Nanog regulates Pou3f1 expression at the exit from pluripotency during gastrulation

Pluripotency is a steady state in which cells can self-renew and remain undifferentiated, retaining the capacity to give rise to derivatives of any germ layer. This cell state is maintained by an intricate gene regulatory network (GRN) that is tightly regulated by a core set of transcription factors (TF): NANOG, OCT4 and SOX2 (Navarro et al., 2012; Rodda et al., 2005; Teo et al., 2011; Trott and Martinez Arias, 2013). These three TFs are involved in establishing and maintaining embryonic pluripotency, both in the blastocyst and in cultured embryonic stem (ES) cells (Chambers and Tomlinson, 2009). This GRN regulates pluripotency by repressing genes involved in differentiation and activating other genes important for pluripotency (Navarro et al., 2012; Thomson et al., 2011). The same GRN also initiates the process of exiting pluripotency by responding to extrinsic and intrinsic signals and changing the regulatory regions and partners these factors bind to (Hoffman et al., 2013; Kalkan and Smith, 2014; Mohammed et al., 2017; Pfeuty et al., 2018).

ES cells can be maintained in different stages of differentiation, the most studied being ground, naïve and primed pluripotent cells (Joo et al., 2014; Morgani et al., 2017; Nichols and Smith, 2009). These states are defined so that they can be maintained and passaged in vitro indefinitely: the first with Leukemia Inhibitory Factor (LIF) and 2i (MEK and GSK3 inhibitors), the second with LIF and serum (Ying et al., 2008), and the latter with Activin and FGF (Tesar et al., 2007). However, while ground and naïve pluripotent cells contribute to all embryonic lineages in blastocyst chimeras, cells in the primed state have lost this potential (Festuccia et al., 2013). In ground pluripotent ES cells, NANOG is highly and homogeneously expressed, while in the primed ES cells NANOG expression levels fluctuate. Transition between these two cell states determines the onset of differentiation. In fact, it has been demonstrated that lowering levels of Nanog expression in ES cells triggers differentiation and its overexpression is sufficient to maintain the cells in a LIF-independent pluripotent state (Chambers et al., 2007). In spite of multiple studies that have addressed ES cell differentiation (Mendjan et al., 2014; Radzisheuskaya et al., 2013; Thomson et al., 2011), the role of NANOG during the exit from pluripotency in vivo is still not well understood (Osorno et al., 2012; Tam and Behringer, 1997). During implantation, Nanog disappears from the epiblast and is re-expressed in the proximal posterior region of the epiblast after implantation, the region in which gastrulation starts (Hart et al., 2004). Thus, we hypothesized that Nanog not only has a role in pluripotency maintenance, but also in defining lineage commitment upon gastrulation (Mendjan et al., 2014). We have recently shown that, at the onset of gastrulation, Nanog has a determinant role in repressing primitive hematopoiesis and Hox genes expression (Lopez-Jimenez et al., 2019 preprint; Sainz de Aja et al., 2019).

To gain further insight into the roles of Nanog beyond pluripotency, we studied the effects of altering the levels of NANOG in different ES cell lines and in mouse embryos. By combining the analysis of different RNA-seq data sets, we found that Pou3f1 expression is regulated by Nanog. Pou3f1, which encodes a TF involved in promoting neural fate, is initially expressed throughout the epiblast at early implantation stages (Song et al., 2015; Zhu et al., 2014). However, at the onset of gastrulation when Nanog is re-expressed in the embryo (Yamaguchi et al., 2005), its expression becomes quickly restricted to the anterior epiblast. While the role of POU3F1 in antagonizing extrinsic neural inhibitory signals is well known (Zhu et al., 2014), little information is available about the transcriptional regulation of this gene in the early stages of gastrulation. By deleting NANOG binding sites located next to the Pou3f1 locus, we observed that Nanog prevents the expression of Pou3f1 in the posterior region of the gastrulating embryo. Therefore, we present a previously unknown mechanism by which Nanog constrains Pou3f1 expression to the anterior region of the embryo, a necessary step for its role in neural development.

