial cell activation’ and thus their altered expression may contribute to the observed cardiovascular defects. H2-Q2, the purported mouse homolog of HLA-G (Comiskey et al., 2003; Figure 4—figure supplement 1A), was also upregulated consistent with reports that Rreb1 is a transcriptional repressor or HLA-G in humans (Flajollet et al., 2009). Subsequent GO analysis revealed that upregulated genes were enriched for four main categories; ‘cytoskeleton’, ‘membrane and vesicle trafficking’, ‘cell junctions’, and ‘extracellular space’ (Figure 3—source data 2). Factors associated with the cytoskeleton included microtubule components (Tubb3), microtubule-interacting proteins (Map6, Jakmip2, Fsd1), microtubule motors (Kif5a, Kif5c, Kif12), actin-binding proteins (Coro1a), and factors that connect adherens junctions to the cytoskeleton (Ctnna2, Ablim3) (Figure 4A). Genes within the ‘vesicle trafficking’ category were also related to the cytoskeleton. For example, Rab family members (Rab6b, Rab39b) (Figure 4A) regulate vesicle transport along actin and microtubule networks.

We therefore asked whether these transcriptional changes in cytoskeleton genes corresponded to a change in cytoskeletal organization in Rreb1 mutants. In the normal (wild-type, Rreb1^+/+) epiblast epithelium, F-ACTIN was arranged in linear filaments oriented parallel to cell junctions (Figure 4B,C). In contrast, F-ACTIN was punctate at epiblast cell junctions within Rreb1^-/- embryos (Figure 4B,C). The cytoskeleton interacts with and influences the localization of adherens junction components (Chen et al., 2003; Liang et al., 2015; Mary et al., 2002; Mège and Ishiyama, 2017; Sako-Kubota et al., 2014; Stehbens et al., 2006; Teng et al., 2005). As we noted a significant upregulation of Ctnna2 and Ablim3, which encode proteins that connect the cytoskeleton to adherens junctions (Figure 4A), we asked whether the change in F-ACTIN was associated with a rearrangement of cell junctions. Cadherins are critical components of adherens junctions and, during gastrulation, E-CADHERIN is expressed within the epiblast, VE, and extraembryonic ectoderm (Pijuan-Sala et al., 2019). In wild-type embryos, E-CADHERIN, similar to F-ACTIN, formed a continuous belt between epithelial epiblast cells but, in Rreb1 mutants, showed a punctate localization (Figure 4D,E, Figure 4—figure supplement 1B–D). Beta-CATENIN was also more punctate at Rreb1 mutant compared to wild-type epiblast junctions (Figure 4—figure supplement 1E), indicating that adherens junction complexes were altered. The punctate adherens junction organization was observed within the proximal but not distal Rreb1^-/- epiblast (Figure 4—figure supplement 1F). The change in E-CADHERIN and Beta-CATENIN localization was not associated with a transcriptional change in these genes, or in the expression of other adhesion-associated factors, such as tight junction components (Figure 4—figure supplement 1G). Thus, the altered protein localization was due to post-transcriptional mechanisms. Therefore, loss of Rreb1 results in a change in the expression of cytoskeleton-associated factors and a change in the organization of the cytoskeleton and adherens junctions within the epiblast.

The cytoskeleton is the scaffold of the cell that regulates cell-cell adhesion (Elson, 1988; Gavara and Chadwick, 2016; Grady et al., 2016; Ketene et al., 2012) and epithelial organization (Bachir et al., 2017; Ivanov et al., 2010; Sun et al., 2015; Vasileva and Citi, 2018). In cancer, a cytoskeleton-mediated switch from linear to punctate E-CADHERIN results in weaker cell-cell adhesion and loss of epithelial integrity (Aiello et al., 2018; Ayollo et al., 2009; Gloushankova et al., 2017; Jolly et al., 2015; Kovac et al., 2018; Saitoh, 2018). In keeping with this, Rreb1^-/- gastrulating embryos exhibited perturbed epithelial architecture. In wild-type embryos, VE cells formed an ordered monolayer overlying the embryonic epiblast and the ExE (Figure 5A,B, Figure 5—figure supplement 1A), while in Rreb1^-/-, cells protruded from the VE (Figure 5A), the extraembryonic VE was frequently ruffled (Figure 5B, Figure 5—figure supplement 1A), and abnormal masses of cells accumulated at the anterior embryonic-extraembryonic boundary (Figure 5C, Figure 5—figure supplement 1B).

