Zfp281 is essential for mouse epiblast maturation through transcriptional and epigenetic control of Nodal signaling

Pluripotency is defined by a cell's potential to differentiate into any somatic cell type. How pluripotency is transited during embryo implantation, followed by cell lineage specification and establishment of the basic body plan, is poorly understood. Here we report the transcription factor Zfp281 functions in the exit from naive pluripotency occurring coincident with pre-to-post-implantation mouse embryonic development. By characterizing Zfp281 mutant phenotypes and identifying Zfp281 gene targets and protein partners in developing embryos and cultured pluripotent stem cells, we establish critical roles for Zfp281 in activating components of the Nodal signaling pathway and lineage-specific genes. Mechanistically, Zfp281 cooperates with histone acetylation and methylation complexes at target gene enhancers and promoters to exert transcriptional activation and repression, as well as epigenetic control of epiblast maturation leading up to anterior-posterior axis specification. Our study provides a comprehensive molecular model for understanding pluripotent state progressions in vivo during mammalian embryonic development.


Introduction
Development of an organism from a fertilized egg involves the coordination of cell lineage specification coupled with the establishment of cardinal axes (including the anterior-posterior (A-P) axis), to build a blueprint for the body plan (Arnold and Robertson, 2009). The earliest stages of mammalian development culminate in the formation of a blastocyst comprising three cell lineages: the pluripotent epiblast (Epi) which gives rise to somatic and germ cells, and two extra-embryonic lineages, the primitive endoderm (PrE) and trophectoderm (TE) (Schrode et al., 2013). Mouse embryonic stem cells (ESCs) are derived from, and represent an in vitro self-renewing counterpart of, the so-called 'naive' pluripotent epiblast cells of the blastocyst Boroviak et al., 2014). Pluripotency comprises a continuum of states sequentially encompassing naive, formative and ultimately primed pluripotency (Smith, 2017). Upon blastocyst implantation into the maternal uterus, epiblast cells acquire characteristics of the more developmentally advanced formative and then

Results
Zfp281 is expressed in early mouse embryos and required for early post-implantation development To begin to investigate the role of Zfp281 in vivo during mouse embryonic development when the pluripotent epiblast population is established and matures, we determined the localization of the . Single-cell quantitative immunofluorescence (Lou et al., 2014;Saiz et al., 2016) and single-cell microarray data (Ohnishi et al., 2014)  Despite its critical function in maintaining ESC pluripotency (Wang et al., 2008;Fidalgo et al., 2011) and promoting the transition from the naive-to-primed state of pluripotency in vitro , whether Zfp281 plays a role within the epiblast lineage of the developing embryo remained an open question. A conventional knockout (KO) allele of Zfp281 generated using a gene targeting approach (Fidalgo et al., 2011) was used for embryo analysis. No homozygous mutant mice were recovered at birth from intercrosses of heterozygous animals, demonstrating a requirement for Zfp281 in embryonic development (Figure 1-figure supplement 2A). The analysis of staged embryos revealed that mutants died around E8.0 ( Figure 1-figure supplement 2B). At E5.5, Zfp281KO embryos were recovered at Mendelian ratios and were indistinguishable from their WT and heterozygous littermates by gross morphology (Figure 1-figure supplement 3A). However, by E6.0-6.5, Zfp281 mutants became readily distinguishable from their WT littermates by their smaller size and distinct morphology, exhibiting a thickened VE layer ( Figure 1D, insets, Figure 1figure supplement 3B, insets with yellow bars). These results suggest a requirement of Zfp281 for both the VE and epiblast, which are likely to be non-cell-autonomous and cell-autonomous, respectively, given the epiblast-specific expression of Zfp281.