To explore the role of Nanog and to identify putative targets during the transition from pluripotency to lineage specification, we analyzed expression changes in ES cells mutant for Nanog and compared them to the parental wild-type ES cell line as control (Chambers et al., 2007). Cells were first cultured with 2i/LIF/KOSR and subsequently changed to serum to induce exit from pluripotency (Heo et al., 2005; Martin Gonzalez et al., 2016). To follow the earliest events taking place, we sampled the cultures at 0, 12, and 24 h (Fig. 1A; Table S1). Then, we performed RNA-seq and selected genes that changed their expression dynamics from 0–24 h. We identified genes repressed by Nanog as those with stable expression in control ES cells but increased expression in Nanog KO cells along time (Fig. 1B), and genes that are positively regulated by Nanog as those activated in controls but unchanged in mutant cells (Fig. 1C).

Principal component analysis (PCA) of the RNA-seq data showed a clear separation of the samples based on the genotype of the cells (dim1, Fig. S1) and timing of differentiation (dim2, Fig. S1). The genotypic difference resulted in close to 43% variability, whereas timing of differentiation explained 26% of the variability. Interestingly, the comparison at time 0 between control and Nanog KO ES cells showed minimal differences in the expression of core pluripotency genes like Oct4 or Sox2 (Fig. 1D). The similarities between Nanog KO and wild-type cells in the pluripotent stage agree with previous observations on the dispensability of Nanog at the pluripotent state (Chambers et al., 2007). We analyzed changes in gene expression, factoring in their expression over time, and identified two clusters with the predicted pattern of change (Fig. 1E,F). Genes that are upregulated in differentiating Nanog mutant cells but not in controls are enriched in neural specifiers such as Pou3f1, Sox8 or Fgf3 (Fig. 1G; Table S1A) (Bell et al., 2000; O'Donnell et al., 2006; Zhu et al., 2014). On the other hand, genes that fail to be upregulated in mutant cells are involved in mesoderm development, such as Mef2c, Hand2 or Tbx2 (Fig. 1G; Table S1B). This analysis indicates that Nanog might be involved in the repression of genes implicated in the development of the anterior-neural fate while promoting posterior-mesodermal fate at the exit from pluripotency in ES cells.

To further explore the putative role of Nanog in neural anterior fate in vivo, we took advantage of published E6.5 embryo single-cell RNA-sequencing (scRNA-seq) data (Mohammed et al., 2017; Scialdone et al., 2016). E6.5 is the stage at which Nanog is re-expressed in the posterior part of the mouse embryo (Hart et al., 2004) and several genes including Sox2 or Pou3f1 are already restricted to the epiblast. We merged two single cell RNA-seq expression data sets and selected those single cells expressing Nanog above 0.4 cpm. The expression of all the genes with 0.4 cpm in at least four cells of at least two samples were adjusted with a linear mixed effect model to the expression of Nanog (Tables S1C, S1D). Next, we established the correlation of all expressed genes to that of Nanog (Fig. 2A; Tables S1C,D). These results confirmed our previous observations in cultured cells. Genes that correlated positively with Nanog were related to gastrulation and mesoderm formation, such as Fgf8, Nodal or Eomes (Fig. 2B; Table S1C). Genes that negatively correlated with Nanog include Pou3f1 and other neural genes such as Nav2 (Fig. 2B; Fig. S2A, Table S1D). Other early anterior genes, such as Sox2, did not show any correlation with Nanog levels (Fig. S2B), suggesting that Nanog might not have a broad impact on anterior specification, but rather has a specific effect on certain genes. Interestingly, among the negatively correlated genes we also found Utf1 (Fig. 2B), a pluripotency associated gene that is restricted to the anterior region of the embryo during gastrulation and to extraembryonic tissues (Okuda et al., 1998). Enrichment analysis of the clustered genes matching the Jansen tissues gene set library (Chen et al., 2013; Kuleshov et al., 2016), allowed us to observe that negatively correlated genes included many genes related to neural development (spinal cord, frontal lobe), and with a lower z-score than endodermal tissues (gut, intestine) (Fig. S2C). Genes that positively correlate with Nanog expression included many genes related to mesodermal tissues (monocyte, B lymphoblastoid cell, bone) (Fig. S2D).