Rreb1^-/- initiated gastrulation in the posterior of the embryo, as marked by downregulation of the pluripotency-associated transcription factor SOX2 and upregulation of the primitive streak marker BRACHYURY (Figure 5D). Furthermore, Rreb1^-/- epiblast cells underwent an EMT at the primitive streak, delaminated from the epithelium, and migrated anteriorly in the wings of mesoderm (Figure 5E). Cells within Rreb1^-/- embryos also differentiated into mesoderm and DE, marked by GATA6 and SOX17 expression respectively (Figure 5D, Figure 5—figure supplement 1C). Hence, Rreb1^-/- mutant cells specify and begin to pattern the embryonic germ layers. As the majority of downregulated genes were factors expressed by the endoderm, we investigated endoderm specification and morphogenesis in more detail. We observed high levels of non-specific antibody staining, including for BRACHYURY and N-CADHERIN in the Rreb1^-/- VE. Such non-specific staining is often observed in the VE of wild-type embryos prior to intercalation of the DE (Kwon et al., 2008; Morgani et al., 2018a), which has ben attributed to its extensive vacuolation. Thus, we hypothesized that there may be defects in DE intercalation in mutant embryos. Consistent with this, the Afp-GFP reporter mouse line revealed delayed dispersal of the embryonic VE, a process driven by DE intercalation. By E7.5, the VE was fully dispersed in wild-type embryos, characterized by mosaic GFP labeling of the outer endoderm layer (Figure 5F,G), but noticeably reduced in Rreb1^-/- embryos of the same stage (Figure 5F,G). The VE was successfully dispersed in mutant embryos by E8.5 (Figure 3—figure supplement 1E). Wild-type SOX17 +DE cells intercalated into the lateral VE as they migrated anteriorly, but Rreb1^-/- SOX17 + cells migrated to the anterior pole without intercalating (Figure 5H, Figure 5—figure supplement 1C), contributing to the abnormal bulges at the anterior embryonic-extraembryonic boundary (Figure 5C, Figure 5—figure supplement 1B). Notably, unlike wild-type DE, Rreb1^-/- SOX17 + cells expressed the mesoderm and primitive streak marker BRACHYURY (Figure 5H), suggesting that they may be incorrectly specified.

As within the VE, the epiblast of mutants showed a range of morphological defects including uncharacteristic folding of the epithelial layer (Figure 5E, I i, Figure 5—figure supplement 1D), abnormal accumulations of cells (Figure 5I ii), increasingly variable cell orientation (Figure 5—figure supplement 1E–G), separation of typically closely apposed tissue layers, such as the mesoderm and endoderm (Figure 5D,E, Figure 5—figure supplement 1H), and cells falling out of the epiblast (Figure 5—figure supplement 1I). In wild-type embryos, epiblast cells divide at the apical, cavity-facing surface while being maintained within the epithelial layer but, in Rreb1^-/- embryos, we observed dividing cells that left the epithelium (Figure 5—figure supplement 1J). Additionally, the epiblast and endoderm are monolayer epithelia in wild-type embryos but formed multilayered regions in Rreb1^-/- mutants (Figure 5I iii).

Epithelial homeostasis requires tight regulation of proliferation and the maintenance of cell polarity. However, Rreb1^-/- embryos showed no difference in the absolute or relative number of dividing cells within the epiblast, VE, or mesoderm when compared to wild-type littermates (Figure 5—figure supplement 1K,L). Furthermore, apicobasal polarity of the Rreb1^-/- epiblast cells was unaffected, demonstrated by the correct positioning of the tight junction protein ZO-1 at the apical surface and the basement membrane protein LAMININ at the basal surface (Figure 5—figure supplement 1M,N). Together these data show that loss of Rreb1 results in disrupted epithelial architecture of both embryonic and extraembryonic tissues, associated with altered cytoskeleton and adherens junction organization.