To confirm the epiblast-specific function of Zfp281, we produced embryos in which Zfp281 was specifically absent in the epiblast, but in which TE and PrE derivatives were wild-type (WT). To do  (B). Immunohistochemistry of Zfp281KO embryo at E5.75 shows that the protein is not expressed, confirming the mutant as a protein null. It also reveals VE-specific background. High-magnification insets (topright) show protein distribution in regions highlighted, white dashed lines delimit the VE layer. At onset of gastrulation (E6.5), Zfp281 is expressed in all epiblast-derived cells. (B) Quantification of nuclear levels of Zfp281 using MINS software at mid-(E3.5) and late (E4.25-4.5) blastocyst stage, revealing protein expression in all three cell types. n = 5 embryos (308 cells) for mid-blastocyst stage and three embryos (448 cells) for late blastocyst stage. See Figure 1-figure supplement 1 for immunohistochemistry of additional stages. (C) Quantification of nuclear levels of Zfp281 in VE and Epi cells in WT and Zfp281KO embryos at E5.75 using Imaris software. n = 6 embryos for the WT (20 cells per genotype) and n = 3 embryos for Zfp281KO. The expression of Zfp281 in the VE was ruled out through quantitative fluorescence level comparisons of wild-type (WT) and Zfp281KO embryos, which lack Zfp281 protein. (D) At E6.5, the A-P axis is established and WT embryos initiate gastrulation at their posterior, while Zfp281KO embryos display a thickened visceral endoderm epithelium (black arrowhead) and no A-P polarity. Insets (top-right) depict thickened VE layer (delineated by black dashed lines) in Zfp281KO embryo compared to the WT. (E) Zfp281KO embryos die around E8.0 and exhibit aberrant gross morphology at E7.75 when compared to WT with either cells of the epiblast layer undergoing apoptosis and/or constriction at the embryonic/extra-embryonic junction (white arrowheads). A = Anterior, p=Posterior, Pr = proximal, D = Distal, BF = brightfield, Scale bars represent 50 mm. Statistical significance was calculated on the average level of corrected fluorescence per embryo using Student T-test. DOI: https://doi.org/10.7554/eLife.33333.002 The following figure supplements are available for figure 1: this we generated tetraploid (4n) WT <->Zfp281KO ESC chimeras and analyzed them at early postimplantation stages (Figure 1-figure supplement 3C). Epiblast-specific loss of Zfp281 produced embryos with a comparable phenotype to that of constitutive gene ablation, with both types of mutant embryos exhibiting a thickened VE (Figure 1-figure supplement 3D). The defect observed in tetraploid chimeras comprising Zfp281KO ESCs could be partially rescued when a Zfp281 cDNA was transfected into Zfp281KO ESCs (referred to as Zfp281KO + Zfp281 cDNA), further confirming the epiblast tissue specificity and Zfp281 gene specificity in the observed mutant phenotypes.

Deregulation of Nodal signaling and A-P axis specification related genes in Zfp281KO embryos
To identify the molecular changes associated with loss of Zfp281, we characterized the transcriptomic differences between WT and Zfp281KO embryos. We performed RNA sequencing (RNA-seq) on individual E6.5 embryos, corresponding to the earliest stage at which mutants could be morphologically distinguished from WT littermates (Figure 2-figure supplement 1). We identified 968 and 792 transcripts that were significantly downregulated and upregulated, respectively, in Zfp281KO versus WT embryos ( Figure 2A, Figure 2-source data 1). Among the significantly downregulated genes were components of the Nodal signaling pathway (e.g.,Nodal,Foxh1,Cripto,Lefty1,Lefty2), and genes regionally-restricted in the epiblast and its derivatives (e.g., Fgf5, Otx2, Gsc) or the VE (e. g., Cer1, Lhx1, Dkk1, Hex, Hesx1) ( Figure 2A, Figure 2-figure supplement 2). We also identified a number of genes that were significantly upregulated in mutant embryos, including Afp, Patched1 (Ptch1), Gsn ( Figure 2A). Gene ontology (GO) analysis for the downregulated genes revealed the top enriched biological process to be A-P axis specification ( Figure 2B). Gene set enrichment analysis (GSEA) revealed that Nodal was the top-ranking signaling pathway enriched in the downregulated genes ( Figure 2C). Of note, this gene set is also included in the GO term A-P axis specification. We next performed RT-qPCR analysis on E6.5 embryos, and confirmed differential expression of genes that are components of the Nodal signaling pathway, Anterior/AVE markers, and posterior/lineage markers, as well as a few genes that were upregulated in Zfp281KO embryos compared to WT embryos ( Figure 2D), consistent with our transcriptome data ( Figure 2A). Together, these data demonstrate that Nodal signaling and A-P axis specification, two key events associated with epiblast maturation, were perturbed in the absence of Zfp281.
