Culture of haploid blastocysts in FGF4 favors the derivation of epiblast stem cells with a primed epigenetic and transcriptional landscape

Pluripotent stem cells within the inner cell mass and epiblast of mammalian embryos have the capacity to form all lineages in the adult organism, while multipotent trophoblast stem (TS) cells derived from the trophectoderm are capable of differentiating into fetal lineages of the placenta. While mouse embryonic stem (ES) cells and epiblast stem cells (EpiSCs) exhibit distinct expression patterns and utilize distinct external signaling pathways for self-renewal, because mouse EpiSCs resemble human ES cells they are a useful model to investigate mechanisms of human ES cell self-renewal and differentiation. Recent studies have shown that haploid embryos and ES cells can be generated from chemically-activated unfertilized mouse oocytes. However, it is unclear whether EpiSCs or TS cells can be derived from haploid embryos. Here, we describe the derivation of EpiSCs from haploid blastocyst-stage embryos using culture conditions that promote TS cell self-renewal. Maternal (parthenogenetic/gynogenetic) EpiSCs (maEpiSCs) functionally and morphologically resemble conventional EpiSCs. Established maEpiSCs and conventional EpiSCs are diploid and exhibit a normal number of chromosomes. Moreover, global expression analyses and epigenomic profiling revealed that maEpiSCs and conventional EpiSCs exhibit similarly primed transcriptional programs and epigenetic profiles, respectively. Altogether, our results describe a useful experimental model to generate EpiSCs from haploid embryos, provide insight into self-renewal mechanisms of EpiSCs, and suggest that FGF4 is not sufficient to derive TS cells from haploid blastocyst-stage embryos.

maEpiSCs have a similar gene expression profile compared to conventional EpiSCs. To evaluate whether maEpiSCs exhibit an EpiSC-like transcriptional profile we utilized RNA-Seq. The transcriptome of a representative maEpiSC cell line was interrogated relative to conventional EpiSCs, ES cells, and embryoid body (EB) differentiated (day 10 and day 14) ES cells, mouse embryonic fibroblasts (MEFs), and TS cells. Like EpiSCs, ES cells also express elevated levels of OCT4, and are pluripotent. ES cells represent another control for a pluripotent cell type, while EBs and MEFs represent differentiated cells. Hierarchical clustering (HCA) and k-means clustering were used to identify variations in patterns of gene expression. These results showed that maEpiSCs and conventional EpiSCs clustered closer together relative to ES cells, TS cells, and differentiated EBs and MEF cells (Fig. 1K). Moreover, the expression profile of maEpiSCs was more similar to conventional EpiSCs (E3) relative to ES cells, TS cells, or differentiated cells (EBs and MEFs) (Fig. 1L). Custom RNA-Seq UCSC genome browser views showed that maEpiSCs and conventional EpiSCs express the pluripotency-factors Nanog, Pou5f1, and the EpiSC marker Fgf5 (Fig. 1M), while MEFs and differentiated EBs (day 10 and day 14) did not express these genes.
Comparable epigenetic landscape of maEpiSCs and conventional EpiSCs. Next, we investigated global profiles of histone modifications in maEpiSCs using ChIP-Seq and previously described methods [14][15][16][17] . By comparing maEpiSC histone modification profiles with histone modification patterns in conventional EpiSCs, ES cells, and MEFs, using 2 kb genomic bins, we found that genome-wide histone modification profiles of maEpiSCs are similar to conventional EpiSCs (E3) ( Fig. 2A). Furthermore, boxplots reveal that H3K4me3 levels at conventional EpiSC peaks are overall similar between conventional EpiSCs and maEpiSCs (Fig. 2B, top left), while MEFs (GSE21271) and ESCs (GSE53087) displayed decreased levels at conventional EpiSC ChIP-Seq peaks (SICER-defined peaks, see methods). Moreover, while levels of H3K4me3 at H3K4me3/H3K27me3 bivalent chromatin regions in EpiSCs were similar between conventional EpiSCs, maEpiSCs, and ES cells (Fig. 2B, bottom left), levels of H3K27me3 at EpiSC bivalent peaks were lower in maEpiSCs relative to conventional EpiSCs. However, levels of H3K27me3 at EpiSC peaks were lower in maEpiSCs relative to conventional EpiSCs (Fig. 2B, top right).