Chimeras containing Rreb1^-/- cells showed an accumulation of cells at the primitive streak (Su et al., 2020). However, this phenotype was observed in only a small fraction of in Rreb1^-/- embryos (3/52). The difference in phenotypic penetrance and expressivity between these experiments could exist for a variety of reasons, including contribution of the extraembryonic tissues, which are wild-type in chimeras but mutant in the Rreb1^-/- mouse line, variability in the proportion of wild-type versus mutant cells and interactions between wild-type and mutant cells in chimeras, as well as genetic background. While the majority of Rreb1 mutant embryos analyzed in this study were of a CD1 outbred background, reflecting the genetic diversity within the human population, chimeric embryos were generated by introducing 129 ESCs into C57BL/6 embryos, both of inbred backgrounds. Mutant inbred mice tend to display more severe defects than their outbred counterparts and thus phenotypic differences could reflect genetic background. To assess this, we collected and analyzed a litter of E7.5 C57BL/6 Rreb1^-/- embryos. We found that 4/4 C57BL/6 Rreb1^-/- embryos exhibited severe defects in the exit of cells from the posterior epiblast (Figure 5—figure supplement 2A,B) and a reduction in LAMININ basement membrane break down at the primitive streak compared to wild-type Rreb1^+/+ littermates (Figure 5—figure supplement 2C,D). There was also a more pronounced buckling of the epiblast epithelium than in outbred CD1 embryos, with 4/4 embryos displaying abnormal epiblast folding (Figure 5—figure supplement 2B). Thus, penetrance and expressivity of the Rreb1^-/- phenotype is influenced by genetic background.

In the context of cancer, cells with punctate E-CADHERIN localization represent an intermediate epithelial-mesenchymal state (Sha et al., 2019; Yang et al., 2020), characterized by an increased propensity for collective invasion and metastasis (Aiello et al., 2018; Ayollo et al., 2009; Gloushankova et al., 2017; Jolly et al., 2015; Kovac et al., 2018; Saitoh, 2018). This state is linked to the downregulation of the transcription factor Ovol1 (Jia et al., 2015; Saxena et al., 2020), which suppresses a mesenchymal identity, and the tight junction component, Claudin7 (Aiello et al., 2018; Kim et al., 2019; Wang et al., 2018). Both of these factors were also significantly downregulated in Rreb1^-/- embryos (Figure 6—figure supplement 1A). Furthermore, we observed Rreb1^-/- epiblast cells that had acquired mesenchymal characteristics (n = 1/3 Rreb1^-/-). In wild-type embryos, the mesenchymal marker and EMT regulator SNAIL was expressed within the primitive streak and the wings of mesoderm (Figure 6A). However, in Rreb1^-/-, SNAIL was ectopically expressed within epiblast cells precociously exiting the epithelium (Figure 6A). Moreover, these cells exhibited punctate Beta-CATENIN, in contrast to the linear localization in neighboring SNAIL-negative epiblast (Figure 6—figure supplement 1B). Moreover, in 3/52 Rreb1^-/- embryos, we observed chains of cells that traversed tissue layers, including cells expressing the epiblast marker SOX2 that crossed the VE (Figure 6B, Figure 6—figure supplement 1) and cells expressing the mesoderm and endoderm marker GATA6 that spanned the epiblast (Figure 6—figure supplement 1C). In most cases, these aberrant cells crossed the VE and SOX2-positive (SOX2+) pyknotic nuclei were detected on the adjacent exterior surface of the embryo (Figure 6—figure supplement 1C,D).

These events were rare in the Rreb1^-/- mouse line, precluding a detailed analysis of the identity of aberrant cell populations. However, we frequently observed ectopic SOX2 + cells dispersed throughout chimeric embryos, where the embryonic epiblast-derived tissues are a mosaic of wild-type and mutant origin and extraembryonic tissues are wild-type (30/63, 48% of Rreb1^-/- chimeric embryos, Figure 6C–E, Figure 6—figure supplement 1E and 33–90 ectopic SOX2 + cells/per embryo). Abnormal SOX2 + cells were predominantly positioned between the epiblast and endoderm tissue layers and less commonly found within the epiblast, cavity, and wings of mesoderm (Figure 6—figure supplement 1F). These cells divided and persisted until later stages of development (Figure 6—figure supplement 1G) and were also observed at the onset of gastrulation (Figure 6—figure supplement 1H).