We next addressed the effect of expressing Nanog throughout the early embryo when using an inducible tetON transgenic model (Nanog^tg) in which Nanog expression is induced by the administration of doxycycline (dox) (Piazzolla et al., 2014). We analyzed bulk RNA-seq data of embryos where Nanog was induced from E4.5 to E7.5 and examined changes in gene expression using untreated females of the same genotype as controls (Lopez-Jimenez et al., 2019 preprint). In this dataset, many genes involved in the early aspects of embryo pattering, such as Hox genes, were downregulated (Lopez-Jimenez et al., 2019 preprint), but the most strongly downregulated gene when Nanog was expressed throughout the early embryo was Pou3f1 (Fig. 2C). The expression of other anterior neural genes, for example Sox2, Hesx1 or Zic3, was not changed. We confirmed these observations by wholemount in situ hybridization of treated and untreated E7.5 Nanog^tg embryos. Induction of Nanog led to a partial downregulation of Pou3f1 in the anterior epiblast of treated embryos, while expression of Sox2 was unchanged (Fig. 2D). Interestingly, when Nanog^tg embryos were recovered at E9.5 after treatment with dox from E6.5, they presented craniofacial defects (white arrowheads) that might be a direct consequence of the deregulation of yet to be identified Nanog target genes (Fig. S2E).

We merged the data from these previous transcriptomic analysis, finding for example that genes whose expression positively correlated with that of Nanog in E6.5 single cells and that were upregulated in dox-treated Nanog^tg embryos were mostly related to early gastrulation and mesoderm specification, such as Eomes, Fgf8, Tdgf1 (Cripto) or Mixl1. However, only three genes were shared by those upregulated in Nanog KO ES cells during early differentiation, genes having a significant negative correlation with Nanog in E6.5 single cell transcriptomics, and that were downregulated in E7.5 Nanog gain-of-function embryos: Pou3f1, Lrp2 and Clic6 (Fig. 2E). Interestingly, Lrp2 and Clic6 are expressed in primitive endoderm and late derivatives (Gerbe et al., 2008; Sherwood et al., 2007), which are lineages in which Nanog has a well-defined negative regulatory role (Chazaud et al., 2006; Dietrich and Hiiragi, 2007; Yamanaka et al., 2010). Therefore, Pou3f1 is a prime candidate to be a direct target of Nanog, mediating its role in suppressing anterior epiblast fate.

To confirm whether Nanog is blocking anterior fate progression, we derived ES cells from the Nanog^tg line and differentiated them towards anterior neural fate (Gouti et al., 2014, 2017), culturing them with or without dox for up to 6 days. Analysis of gene expression by RT-qPCR showed that upon induction of Nanog, neural specification genes (Pou3f1, Sox1, Pax6 and Otx2) were not upregulated during the differentiation process to the same extent as in control-treated cells. Sox2, which has roles both in pluripotency and in early neural development, showed a similar pattern of expression by qPCR in both Nanog^tg +dox and −dox (Fig. 3A). Immunofluorescence of TUJ1, revealed lack of differentiation at a protein level in the differentiation of Nanog^tg cells treated with dox (Fig. 3B). When cells were differentiated towards a more posterior neural fate by treatment with high doses of retinoic acid (Gouti et al., 2014, 2017), differences in the expression of neural markers were less marked, although following a similar trend (Fig. S3). We also observed a reduction in the expression of Hoxa1, a marker for posterior neural (hindbrain) fate (Fig. S3), in line with recent findings (De Kumar et al., 2017; Lopez-Jimenez et al., 2019 preprint). These results indicate that during neural differentiation, Nanog prevents the upregulation of genes important for neural specification.