While scatter plots demonstrated that maEpiSC H3K4me3 profiles were more similar to conventional EpiSCs relative to ES cells or MEFs (Fig. 2C), a comparison of conventional EpiSCs and maEpiSCs demonstrated that levels of H3K4me3 were lower in maEpiSCs at a subset of regions. PCA further confirmed that maEpiSCs are more similar to conventional EpiSCs (E3) relative to ES cells or MEFs (Fig. 2D). PCA revealed that maEpiSCs and conventional EpiSCs clustered closer to one another within the two-dimensional space, and along the PC1 and PC2 axes, relative to ES cells or MEFs. Heat maps also showed that conventional EpiSCs and maEpiSCs exhibited similar global profiles of H3K4me3 (Fig. 2E, top), where H3K4me3 levels were higher in maEpiSCs and conventional EpiSCs relative to ESCs and MEFs, at genes whose expression is elevated in EpiSCs. Moreover, while conventional EpiSCs, maEpiSCs, and ES cells exhibited relatively similar distributions of H3K27me3 (Fig. 2E, bottom), MEFs exhibited higher levels of H3K27me3 at genes whose expression is elevated in EpiSCs (Fig. 2E, bottom). In addition, our ChIP-Seq results show that 83% of regions containing H3K4me3-marks in conventional EpiSCs were also marked by H3K4me3 in maEpiSCs, whereas 46% of regions containing H3K4me3-marks in ES cells were also occupied by H3K4me3 in maEpiSCs (Fig. 2F), and 69% of regions containing H3K4me3-marks in MEFs were also occupied by H3K4me3 in maEpiSCs (Fig. 2F), suggesting that H3K4me3 levels in maEpiSCs are more similar to conventional EpiSCs than ES cells or MEFs. Also, 69% of regions containing H3K27me3 in conventional EpiSCs were also occupied by H3K27me3 in maEpiSCs (Fig. 2F, bottom). In contrast, 43% and 45% of regions containing H3K27me3 in MEFs and ES cells, respectively, were also occupied by H3K27me3 in maEpiSCs (Fig. 2F, bottom). These findings suggest that H3K27me3 profiles in maEpiSCs resemble conventional EpiSCs more than ES cells or MEFs.
Conventional EpiSCs and maEpiSCs both had elevated H3K4me3 levels at Fgf5, whereas ES cells and MEFs had low H3K4me3 levels at Fgf5, a gene whose expression is enriched in EpiSCs (Fig. 2G, top left), further showing that maEpiSCs epigenetically resemble conventional EpiSCs compared to MEFs or ES cells. Moreover, we observed H3K4me3 enrichment at pluripotency-related genes, including Nanog and Pou5f1, in UCSC genome browser views (Fig. 2G, left). Moreover, H3K4me3 levels were low at the ESC-enriched gene, Pecam1, in conventional EpiSCs and maEpiSCs (Fig. 2G, top right). Furthermore, maEpiSCs display H3K4me3/H3K27me3 co-occupancy at a custom UCSC browser view of a bivalently marked gene (Fig. 2G, bottom right). Overall, these findings show that maEpiSCs and conventional EpiSCs exhibit similar epigenetic profiles.

Similar profile of H3K27ac-marked enhancers in maEpiSCs and conventional EpiSCs.