In chimeric embryos, ectopic cells expressed higher levels of SOX2 than those within the epiblast (Figure 6—figure supplement 1I), suggesting that their identity may be altered during exit from the epithelium. To investigate this, we performed immunofluorescence analysis of a panel of key markers expressed within embryos at this time. In addition to SOX2, ectopic cells expressed the pluripotency-associated factors NANOG and OCT4 (POU5F1) (Figure 6—figure supplement 2A). However, they did not express the epiblast marker OTX2 (Figure 6—figure supplement 2B). OTX2 is additionally expressed within the mesoderm and VE, hence its absence excluded the possibility that cells transdifferentiated toward these lineages. OTX2 blocks pluripotent cells from adopting a primordial germ cell (PGC) identity (Zhang et al., 2018). Therefore, we asked whether the absence of OTX2 correlated with an upregulation of PGC-associated genes. PGCs express a myriad of pluripotency factors, for example SOX2, NANOG, and OCT4, but not the naïve pluripotency marker KLF4. Ectopic SOX2 + cells did not express KLF4 (Figure 6—figure supplement 2C) but did express the PGC marker AP2γ (Figure 6—figure supplement 2D). In summary, Rreb1^-/- cells that ectopically exited the epiblast in chimeric embryos were SOX2^HINANOG + OCT4+OTX2- KLF4- AP2γ+, a profile found only in PGCs at this developmental stage. Thus, loss of Rreb1 caused cells to ectopically exit the epiblast epithelium in early mouse embryos, correlating with a change in cell fate.

In chimeric embryos, ectopic SOX2 + cells were of both wild-type and mutant origin (Figure 7A, Figure 7—figure supplement 1A), indicating that invasive-like behaviors were not driven solely by cell-autonomous properties, such as changes in the cytoskeleton and adherens junctions. Remodeling of the extracellular matrix (ECM) could promote invasive behaviors of both wild-type and mutant cells. We noted that many of the genes that were significantly altered in Rreb1^-/- embryos were associated with ECM and cell-ECM adhesion. For example, Tff3 (Ahmed et al., 2012; Pandey et al., 2014), Hpse (Liu et al., 2019), Slit1 (Gara et al., 2015), Spon1 (Chang et al., 2015), Spock1, and Spock3 (Chen et al., 2016) are associated with increased cancer cell invasion and were upregulated in Rreb1^-/-, and Selenbp1 (Caswell et al., 2018; Schott et al., 2018) and Serpin6b (Chou et al., 2012) are tumor suppressor genes that were downregulated. Therefore, we asked whether the basement membrane underlying the epiblast was perturbed in Rreb1^-/- chimeras.

In wild-type chimeras, the basement membrane at the epiblast-VE interface is broken down in the posterior of the embryo at the primitive streak during gastrulation, as cells undergo an EMT (Figure 7C). In Rreb1^-/- embryo chimeras, the basement membrane was broken down at the primitive streak but also in anterior and lateral regions of the epiblast (Figure 7C, Figure 7—figure supplement 1B). SOX2 + cells were observed traversing these ectopic basement membrane breaks (Figure 7C). Furthermore, aberrant SOX2 + cells were surrounded by higher levels of Laminin than their neighbors and associated with Laminin tracks, up to 68 µm (approximately 7 cell diameters) in length (Figure 7D, Figure 7—figure supplement 1C). Thus, loss of Rreb1 in the mouse embryo caused epiblast epithelial cells to cross the basement membrane underlying the epiblast epithelium, reminiscent of the invasive cell behaviors observed in cancer metastasis. These defects were associated with cell-autonomous changes in the cytoskeleton as well as non-cell-autonomous changes in the ECM. KEGG pathway analysis also revealed that the genes upregulated in Rreb1^-/- embryos were enriched for pathways associated with cancer, including ‘Pathways in cancer’, ‘MicroRNAs in cancer’, and ‘Gastric cancer’ (3/5 most enriched pathways, Figure 7E). Together these data suggest that the embryonic role of Rreb1 may be functionally linked to its role in cancer (Figure 7F).

Mice were housed under a 12 hr light-dark cycle in a specific pathogen-free room in the designated facilities of MSKCC. Natural matings were set up in the evening and mice were checked for copulation plugs the following morning. The date of vaginal plug was considered as E0.5. Genotyping was carried out at the time of weaning. Mice were outbred to CD1 animals and maintained on a mixed bred CD-1/129 Sv/C57BL6/C2J background in accordance with the guidelines of the Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Animal Care and Use Committee (IACUC).