Zfp281 controls hallmark molecular events in the exit from naive pluripotency Changes in the expression of some of the stage-specific pluripotency-associated genes ( Figure 2) prompted us to investigate this hallmark molecular event involved in the naive-to-primed transition. Under normal development, the pluripotency factors Sox2 and Oct4 are similarly expressed throughout the naive-to-primed transition. On the other hand, Nanog is rapidly shut down after E4.5 (Chambers et al., 2003) and expressed again from E6.0 onwards in the epiblast (Hart et al., 2004). Nanog is also downregulated in EpiSCs (Silva et al., 2009). Consistent with our RNA-seq data, we noted that Sox2 and Nanog protein levels were unaffected in Zfp281KO embryos ( Figure 3A,D and Additionally, Otx2, which is activated during the naive-to-primed transition and critical for activation of epiblast gene-related enhancers (Buecker et al., 2014), was expressed at significantly reduced levels ( Figure 3C,D) and mislocalized at the distal tip of mutant embryos, rather than being anteriorly restricted as in WT (Figure 3-figure supplement 2B). This downregulation of Otx2 protein expression was already visible at an earlier stage in Zfp281KO embryos and in tetraploid chimeras,    Another hallmark of the naive-to-primed transition is the expression of lineage-specific genes at the posterior part of the embryo, marking the site of the primitive streak and onset of gastrulation. We noted that the T protein is expressed in Zfp281KO embryos ( Figure 3E). However, cells that started expressing T, had not downregulated Nanog (see inset in Figure 3E), indicating a failure in the extinguishment of pluripotency-associated genes as epiblast cells committed to differentiate in Zfp281KO embryos. The domains of both T and Nanog expression in mutant embryos were also proximally radialized, instead of being posteriorly restricted as in WT embryos ( Figure 3E). This radialization of normally posteriorly-localized markers was also observed in tetraploid Zfp281KO ESCs chimeras (Figure 3-figure supplement 3A). Notably, this phenotype was rescued in tetraploid WT <->Zfp281KO + Zfp281 cDNA ESC chimeras ( Figure 3-figure supplement 3A). Analysis of RNA expression by wholemount mRNA in situ hybridization (WISH) of other markers of the primitive streak such as Fgf8, Axin2 and Eomes showed that, similarly to T, they were expressed and proximally radialized in mutant embryos ( Figure 3F and Figure 3-figure supplement 3B,C). Our WISH and RNA-seq data both also revealed that levels of Fgf8 were reduced upon Zfp281 loss. However, T, Axin2 and Eomes RNA levels were not significantly reduced in Zfp281KO embryos ( Figure 2D). Therefore, Zfp281 is required for induction and proper localization of lineage specification markers.
Together, our data establish a requirement for Zfp281 in controlling key molecular events involved in the naive-to-primed transition in embryos, including extinguishment of pluripotency-associated and induction (or localization) of lineage specification transcriptional programs during epiblast maturation.
Zfp281 plays a critical role in Nodal signaling activation to promote DVE/AVE migration Execution of the naive-to-primed transition in the epiblast of early post-implantation embryos culminates in A-P axis formation, determined via cross-talk between the epiblast and VE through Nodal and Wnt signals (Kiecker et al., 2016). The radialized expression of Nanog and several lineage specification factors in Zfp281KO embryos ( Figure 3E , suggesting that the DVE/AVE population was not specified, or could not be maintained or migrate. To distinguish these possibilities, we employed a Hex-GFP reporter line  to visualize the AVE in Zfp281KO embryos. In agreement with our WISH data, Hex-GFP expression was reduced and distally-localized, consistent with a failure in maintenance leading to impaired migration of the DVE/AVE population ( Figure 4B and   Nodal and Wnt signaling pathways play a key role in DVE/AVE specification and migration (Kiecker et al., 2016), and rank as top enriched GO terms in the downregulated genes from our RNA-seq data ( Figure 2C). We already showed that levels of Axin2 and T, two Wnt signaling targets, were unaffected by loss of Zfp281, although their expression was radialized due to failure in A-P axis specification ( Figures 2D and 3F and Figure 3-figure supplement 3B). By contrast, the Nodal target and negative regulator Lefty2 (expressed in the posterior epiblast of WT embryos) were absent in the epiblast of Zfp281KO embryos ( Figure 4C), similarly to the related Lefty1 gene, which is expressed earlier in the VE, and also absent in mutants ( Figure 4A). Nodal and Eomes are usually posteriorly localized in the epiblast at the site of the primitive streak (Nowotschin et al., 2013). However, their expression was radialized in Zfp281KO embryos ( Figure 4C and Together, our data suggest an impairment of Nodal signaling in Zfp281KO embryos, leading to defects in DVE/AVE specification and migration, further corroborating a critical role for Zfp281 in promoting epiblast maturation.

Zfp281 regulates lineage-specific genes for transcriptional activation during epiblast maturation
To understand the molecular mechanisms by which Zfp281 promotes epiblast maturation, we turned to the in vitro naive-to-primed transition model to identify Zfp281-regulated genes. We first examined expression of Zfp281-regulated genes in WT and Zfp281KO ESCs and epiblast-like cells (EpiLCs), which is an alternative of primed cells because Zfp281KO is detrimental to the self-renewal of EpiSCs (Tsakiridis et al., 2014;Fidalgo et al., 2016). EpiLCs are derived by short-time (48 hr) adaption of ESCs to the primed cell culture condition, representing an intermediate state, known as the formative state Smith, 2017), in the naive-to-primed transition (Hayashi et al., 2011;Buecker et al., 2014). As expected, EpiLCs (WT) expressed higher levels of lineage-specific genes than ESCs ( Figure 5A). However, levels of expression of these genes were reduced or abrogated in Zfp281KO relative to WT EpiLCs ( Figure 5A), indicating Zfp281 is important for establishing or maintaining expression of these genes during the transition. In addition, while genes involved in Nodal signaling were similarly expressed in both WT ESCs and EpiLCs, their levels of expression were downregulated in Zfp281KO versus WT EpiLCs (Figure 5-figure supplement 1), consistent with their downregulation in Zfp281KO versus WT embryos (Figure 2A,D).