The evidence presented so far suggests that Pou3f1 is likely a direct transcriptional target of NANOG during anterior-posterior axis specification in the epiblast. To explore this possibility, we analyzed published ChIP-seq data for NANOG binding in ES and epiblast-like cells (EpiLCs) (Murakami et al., 2016). This work describes a broad resetting of NANOG-occupied genomic regions in the transition from ES cells to EpiLCs, resembling the developmental progress from the naïve inner cell mass of the blastocyst to the primed epiblast at gastrulation. We examined the Pou3f1 locus and identified three prominent regions of NANOG binding at 11.5 and 9 kilobases (kb) upstream and 9 kb downstream from the transcription start site. Interestingly, NANOG binds these regions in EpiLC but not in ES cells, suggestive of a specific input of Nanog on Pou3f1 in the epiblast but not at earlier pluripotent stages (Fig. 4A).

We hypothesized that the deletion of the NANOG bound regions would de-repress Pou3f1 and therefore expand its expression domain towards the posterior region of the embryo at E6.5. To investigate this hypothesis, we separately deleted each of the three NANOG bound regions we identified at the Pou3f1 locus by CRISPR/Cas9 genome editing in a transient transgenic embryo assay. Embryos microinjected with Cas9-gRNAs ribonucleoproteins were recovered at E6.5, processed for wholemount in situ hybridization, and subsequently genotyped for the expected deletion (Fig. S4). This assay showed that only deletion of the +9 kb downstream region caused a reproducible change in Pou3f1 expression, consisting in a posterior expansion of its expression domain in the epiblast (Fig. 4B,C). To further confirm this observation, we generated a stable mouse line carrying the deletion of the +9 kb NANOG bound region (Fig. 4D). Mice homozygous for the deletion were viable and fertile, which was not completely unexpected as homozygous null Pou3f1 mice survive up to birth (Bermingham et al., 1996). We crossed mice heterozygous and homozygous for the deletion and compared littermates for the expression of Pou3f1 by wholemount in situ hybridization. We observed that three out of five homozygous embryos presented a phenotype of posterior expansion while none of the heterozygous embryos did so (Fig. 4E). These results show that the +9 kb NANOG bound region is important for the restriction of Pou3f1 expression to the anterior epiblast.

The transition from pluripotency towards early differentiation that occurs during the initial stages of mammalian embryonic development can be recapitulated in vitro, at least partially, under defined culture conditions (Hackett and Surani, 2014; Nichols and Smith, 2009). This provides an opportunity to study and follow different pluripotent states, as cells move from the naïve or ground state, equivalent to the epiblast of the blastocyst, to a primed state that more closely resembles the early post-implantation embryo, where first lineage decisions take place (Yang et al., 2019). The role of a core set of transcription factors (that includes OCT4, NANOG or SOX2) controlling the establishment and maintenance of embryonic pluripotency has been extensively studied (Chambers and Tomlinson, 2009). Moreover, recent studies indicate that these same factors can play important roles in regulating the exit from pluripotency towards committed states (Festuccia et al., 2013) as well as later developmental decisions in the embryo (Aires et al., 2016; Lopez-Jimenez et al., 2019 preprint; Sainz de Aja et al., 2019).