To generate the Rreb1^LacZ reporter mouse line, in vitro fertilization was performed using C57BL/6N-A^tm1Brd Rreb1^{tm1a(EUCOMM)Wtsi}/WtsiPh (RRID:IMSR_EM:10996) sperm obtained from the European Conditional Mouse Mutagenesis Program (EUCOMM). The Tm1a (knockout-first) allele was genotyped by PCR using the following primers: Rreb1 5’ arm: CTTCTGTCCCAGAAGCTACATTGC, Rreb1 3’ arm: GGACAACGGTCACTGAGAAGATGG, Lar3: CAACGGGTTCTTCTGTTAGTCC and the protocol: Step1–95°C for 3 min, Step 2–35x: 95°C for 30 s, 63°C for 30 s, 72°C for 30 s, Step 3–72°C for 3 min. This results in a wild-type allele amplicon band of 751 bp and a transgenic allele amplicon of 502 bp. Tm1a mice were then crossed with a Flp recombinase mouse line (Rodríguez et al., 2000) to remove the neomycin cassette and Exon 6, producing the Tm1b LacZ-tagged null allele. Rreb1^LacZ/+ embryos were analyzed by X-gal staining to determine the Rreb1 expression pattern.

Rreb1^-/- mutant mice were generated by CRISPR-mediated genetic knockout. The CRISPR gRNAs used for deleting exon 6 of the Rreb1 gene were designed using the approach of Romanienko et al., 2016. The sequences of the guides are: crRNA#1: TATTATGAACTCCTCTGGAC, crRNA#2: AGTGTCTTCGAAAGAGCCAA, crRNA#3: CGTTACAACAAAGCACCCTT, crRNA#4: AGGAAAACTCGTAGTGGCAC. To initiate cleavage and subsequent deletion of the target locus in mice, guides were injected in pairs, either #1 and #3 or #2 and #4, into the pronuclei of mouse zygotes at a concentration of 50 ng/µl each, with 100 ng/µl purified Cas9 protein (PNABio, Newbury Park, CA), using conventional techniques (Behringer et al., 2014). Founder mice were analyzed for the deletion by PCR using the primers RREB2: GACACCTAGTCACCGAGGAAAC and RREB6: CTGTGGCAGATCTGGTAGGC. This primer pair is located outside of the gRNA cleavage sites, thereby revealing the size of the deletion based on the nucleotide length of the amplicon obtained. The wild-type amplicon size is 1019 bp. The deletion amplicons, if there had been a simple cut and rejoining, would be: Cr#1 and #3: 275 bp. Cr#2 and #4: 456 bp. Genotyping of the Rreb1 locus was performed by PCR with primers RREB1_1: GTGACAGAGGGAACAGTGGG, RREB1_2: GACACCTAGTCACCGAGGAAAC, RREB1_3: GTGTCTGTGTTGTGCTGCA using the following protocol: Step1–94°C for 3 min, Step 2–35x: 95°C for 30 s, 64°C for 90 s, 72°C for 1 min, Step 3–72°C for 5 min, resulting in a 358 bp amplicon for the wild-type allele and a 275 bp amplicon for the mutant allele. Rreb1^-/- mice were embryonic lethal at midgestation but no peri-natal lethality was observed for Rreb1^-/+ mice. Therefore, the Rreb1 mouse line was maintained and Rreb1^-/- embryos were obtained through heterozygous Rreb1^-/+ intercrosses.

Approximately 10–15 Rreb1^-/- ESCs, described in Su et al., 2020, harboring a constitutive mCherry fluorescent lineage tracer were injected into E3.5 blastocysts (C57BL/6J, Jackson Laboratory, Bar Harbor, ME) as previously described (Su et al., 2020). Injected blastocysts were cultured in KSOM/AA (Millipore, Billerica, MA) at 37°C in an atmosphere of 5% CO2 to allow for recovery of blastocyst morphology and then implanted into the uterine horns (up to 10 embryos per horn) of E2.5 pseudopregnant females (C57BL/6J;CBA F1, Jackson Laboratory) using standard protocols. Chimeric embryos were recovered between E7.5-E9.5.