To investigate how Zfp281 exerts transcriptional control of downstream target genes, we performed chromatin immunoprecipitation (ChIP) with massively parallel sequencing (ChIP-seq) in both WT ESCs and EpiSCs, which revealed enrichment of Zfp281 binding at regions near gene transcription start sites (TSSs) and enhancers ( Figure 5-figure supplement 2A,B), suggesting that Zfp281 is actively involved in transcriptional regulation in both naive and primed pluripotent states. There were 9358 common peaks for Zfp281, and 11,408 and 3467 peaks specific to ESCs and EpiSCs, respectively, which were lost and gained during the transition between these two states ( Figure 5figure supplement 2C,D), suggesting Zfp281 targets may be dynamically regulated during pluripotent state transition.
Next, we determined whether Zfp281 coordinately controls transcriptional programs associated with pluripotent state transition and lineage commitment through binding of regulatory regions of its target genes. Surprisingly, promoters (Pro) of lineage-specific genes such as T, Otx2, Eomes, Gsc were bound by Zfp281 with a higher enrichment in ESCs than in EpiSCs ( Figure 5B). ChIP-qPCR analyses confirmed that Zfp281 binding intensities were reduced at these promoters in EpiSCs ( Figure 5C). Consistent with the fact that these lineage-specific genes will be activated in primed  cells, we also observed diminished intensities of repressive histone mark H3K27me3 on their promoters ( Figure 5D). By contrast, there was no Zfp281 peak at promoters of Fgf5 or Fgf8, indicating a promoter-independent regulation of Zfp281 on these two genes ( Figure 5E). However, Zfp281 bound at promoter-distal regions of Fgf5 and Fgf8 ( Figure 5E-F) accompanied with increased H3K27ac ( Figure 5G) in EpiSCs, suggesting Zfp281 may be involved in enhancer activation on these two targets during the naive-to-primed transition. Indeed, a previous study has shown that the two Zfp281 peaks (P1, P2) comprising enhancers of Fgf5 are critical for the naive-to-primed transition (Buecker et al., 2014). Taken together, our data indicate that Zfp281 regulates lineage-specific genes during epiblast maturation through both promoter-and enhancer-related mechanisms.

Zfp281 is associated with chromatin modifiers for promoter activation of lineage-specific genes
To further understand how Zfp281 controls transcription of target genes in relation to their promoter chromatin architecture, we investigated the genome-wide association of Zfp281 with other epigenetic regulators and TFs in ESCs. Hierarchical clustering analysis for ChIP-seq association revealed that Zfp281 and Ep400 have the most similar binding patterns. Furthermore, Zfp281 and Ep400 also show similar binding patterns with the Polycomb repressive complex 2 (PRC2) components Ezh2 and Suz12 ( Figure 6A). Ep400 is a component of the Tip60-Ep400 histone acetyltransferase complex that is necessary to maintain ESC self-renewal (Fazzio et al., 2008;Chen et al., 2015). A previous study showed that Ep400 localization to promoters depends on H3K4me3, and Ep400 promotes histone H4 acetylation at both active and silent target promoters in ESCs (Fazzio et al., 2008). We profiled ChIP-seq intensities of Ep400, Suz12 (a component of PRC2 that modifies H3K27me3), Mbd3 and histone marks H3K4me3, H3K27ac, and H3K27me3 at Zfp281 peak regions ( Figure 6B). Mbd3 is a core component of NuRD histone deacetylation complex that can be recruited by Zfp281 for repression of pluripotency genes (Fidalgo et al., 2012). Zfp281 and Ep400 peaks exhibited a high correlation across the genome, but were mutually exclusive with Suz12 and Mbd3 peaks, dividing Zfp281 peaks into two classes: (I) Zfp281/Ep400/Suz12-cobound, and (II) Zfp281/Ep400/Mbd3cobound ( Figure 6B). Class I and Class II characterize the epigenetic features of target genes in ESCs, and are highly enriched for GO terms that signify development and pluripotency, respectively ( Figure 6-figure supplement 1). The Class I genes are bivalent with a feature of H3K4me3 and H3K27me3 co-enrichment ( Figure 6B). Bivalent genes remain silent in ESCs while undergoing fast activation in response to differentiation signals (Bernstein et al., 2006). Our study demonstrates that Zfp281 regulates active and bivalent genes by associating with identical (Ep400 for both Class I and Class II) or distinct (PRC2 for Class I versus NuRD for Class II) epigenetic regulators.