In this work, we sought to capture the first steps in the differentiation of naïve ES cells to assess the role of Nanog in the exit from pluripotency. We observed that globally, Nanog represses the differentiation of the anterior fate while promoting posterior differentiation. In an effort to identify direct transcriptional targets of NANOG in this process, we merged three sets of data: transcriptomic data from naïve-to-primed differentiation of Nanog mutant ES cells; single-cell RNA-seq data from E6.5 embryos (Mohammed et al., 2017; Scialdone et al., 2016); and transcriptional analysis of the forced expression of Nanog in E7.5 embryos (Lopez-Jimenez et al., 2019 preprint). Only three genes were identified that met the requirements to be negatively regulated by Nanog (upregulated during differentiation of Nanog KO cells, negatively correlated with Nanog in scRNA-seq data, and downregulated upon Nanog expression in embryos). Two of them, Lrp2 and Clic6, are most prominently expressed in the primitive endoderm and later in other endoderm derivatives (Nowotschin et al., 2019), and in fact LRP2 had been previously identified as a marker of primitive endoderm precursors of the blastocyst (Gerbe et al., 2008). This fits well with the known role of Nanog in the epiblast/primitive endoderm decision occurring in the preimplantation embryo (Bassalert et al., 2018; Frankenberg et al., 2011) and suggests that Lrp2 and Clic6 could be directly repressed by NANOG in epiblast cells of the blastocyst.

The third gene identified as a potential NANOG target is Pou3f1. This gene encodes a POU family transcription factor, initially expressed throughout the epiblast of E6.5 embryos and later restricted to the anterior epiblast at E7.5 and afterwards, to the anterior neural tube (Iwafuchi-Doi et al., 2012; Zwart et al., 1996). Pou3f1 has also been shown to drive the progression of neural differentiation both in ES cells and in epiblast like stem cells (EpiLCs) through the activation of intrinsic neural lineage genes such as Sox2 or Pax6 (Song et al., 2015; Zhu et al., 2014). Pou3f1 is strongly downregulated in embryos in which we induce the expression of Nanog, and we observe the same in ES cells differentiated towards neural fates. Interestingly, this effect is more pronounced when ES cells are directed to anterior/forebrain identities rather than posterior/hindbrain fates (Gouti et al., 2014; 2017), resembling the dynamics of Pou3f1 expression in the developing neural tube (Zwart et al., 1996).

Analysis of published ChIP-seq data in ES and EpiL cells (Murakami et al., 2016) allowed us to identify three genomic regions that could be mediating the transcriptional repression of Pou3f1 by NANOG. It is noteworthy that these sites are not bound in ES cells but only in EpiLCs, indicating that NANOG is not simply repressing Pou3f1 as part of the core pluripotency program, but involved in fine-tuning the timing of its expression once differentiation programs are initiated at the primed state. We detected expansion of Pou3f1 expression to the posterior epiblast only when the +9 kb region was deleted. However, this does not rule out a possible input of the other two regions (−11.5 kb and −9 kb); had all three regions been deleted in the same embryo, we might have observed a more robust de-repression of Pou3f1.

The results we describe here, together with our previous observations regarding the function of Nanog in primitive hematopoiesis (Sainz de Aja et al., 2019), suggest that Nanog has an active role in the primed epiblast as a brake for ongoing lineage determination. Apparently, this occurs through well-known neural (Pou3f1) or mesodermal (Tal1; Sainz de Aja et al., 2019) specifiers, but it is tempting to speculate, based on our observations for Lrp2 and Clic6, that it is also occurring in endodermal lineages. Only when Nanog expression is extinguished, transcriptional repression of these epiblast-specific targets lifted, and differentiation allowed to proceed. Understanding the regulatory mechanism that controls the re-expression of Nanog in the epiblast (Hart et al., 2004) and how it is definitively silenced will allow us to better understand how pluripotency is dismantled and how particular lineage specific programs come to be deployed.

ES cells were maintained in serum-free conditions with Knockout serum replacement (Thermo Fisher Scientific), LIF (produced in-house), and 2i (CHIR-99021, Selleckchem; and PD0325901, Axon) over inactive mouse embryonic fibroblast (MEFs). The Nanog^KO (BT12) and their parental wild-type control (E14Tg2a) ES cells were kindly provided by Ian Chambers (MRC Centre for Regenerative Medicine, Edinburgh) and Austin Smith (Wellcome-MRC Cambridge Stem Cell Institute) (Chambers et al., 2007). The Nanog gain-of-function ES cells were derived from the Nanog/rtTA mouse line following standard procedures (Nagy et al., 2014). Karyotyping of the obtained lines was performed by the Pluripotent Cell Technology Unit at CNIC.