To produce the Rreb1 riboprobes, RNA was isolated from pooled E12.5 CD1 mouse embryos using an RNeasy Plus Mini Kit (Qiagen, Hilden, Germany) and then used to generate cDNA with a QuantiTect Reverse Transcription Kit (Qiagen), as per manufacturer’s instructions. Primers (5’ UTR L: GGGCCTTTGTCTCATGCTCC, 5’ UTR R: CGCAGAATGTTTTCCTCAACAG) were designed against a unique 502 bp region within the Rreb1 5’ UTR and used to PCR amplify this fragment from E12.5 embryo cDNA. The PCR product was purified using a QIAquick PCR Purification Kit (Qiagen) and a TOPOTA Cloning Kit (K461020, Thermo Fisher Scientific) used to introduce the fragment into a pCRII-TOPO Vector and transformed into E. coli. Colonies were picked, expanded and the plasmid isolated for sequencing. A plasmid containing the correct sequence (5’-CGCAGAATGTTTTCCTCAACAGTTGACAATTTTAGGATAAATAGAACTTTAGAAAAATTACTACTATCAATCATCTAAGTATTCCGAATAGGAAAAAAAGTCAAAATAAGTAAGGGACGCTGGAGCTACCTCAGTGAAGGGGAAAAAATATCCAATCCCACTTTTCTGTATTACATGTGTGGTAGCTAAAGAACTCCATAGAATGTTCAAAAAAAAAAAAAAAAGACGGCACTGAAGATTATCATGTCAAAGCACCAAGCTCATTACATCACTGTTACCTTAATGCAAAGTCCCACTTCTCCGGAATGGCCTCCATACTTAGAAACTCTTGGAACTTGTCAGGCAAAGGTTATGGGGAGGGAAGTGAAGGAGCCTATGACCACTGTCACTGTGTCTGATACATTTATTTACAGATAAGCCTTGGTGGCTCAGACCACAGGCACAGATTATATGGAAAGTAACAGCCTGTGACTTCTGAGACAAAGAATGGAGCATGAGACAA-3’) was selected, linearized and the dual promoter system within the pCRII-TOPO Vector used to amplify and DIG label both a control sense and an antisense probe. Wholemount mRNA in situ hybridization was then carried out as previously reported (Conlon and Rossant, 1992).

X-gal staining of cells and embryos containing the Rreb1-LacZ reporter was performed using a β-Gal Staining Kit (K146501, Invitrogen, Waltham, MA) as per manufacturer’s instructions. Embryos and cells were fixed for 15 mins at room temperature followed by staining until the blue color was detectable (2–3 hr) at 37°C.

Cells were maintained in standard serum/LIF ESC medium (Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, Gaithersburg, MD) containing 0.1 mM non-essential amino-acids (NEAA), 2 mM glutamine and 1 mM sodium pyruvate, 100 U/ml Penicillin, 100 μg/ml Streptomycin (all from Life Technologies, Carlsbad, CA), 0.1 mM 2-mercaptoethanol (Sigma, St. Louis, MO), and 10% Fetal Calf Serum (FCS, F2442, Sigma) and 1000 U/ml LIF) as previously described (Morgani et al., 2018a). C57BL/6N-A^tm1Brd Rreb1^{tm1a(EUCOMM)Wtsi}/WtsiPh (RRID:IMSR_EM:10996) embryonic stem (ES) cell lines were used to analyze Rreb1 expression and also converted to an epiblast stem cell (EpiSC) state through prolonged culture (more than five passages) in N2B27 medium containing 12 ng/ml FGF2 (233-FB-025, R and D Systems) and 20 ng/ml ACTIVIN A (120–14P, Peprotech, Rocky Hills, NJ), as previously described (Tesar et al., 2007). ES cells were routinely tested and found to be mycoplasma negative. They were validated by genotyping and used to generate germline transmitting mouse chimeras.

Cell lines were immunostained as previously described (Morgani et al., 2018a). Post-implantation embryos were fixed in 4% paraformaldehyde (PFA) for 15 min at room temperature (RT). Embryos were washed in phosphate-buffered saline (PBS) plus 0.1% Triton-X (PBST-T) followed by 30 min permeabilization in PBS with 0.5% Triton-X. Embryos were washed in PBS-T and then blocked overnight at 4°C in PBS-T, 1% bovine serum albumin (BSA, Sigma) and 5% donkey serum. The following day, embryos were transferred to the primary antibody solution (PBS-T with appropriate concentration of antibody) and incubated overnight at 4°C. The next day, embryos were washed 3 × 10 min in PBS-T and then transferred to blocking solution at RT for a minimum of 5 hr. Embryos were transferred to secondary antibody solution (PBS-T with 1:500 dilution of appropriate secondary conjugated antibody) and incubated overnight at 4°C. Embryos were then washed 3 × 10 min in PBS-T with the final wash containing 5 μg/ml Hoechst. Where F-ACTIN staining was performed, Alexa Fluor conjugated phalloidin (Thermo Fisher Scientific, Waltham, MA) was added to the primary and secondary antibody solutions at a 1:500 dilution.