We next asked whether Zfp281 associates with these epigenetic modifiers during the naive-toprimed transition. While we confirmed interactions of the PRC2 (Suz12) and NuRD (Chd4, Mbd3) complexes with Zfp281 in ESCs, we found their associations with Zfp281 were reduced in EpiSCs despite their expression levels being similar in the two populations ( Figure 6C). By contrast, Zfp281's interaction with components of the Tip60-Ep400 complex (Ep400, Trrap) was maintained in both ESCs and EpiSCs ( Figure 6C). Loss of interaction between Zfp281 and the PRC2 complex may be responsible for the activation of Zfp281 target genes during the transition, providing a parsimonious explanation for the upregulation of lineage specification genes ( Figure 5B-C) and reduction of H3K27me3 ( Figure 5D) in EpiSCs versus ESCs. The lack of association between Zfp281 and NuRD in EpiSCs may also explain why pluripotency genes including Sox2 and Nanog are not regulated by Zfp281 in the epiblast ( Figure 3A,D), which is in contrast with NuRD-mediated Nanog repression in ESCs due to their physical association (Fidalgo et al., 2012). We confirmed the reduced chromatin occupancy of Suz12 at bivalent promoters of T and Eomes in Zfp281KO relative to WT ESCs ( Figure 6D-E), indicating a Zfp281-dependent recruitment of PRC2 at the bivalent promoters. However, we cannot exclude the possibility that the residual binding of Zfp281 to bivalent promoters (e. g., T and Eomes in Figure 5C) with maintained Tip60-Ep400 complex association during the naiveto-primed transition may further reinforce the activation of lineage-specific genes through the presumed role of Tip60-Ep400 in outcompeting PRC2, and thus downregulating the promoter H3K27me3 levels (Chen et al., 2015). Taken together, our data reveal an important role of Zfp281 in regulating bivalent promoters during the naive-to-primed pluripotency transition. Reduced association between Zfp281 and the PRC2 complex, but preservation of the Zfp281-Ep400 association, at bivalent promoters results in decreased H3K27me3 during this transition, leading to transcriptional activation of lineage-specific genes concomitant with epiblast maturation.

Zfp281 cooperates with Oct4 and P300 for regulation of Nodal signaling components in epiblast maturation
We have shown that Zfp281 is necessary for activation of Nodal signaling components in the embryo (Figure 2 and Figure 4). Next, we investigated activity of Nodal signaling by examining Smad2 phosphorylation (p-Smad2) in WT and Zfp281KO ESCs. P-Smad2 is significantly reduced in Zfp281KO ESCs compared to that in WT ESCs ( Figure 7A). Since Zfp281KO EpiSCs cannot be maintained in long-term culture , we performed shRNA-mediated knockdown (KD) of Zfp281 in EpiSCs. P-Smad2 is not affected by Zfp281KD in EpiSCs, probably because of the Activin in culture constitutively activating p-Smad2. However, protein expression of the Nodal signaling target Lefty significantly decreased by Zfp281KD (Figure 7-figure supplement 1A), suggesting that Zfp281 may directly regulate Lefty expression. Similarly, reduction of Lefty protein expression is reproduced by treatment of ALK receptor inhibitor (ALKi) that specifically blocks p-Smad2 and Nodal signaling (Figure 7-figure supplement 1B). In addition, as activation of WNT and Nodal signaling pathways can differentiate ESCs to primitive-streak (PS)-like cells (Mulas et al., 2017), we also investigated the role of Zfp281 in this differentiation. ESCs were treated with Activin (a Nodal ligand) and CHIR (a GSK3b inhibitor to activate WNT pathway), and with ALKi or Zfp281 shRNAs. The morphology of Zfp281KD cells revealed a strong phenotype of differentiation resistance, which is similar to that of ALKi treatment. Dome-shaped ESC-like colonies persisted in Zfp281KD cell cultures after Activin/CHIR treatment, a striking difference compared to WT control cells (Figure 7-figure supplement 2A). RT-qPCR for up to 3 days after treatment indicated both Zfp281KDand ALKi-treated cells exhibited decreased expression of the PS marker genes T and Lefty2 (a Nodal signaling target gene also expressed in PS) (Figure 7-figure supplement 2B,C).