Nanog^tg ES cells were differentiated to anterior (forebrain) or posterior (hindbrain) neural lineages as described (Gouti et al., 2014, 2017). Cells were grown in N2B27 media supplemented with 10 ng/ml bFgf (R&D) for the first 3 days (d1–d3), and then from day 3 to 6 in N2B27 without growth factors for forebrain differentiation, or in N2B27 supplemented with 10 nM retinoic acid for hindbrain differentiation. A Tuj1 monoclonal antibody (Promega #G7121, dilution 1:100) was used to assess neuronal differentiation. All cell lines have been tested for contamination at the CNIC Cell Culture Unit.

RNA from Nanog^KO ES cells and their parental line was extracted using the RNeasy Mini Kit (Qiagen) and then reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Library preparation (New England Biolabs Nest Ultra RNA library prep Kit) and single read next generation sequencing (Illumina HiSeq 2500) were performed at the Genomics Unit at Centro Nacional de Investigaciones Cardiovasculares (CNIC).

Sequencing reads were processed by means of a pipeline that used FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc) to asses read quality, and Cutadapt v1.3 (Martin, 2011) to trim sequencing reads, eliminating Illumina adaptor remains and to discard reads that were shorter than 30 bp. The resulting reads were mapped against the mouse transcriptome (GRCm38 assembly, Ensembl release 76) and quantified using RSEM v1.2.20 (Li and Dewey, 2011). Raw expression counts were processed with an analysis pipeline that used Bioconductor packages EdgeR (Robinson et al., 2010) for normalization (using TMM method) and differential expression testing, and ComBat (Johnson et al., 2007) for batch correction. Only genes expressed at a minimal level of 1 count per million, in at least three samples, were considered for differential expression analysis. Changes in gene expression were considered significant if their Benjamini and Hochberg adjusted P-value (FDR) was lower than 0.05.

RNA-seq data is available at the Gene Expression Omnibus (GEO) database under accession number GSE138818.

Two data sets from different mouse embryo single cell RNA-seq experiments (Mohammed et al., 2017; Scialdone et al., 2016) were normalized by quantiles and batch corrected. After merging the two datasets, genes with zero-variance were eliminated and counts were log2 transformed and scaled. Then, datasets were normalized using the quantiles method and batch corrected. Single cell clustering patterns were visualized after dimensionality reduction with the R package Rtsne. For correlation of genes with Nanog we used the slope of the line adjusted to the points per sample. For plotting we used ggPlot package from R. We separated the plots by sample. Statistical analysis was developed in R. RNA-seq data from E7.5 Nanog^tg embryos was previously described (Lopez-Jimenez et al., 2019 preprint). The intersection analysis of the genes coming from different RNA-seq datasets was performed with the web tool from Bioinformatics and Evolutionary Genomics (http://bioinformatics.psb.ugent.be/webtools/Venn/).

RNA was isolated from ESCs using the RNeasy Mini Kit (Qiagen) and then reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). cDNA was used for quantitative-PCR (qPCR) with Power SYBR^® Green (Applied Biosystems) in a 7900HT Fast Real-Time PCR System (Applied Biosystems). Primers for qPCR detailed in Table S2.