The following primary antibodies were used in this study: Beta-catenin (RRID:AB_397555, BD Transduction labs, Billerica, MA, 610154, 1:500), Brachyury (RRID:AB_2200235, R and D, AF2085, 1:100), CD31 (RRID:AB_394819, BD Biosciences, 553373, 1:100) CD105 (RRID:AB_354735, R and D Systems, AF1320, 1:100), E-cadherin (RRID:AB_477600, Millipore Sigma, U3254, 1:200), Endoglin (CD105) (RRID:AB_354735, R and D Systems, AF1320, 1:300), Flk-1 (RRID:AB_355500, R and D Systems, AF644, 1:200) Gata6 (RRID:AB_10705521, D61E4 XP, Cell Signaling, 5851, 1:500), GFP (RRID:AB_300798, Abcam, ab13970), Laminin (RRID:AB_477163, Millipore Sigma, L9393, 1:500), N-cadherin (RRID:AB_2077527, BD Biosciences, 610920, 1:200), Pecam-1 (RRID:AB_394819, BD Biosciences, 553373, 1:200), RFP (Rockland, Limerick, PA, 600-400-379, 1:300), Snail (RRID:AB_2191738, R and D Systems, AF3639, 1:50), Sox2 (RRID:AB_11219471, Thermo Fisher Scientific, 14-9811-82, 1:200), Sox17 (RRID:AB_355060, R and D Systems, AF1924, 1:100), ZO-1 (RRID:AB_87181, Invitrogen, 33–9100, 1:200).

Embryos were oriented as desired and embedded in Tissue-Tek OCT (Sakura Finetek, Japan). Samples were frozen on dry ice for approximately 30 min and subsequently maintained for short periods at −80°C followed by cryosectioning using a Leica CM3050S cryostat. Cryosections of 10 µm thickness were cut using a Leica CM3050S cryostat and mounted on Colorfrost Plus microscope slides (Fisher Scientific) using Fluoromount G (RRID:SCR_015961, Southern Biotech, Birmingham, AL) and imaged using a confocal microscope as described.

Embryos were imaged on a Zeiss LSM880 laser scanning confocal microscope. Whole-mount embryos were imaged in glass-bottom dishes (MatTek, Ashland, MA) in PBS. Raw data were processed in ImageJ open-source image processing software (Version: 2.0.0-rc-49/1.51d).

Nuclei orientation (Figure 5—figure supplement 1E–G) was measured manually using Fiji (RRID:SCR_002285, Image J) software. Using the angle tool, we measured the angle between the long axis of individual epiblast nuclei and the underlying basement membrane, marked by Laminin staining on confocal optical sections of transverse cryosections. We measured the angle of 143 cells from 3 Rreb1^+/+ embryos and 136 cells from 3 Rreb1^-/- embryos.

We quantified proliferation in Rreb1^+/+ versus Rreb1^-/- embryos (Figure 5—figure supplement 1L) by manually counting the number of phosphorylated histone H3 (pHH3) positive cells in the epiblast, outer endoderm layer or wings of mesoderm in transverse cryosections of Rreb1^+/+or Rreb1^-/- embryos. Initially, cell counts were also categorized as divisions in anterior versus posterior embryonic regions but, as no differences were observed, these data were subsequently combined. We performed counts on cryosections comprising three entire embryos per genotype. Data was analyzed as the absolute numbers of dividing cells per cell type. Additionally, we counted the total number of cells per cell type per section and normalized the number of dividing cells to this value to account for differences based on embryo or tissue size. Statistics were performed on a per embryo rather than a per cell basis.

The level of GFP in the VE of Afp-GFP; Rreb1^+/+ and Rreb1^-/- embryos was quantified by manually selecting the embryonic and extraembryonic region of confocal maximum intensity projection images and measuring the mean fluorescence intensity using Fiji software.

Quantification of SOX2 protein levels (Figure 6—figure supplement 1F) were carried out on cryosections of Rreb1^-/- chimeric embryos containing cells expressing high levels of SOX2 (SOX2^HI cells) to determine the approximate fold change in protein level relative to normal surrounding cells. To make measurements, nuclei were manually identified using the freehand selection tool in Fiji software. Aberrant SOX2^HI cells could readily be distinguished from standard neighboring cells by their elevated signal after immunostaining for SOX2 protein. Mean fluorescence intensity of SOX2 immunostaining was measured within all SOX2^HI nuclei within a particular cryosection and an equivalent number of randomly selected nuclei with normal SOX2 expression within the anterior and posterior epiblast regions were measured. Mean SOX2 fluorescence intensity in each nucleus was normalized to the corresponding mean fluorescence intensity of the Hoechst nuclear stain. All data is shown relative to the mean SOX2 fluorescence intensity measured in ‘normal’ anterior epiblast cells of the same confocal optical section. A total of 8 embryos, 35 cryosections, and 696 cells were analyzed. Statistics were carried out on the average fluorescence levels per embryo.