To understand the molecular regulation of the Nodal signaling pathway by Zfp281, we employed ChIP-seq analysis and revealed that Zfp281 localizes to distal regions of Nodal and Lefty2, as well as promoters of Foxh1 ( Figure 7C). Previous studies indicated that transcription factors Oct4, Otx2, and histone acetyltransferase P300 (the writer of H3K27ac) associate with enhancer reorganization in the naive-to-primed transition (Buecker et al., 2014;Yang et al., 2014). We first confirmed that Zfp281 interacts with Oct4 and P300 in EpiSCs ( Figure 7B). However, we did not detect the Zfp281-Otx2 interaction in EpiSCs (data not shown). Furthermore, ChIP-seq analysis indicates that Zfp281 Source data 1. Accession numbers of ChIP-seq data used in Figure 6A. Figure 6 continued on next page co-localizes with Oct4 and P300 (Buecker et al., 2014) in almost all Zfp281 peaks ( Figure 7C, Figure 7-figure supplement 3), suggesting that Zfp281 may be involved in widespread relocation of Oct4 and P300, including both promoters and enhancers, in the naive-to-primed transition, that is independent of Otx2 association.
Zfp281 localizes to characterized enhancers within the Nodal locus (PEE, NDE, HBE, but not ASE, nomenclature from ) ( Figure 7C). ChIP-qPCR analysis confirmed that Zfp281 intensities are comparable between PEE and NDE, but decreased at HBE in EpiSCs compared to ESCs ( Figure 7D). Zfp281 intensity by ChIP-qPCR was relatively low at ASE, consistent with our ChIP-seq data ( Figure 7C-D). It is reported that, during the naive-to-primed transition, Nodal enhancer activity relocates from the HBE to the ASE . Indeed, H3K27ac intensity was also reduced at HBE (Figure 7E), consistent with reduced Zfp281 binding at this locus during the naive-to-primed transition ( Figure 7D). We also evaluated Zfp281 intensities by ChIP-qPCR at an enhancer of Lefty2 and the promoter of Foxh1, and found high intensities of Zfp281 and H3K27ac, colocalizing with P300 and Oct4 peaks at these loci in ESCs and EpiSCs ( Figure 7C-E). Zfp281 binding at regulatory regions of these Nodal signaling-related genes is biologically important, as expression of Lefty2 is downregulated by Zfp281KD in EpiSCs (Figure 7-figure supplement 1), and many of Nodal signaling genes were perturbed in Zfp281KO EpiLCs relative to their WT counterparts ( Figure 5-figure supplement 1), supporting a critical role of Zfp281 in transcriptional activation of Nodal signaling during the transition to primed pluripotency in vitro. To provide direct evidence that Zfp281 binds and regulates these Nodal signaling related genes in vivo, we performed ChIP experiments on E6.5 WT embryos ( Figure 7F). Zfp281 exhibited high chromatin-binding activity at the HBE enhancer of the Nodal locus, as well as the regulatory regions within the Lefty2 and Foxh1 loci ( Figure 7G), which would be abrogated in Zfp281 mutant embryos. Together, our data demonstrate that Zfp281, together with P300 and Oct4, are important for regulation of components of the Nodal signaling pathway during the maturation of the epiblast of the mouse embryo.

Discussion
Pluripotency is a continuum where the naive and primed states represent the initial and final stages, corresponding to the establishment of pluripotency in the pre-implantation blastocyst and the exit from pluripotency as cells of the post-implantation epiblast initiate gastrulation. While the in vitro pluripotent state transition model provides a useful tool to infer how the epiblast transitions between states and prepares for germ layer differentiation (Weinberger et al., 2016), it cannot substitute for direct investigations into the transition occurring during epiblast maturation in vivo in embryos. Notably, none of the pluripotency-associated factors, for which mouse mutants are available, have described phenotypes directly associated with the exit from naive pluripotency in the embryo. Zfp281 is the first factor demonstrated to be critical in vivo for this developmental transition; its loss results in embryonic lethality due to a failure in epiblast maturation. Our data reveal that Zfp281 is concomitantly required for the initiation of expression of genes encoding Nodal signaling components and the lineage specification program during epiblast maturation. Zfp281 coordinates crosstalk among multiple epigenetic pathways through physical association with histone methylation (PRC2) and acetylation (Ep400, P300) complexes within epiblast cells. Crosstalk converges on the coordinated regulation of bivalent promoters and reorganization of enhancers leading to activation of lineage-specific genes and downregulation of pluripotency genes during the naive-to-primed transition, arguing for a master regulator status of Zfp281 in controlling key molecular events leading to epiblast maturation (Figure 8, top panel). Accordingly, loss of Zfp281 results in a series of developmental defects within the epiblast, which non-cell-autonomously lead to a failure to establish or maintain the DVE/AVE resulting in A-P axis specification defects (Figure 8, bottom panel).  The in vitro pluripotent stem cell models overcome the limitation of embryo material and so have been instrumental in dissecting molecular events involved in the naive-to-primed transition (Buecker et al., 2014;Factor et al., 2014) as well as the regulation of the Nodal signaling pathway during epiblast maturation . Our combined in vitro and in vivo studies have demonstrated high consistency of the regulatory functions of Zfp281 on the lineage-specific genes (Figure 2 and Figure 5). However, there are also some differences between our findings in embryos and those made in pluripotent stem cell models. For instance, embryo RNA-seq data indicated non-significant changes in expression of lineage markers T and Eomes between WT and Zfp281KO embryos ( Figure 2D), while their expression was markedly downregulated in Zfp281KO EpiLCs ( Figure 5A). This disparity could be due to EpiLC representing a pluripotent state that does not match with the embryonic stage examined for RNA-seq in embryos, and/or extra-embryonic expression, in the case of Eomes in vivo (Nowotschin et al., 2013) (Figure 3-figure supplement 3B), that is not represented in EpiLC culture. Similarly, Zfp281KO embryonic phenotypes described for the VE layer cannot be reproduced in vitro since there is no equivalent cell population in ESC/EpiSC cultures. We noted that the morphology of VE cells in Zfp281KO embryos was different from the WT and resembled their adjacent extra-embryonic VE cells (Figure 1-figure supplement 3). As we identified that Afp, a pan-VE marker whose downregulation is coincident with the onset of gastrulation (Viotti et al., 2014), is upregulated in our mutant embryos (Figure 2A,D), we speculate that Zfp281KO embryos may also fail in VE maturation.