For the generation of transgenic embryos, 7-week-old F1 (C57Bl/6xCBA) females were superovulated to obtain fertilized oocytes as described (Nagy et al., 2014). Viable one-cell embryos were microinjected into the pronucleus with commercially available Cas9 protein (30 ng/µl; PNABio) and guide RNAs (sgRNA; 25 ng/µl; Sigma-Aldrich). All those components were previously hybridized in solution to generate ribonucleoprotein complexes. First, we incubated 100 ng/µl of trans-activating crRNA (tracrRNA) and sgRNA for 5 min at 95°C and then for 10 min at room temperature (RT). We then incubated the sgRNAs with the Cas9 for 15 min at RT and stored at 4°C. Injection buffer consisted of Tris 50 nM pH7.4, EDTA 1 nM, H₂O embryo tested and was filtered through a 0.22 um filter. After injection, embryos were cultured in M16 (Sigma-Aldrich) covered with mineral oil (Nid Oil, EVB) up to the two-cell stage. Living embryos were then transferred into a pseudopregnant CD1 female, previously crossed with a vasectomized male. Embryos were recovered at E6.5 for further analysis, or allowed to progress in order to establish a stable line carrying the deletion of the +9 kb region. sgRNA were designed with an online tool (http://crispr.mit.edu/). Details of the sequences for the sgRNAs and primers used for genotyping are shown in Table S2.

We obtained the Nanog/rtTA mouse line (R26-M2rtTA;Col1a1-tetO-Nanog) (Piazzolla et al., 2014) from Manuel Serrano (CNIO, Madrid) and Konrad Hochedlinger (Harvard Stem Cell Institute). This is a double transgenic line that carries the M2-rtTA gene inserted at the Rosa26 locus and a cassette containing Nanog cDNA under the control of a doxycycline-responsive promoter (tetO) inserted downstream of the Col1a1 locus. Mice were genotyped by PCR of tail-tip DNA as previously described (Hochedlinger et al., 2005; Piazzolla et al., 2014).

All mice used in this work were housed and maintained in the animal facility at the Centro Nacional de Investigaciones Cardiovasculares (Madrid, Spain) in accordance with national and European Legislation. Procedures were approved by the CNIC Animal Welfare Ethics Committee and by the Area of Animal Protection of the Regional Government of Madrid (ref. PROEX 196/14).

Embryos were collected in cold PBS, transferred to 4% PFA, and fixed overnight at 4°C. After washing, embryos were dehydrated in increasing concentrations of PBS-diluted methanol (25%, 50%, 75% and 2X 100%). In situ hybridization in wholemount embryos was performed as previously described (Acloque et al., 2008; Ariza-McNaughton and Krumlauf, 2002). Signal was developed with anti-digoxigenin-AP (Roche) and BM-Purple (Roche). Images were acquired with a Leica MZ-12 dissecting microscope. Primers used for the generation of probes are detailed in Table S2. The expansion of Pou3f1 expression into posterior territories was scored on images of embryos after wholemount in situ. For this, embryos were divided vertically by an imaginary line that separated anterior from posterior halves of the epiblast. If Pou3f1 expression extended into the posterior half, it was scored as a phenotype.

Statistical analysis was performed with the use of two-tailed Student's unpaired t-test analysis (when the statistical significance of differences between two groups was assessed). Fisher’s exact test was performed for analysis of contingency tables containing the data of the deleted genotypes and expanded phenotypes. Prism software version 7.0 (GraphPad Inc.) was used for representation and statistical analysis. Enriched functional categories in the mouse gene atlas score was calculated using Enrichr (Chen et al., 2013; Kuleshov et al., 2016).

We thank Austin Smith and Ian Chambers for Nanog^−/− ES cell lines; Manuel Serrano and Konrad Hochedlinger for the Nanog^tg mouse line; the CNIC Pluripotent Cell Technology Unit and Elena Lopez-Jimenez for derivation of ES cell lines; the CNIC Genomics Unit for sequencing; Cristina Gutierrez-Vazquez, Rocio Nieto-Arellano and Teresa Rayon for comments on the manuscript; Briane Larui for English editing and Karen Pepper for input in the writing; Jesus Victorino for discussion and members of Manzanares lab for continued support.