The localization of SOX2^HI cells (identified manually from SOX2 immunostaining) (Figure 6—figure supplement 1G) was scored based on their location within confocal images of cryosectioned Rreb1^-/- chimeric embryos. Scoring was carried out on 76 cryosections from seven independent embryos that contained high numbers of SOX2^HI cells. SOX2^HI cells were scored as being within the Epi itself, at the Epi-VE interface (outside of the epiblast epithelium), within the primitive streak or wings of mesoderm (mesoderm) or within the amniotic cavity.

Statistical analysis of significance was assessed using a one-way ANOVA (p<0.0001) followed by unpaired t-tests to compare particular groups (GraphPad Prism, RRID:SCR_002798, GraphPad Software, Inc, Version 7.0a).

Frozen tissue was homogenized in TRIzol Reagent (ThermoFisher catalog # 15596018) using the QIAGEN TissueLyser at 15 Hz for 2–3 min with a Stainless-Steel Bead (QIAGEN catalog # 69989). Phase separation was induced with chloroform. RNA was precipitated with isopropanol and linear acrylamide and washed with 75% ethanol. The samples were resuspended in RNase-free water. After RiboGreen quantification and quality control by Agilent BioAnalyzer, 150 g of total RNA underwent polyA selection and TruSeq library preparation according to instructions provided by Illumina (TruSeq Stranded mRNA LT Kit, catalog # RS-122–2102), with 8 cycles of PCR. Samples were barcoded and run on a HiSeq 4000 in a 50 bp/50 bp paired-end run, using the HiSeq 3000/4000 SBS Kit (Illumina). An average of 47 million paired reads was generated per sample. The percent of mRNA bases averaged 67%.

The output data (FASTQ files) were mapped to the target genome using the rnaStar aligner (Dobin et al., 2013) that maps reads genomically and resolves reads across splice junctions. We used the two pass mapping method outlined in Engström et al., 2013, in which the reads are mapped twice. The first mapping pass uses a list of known annotated junctions from Ensemble. Novel junctions found in the first pass were then added to the known junctions and a second mapping pass is done (on the second pass the RemoveNoncanoncial flag is used). After mapping, we post-processed the output SAM files using the PICARD tools to: add read groups, AddOrReplaceReadGroups which in additional sorts the file and converts it to the compressed BAM format. We then computed the expression count matrix from the mapped reads using HTSeq (https://www.huber.embl.de/users/anders/HTSeq/doc/overview.html) and one of several possible gene model databases. The raw count matrix generated by HTSeq was then processed using the R/Bioconductor package DESeq (https://www.huber.embl.de/users/anders/DESeq/) which is used to both normalize the full dataset and analyze differential expression between sample groups. The data was clustered in several ways using the normalized counts of all genes that a total of 10 counts when summed across all samples; 1. Hierarchical cluster with the correlation metric (Dij = 1 - cor(Xi,Xj)) with the Pearson correlation on the normalized log2 expression values. 2. Multidimensional scaling. 3. Principal component analysis. Heatmaps were generated using the heatmap.2 function from the gplots R package. For the Heatmaps the top 100 differentially expressed genes are used. The data plot represents the mean-centered normalized log2 expression of the top 100 significant genes. We ran a gene set analysis using the GSA package with gene sets from the Broads mSigDb. The sets used were: Mouse: c1, c2, c3, c4, c5. Gene ontology analyses were performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) Bioinformatics resource (Version 6.8) gene ontology functional annotation tool (http://david.abcc.ncifcrf.gov/tools.jsp) with all NCBI Mus musculus genes as a reference list. KEGG pathway analysis was performed using the KEGG Mapper – Search Pathway function (https://www.genome.jp/kegg/tool/map_pathway2.html). We performed a manual literature search to determine the proportion of significantly changing genes associated with cancer progression and metastasis.

The Gene Expression Omnibus accession number for the RNA-sequencing data reported in this study is GSE148514.