Furthermore, we also observed that Nodal signaling was significantly reduced in Zfp281KO embryos ( Figure 2D), whereas in EpiSCs, p-Smad2 level was not affected by Zfp281KD (Figure 7figure supplement 1). We speculate that this discrepancy may be attributable to the presence of Activin in the primed cell culture medium, which constitutively activates Nodal signaling. Compared to the in vitro system, crosstalk between the in vivo epiblast and its adjacent extra-embryonic tissues is dynamic and context-dependent, and unable to be precisely captured in a single defined cell culture system. Moreover, mRNA and protein levels of Oct4 and Otx2, the two factors required for reorganization of the enhancer landscape during the naive-to-primed transition (Buecker et al., 2014), are significantly downregulated in Zfp281KO embryos ( Figure 2D and Figure 3B-D), which may also explain why Nodal signaling fail to be activated in epiblast maturation (Figure 7). Taken together, our studies highlight the importance of direct in vivo functional investigations using mouse models to refine and/or authenticate in vitro findings of pluripotent state transitions for a better understanding of epiblast maturation.
Nodal signaling has been extensively studied in key sequential events ranging from epiblast maturation to left-right patterning (Shen, 2007). In Nodal mutant embryos, epiblast cells do not mature properly and embryos exhibit a failure in DVE/AVE establishment (Brennan et al., 2001;Mesnard et al., 2006). A recent study suggests that Nodal guides the transition from naive to formative pluripotency in vitro (Mulas et al., 2017). Although pluripotency factors have been reported to occupy and regulate enhancers at the Nodal locus during the ESC-to-EpiSC transition  , how they contribute to lineage specification during the exit from a naive pluripotent state in vivo has remained an open question. Our work reveals that Zfp281 acts as a master regulator of epiblast maturation through the regulation of primed pluripotency genes and Nodal signaling components. Zfp281 binds the PEE, NDE, and HBE Nodal enhancers in vitro and in vivo, with binding at the HBE decreasing during epiblast maturation. In ESCs, HBE is bound by Oct4/ Sox2/Nanog/Klf4 , as well as Zfp281, however neither Zfp281 nor Oct4 binds ASE ( Figure 7C), suggesting that loss of Zfp281 binding at HBE may be a prerequisite for decommissioning the naive-specific Nodal enhancer in transitioning to the primed state. In addition, Nodal expression is reduced or not restricted proximally in Zfp281KO embryos, which could further indicate a failure in enhancer switching. Our findings are thus in agreement with the proposed switch of Nodal enhancers, from HBE to ASE, concomitant with the transition from ESCs to EpiSCs . Together with the crosstalk between Zfp281 and mulitiple histone modifying complexes (PRC2/Ep400/NuRD) (Figure 6), our data have thus uncovered a critical role for the pluripotency factor Zfp281 in coordinating transcriptional and epigenetic control of the exit from naive pluripotency through promoter and enhancer remodeling, leading to activation of lineage-specific genes and genes encoding Nodal signaling components while embryos are progressing to a primed pluripotency state during epiblast maturation ( Figure 8).
As the targets and antagonists of Nodal signaling in embryo development, Lefty genes (Lefty1 and Lefty2) are sensitive to status of DNA methylation at promoters. It is reported that ESCs depleted of Tet1 diminished expression of Lefty1 with DNA hyper-methylation at Lefty1 promoter (Koh et al., 2011). Our previous work indicated that Zfp281 physically interacts with Tet1 , therefore Zfp281 may recruit Tet1 at Lefty1/2 promotors and maintain Lefty expression, suggesting a Nodal signaling-independent role of Zfp281. Although, developmental arrest of Zfp281 mutant embryos (~E6.5) precedes that of Tet-TKO (triple KO) mutants (~E7.5), and defects observed in Tet-TKO mutants were attributed to downregulation of Lefty1/2 through promoter hyper-methylation, leading to constitutive activation of the Nodal signaling pathway due to the disruption of Lefty-mediated negative-feedback (Dai et al., 2016). Therefore, these results raise the question of a possible link between Zfp281 and DNA modification for epigenetic control of proper embryonic development, which is worthy of future investigation.

Mouse strains, embryo collection and staining
Mouse strains used in the study were Zfp281KO (Fidalgo et al., 2011) and Hex-GFP . All mice used in this study were maintained in accordance with the guidelines of the Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Animal Care and Use Committee (IACUC) under protocol number 03-12-017 (PI Hadjantonakis).

Tetraploid WT <-> ESCs chimera
Tetraploid chimeras were generated according to standard protocols (Eakin and Hadjantonakis, 2006;Eggan et al., 2001). Briefly, females were superovulated by i.p. injection of pregnant mares' serum and human chorionic gonadotropin (HCG), and were then mated with males. Fertilized zygotes were isolated from oviducts 24 hr later, cultured until they reached the 2-cell stage, at which point they were electrofused. Fused 1-cell embryos were carefully identified, cultured for another 2 days, and then injected with about fifteen ESCs. ESCs from two Zfp281KO clones 2.6 (XX) and 3.34 (XY), and a Zfp281 cDNA rescued 3.34 clone (XY) were used for tetraploid injection. About 40~60 injected blastocysts were collected for each ES clone, and transferred into 2~3 pseudo-pregnant foster females. All experiments involving 2C-embryo electrofusion and tetraploid injection were performed at the Rodent Genetic Engineering Laboratory at New York University, with dissections of resulting post-implantation embryo chimeras taking place at MSKCC.

Knockdown of Zfp281, ALK inhibitor (ALKi) treatment and western blot analysis
Two shRNAs for Zfp281 knockdown were previously validated in our study . Lentivirus production and infection were performed as described (Ivanova et al., 2006). Concentrated viral supernatants were incubated with ESCs or EpiSCs for 1 hr, then cells were diluted with fresh medium. Blasticidin (10 mg/mL) was used for selection 24 hr later, and cells were harvested 72 hr after viral infection.

Induction of primitive streak (PS)-like cells
ESCs were seeded on fibronectin-coated plates, and treated with Activin A (20 ng/mL) and CHIR99021 (3 mM) for up to 3 days for PS-like cells inductions. Cells were also treated with ALK inhibitor (1 mM) or Zfp281 shRNAs to investigate the effects of Nodal signaling inhibition or Zfp281 knockdown during this differentiation. RNA was collected at different days after treatment.

RT-qPCR
Total RNA was extracted from ESCs or EpiSCs using the RNeasy kit (Qiagen), and from E6.5 embryos using the TRIZOL reagent (Ambion, #15596018). Reverse transcription was performed and cDNA was generated using the qScript kit (Quanta, Cat# 95048). Relative expression levels were determined using a LightCycler 480 SYBR green mix (Roche, 4729749001). qRT-PCR experiments were performed on a LightCycler Real Time PCR System (Roche). Gene expression levels were normalized to Gapdh. Error bars indicate standard error for average expression of two technical replicates. Primers for qPCR are listed in Figure 2-source data 2.

RNA-seq analysis of embryos
Mouse embryos were dissected at E6.5 and total RNAs were extracted using TRIZOL reagent (Ambion, #15596018) following standard protocols. RNA quality was evaluated with an Agilent 2100 BioAnalyzer system, and embryo genotype was determined by morphology and confirmed by expression of Zfp281 by RT-qPCR. Ten to one-hundred ng total RNA from each embryo was processed for RNA-seq library construction using the Ovation Mouse RNA-seq kit (NuGEN, #0348-32) following the manufacturer's protocol, then massively parallel sequencing was performed on an Illumina HiSeq 4000 Sequencing System. Between 20 and 50 million 100 bp single-end reads were obtained per sample.
RNA-seq reads were aligned to the genome using TopHat (v2.0.10) and Bowtie2 (v2.1.0) with the default parameter settings. UCSC mm9 mouse genome, as well as the transcript annotation, was downloaded from the iGenomes site. Transcript assembly and differential expression analyses were performed using Cufflinks (v2.1.1). Assembly of novel transcripts was not allowed (-G), other parameters of Cufflinks followed the default setting. The summed RPKM (reads per kilobase per million mapped reads) of transcripts sharing each gene_id were calculated and exported by the Cuffdiff