Cis-regulatory elements determine lineage-specific transcriptional responses to external stimuli [18][19][20] . To evaluate enhancer profiles in maEpiSCs, conventional EpiSCs, ES cells, and MEFs we performed H3K27ac ChIP-Seq, which marks active enhancers 21,22 . We observed similar levels of H3K27ac between conventional EpiSCs and maEpiSCs at EpiSC-peaks (Fig. 3A), while MEFs (GSE29218) displayed lower levels, and ES cells exhibited higher levels, demonstrating that H3K27ac levels at active enhancer regions in maEpiSCs are more similar to conventional EpiSCs relative to ES cells or MEFs. Furthermore, a comparison of maEpiSCs and conventional EpiSCs showed that levels of H3K27ac in intergenic levels are similar (Fig. 3B, left panel). However, a subset of enhancer regions occupied by H3K27ac exhibited elevated levels in conventional EpiSCs, suggesting that a subset of conventional EpiSC-enhancers may have been deactivated in maEpiSCs. In contrast, levels of H3K27ac in intergenic regions were less similar between maEpiSCs and MEFs or ES cells (Fig. 3B, right).
Recent studies have shown that mediator and master transcription factors co-occupy enhancer clusters, also known as 'super-enhancers' , which have been used to define cellular identity 23,24 . To evaluate the activity of ESC-defined super-enhancers in maEpiSCs, we surveyed H3K27ac levels. These results show that levels of H3K27ac are lower in maEpiSCs relative to ES cells (Fig. 3C), but similar to conventional EpiSCs. These findings likely reflect differential marking of super-enhancers between ES cells and EpiSCs. Also, H3K27ac levels were higher in maEpiSCs relative to MEFs at ESC-defined super-enhancers ( Fig. 3C), showing that the activity of ES cell super-enhancers is higher in maEpiSCs relative to MEFs. PCA analysis confirmed that H3K27ac levels at conventional EpiSC peaks were similar between maEpiSCs and conventional EpiSCs relative to ES cells or MEFs (Fig. 3D), where maEpiSCs and conventional EpiSCs clustered closer to one another within the two-dimensional space relative to ES cells or MEFs. In addition, heat maps showed that conventional EpiSCs and maEpiSCs exhibit relatively similar global profiles of H3K27ac (Fig. 3E). Moreover, our results show that 73% of H3K27ac-marked regions in conventional EpiSCs were also marked by H3K27ac in maEpiSCs, whereas 49% of H3K27ac-marked regions in ESCs were occupied by H3K27ac in maEpiSCs (Fig. 3F), and 44% of H3K27ac-occupied regions in MEFs were also occupied by H3K27ac in maEpiSCs (Fig. 3F), suggesting that H3K27ac levels in maEpiSCs are more similar to conventional EpiSCs than MEFs or ESCs.
To further investigate the enhancer landscape of maEpiSC we evaluated H3K4me2 levels, which are also enriched at enhancer regions 25 . Similar to H3K27ac, H3K4me2 levels are correlated with enhancer activity in a cell-type specific manner. Our findings show that 77% of H3K4me2-occupied regions in conventional EpiSCs were also marked by H3K4me2 in maEpiSCs, whereas 44% of H3K4me2-marked regions in ESCs were co-occupied by H3K4me2 in maEpiSCs (Fig. 3G), and 41% of H3K27ac-occupied regions in MEFs were also occupied by H3K27ac in maEpiSCs (Fig. 3G), suggesting that H3K27ac levels in maEpiSCs are more similar to conventional EpiSCs than MEFs or ESCs. In addition, while ChIP-Seq demonstrated that H3K4me2 levels are similar between maEpiSCs and conventional EpiSCs (Fig. 3H), H3K4me2 levels at EpiSC-occupied enhancers were lower in ES cells (GSE53087) and MEFs (GSE36292) 26 , demonstrating that maEpiSCs cells are more similar to conventional EpiSCs than MEFs or ESCs. In addition, levels of H3K4me2 were similar between maEpiSCs and conventional EpiSCs (Fig. 3I, left). However, levels of H3K4me2 were less similar between maEpiSCs and ESCs or MEFs (Fig. 3I, right, bottom), further suggesting that the epigenetic profile of maEpiSCs is more similar to conventional EpiSCs than MEFs or ESCs. Likewise, levels of H3K4me2 at ES cell defined super-enhancers were similar between maEpiSCs and conventional EpiSCs (Fig. 3J). However, H3K4me2 levels were lower in maEpiSCs/ RNA-Seq data. Clustering was performed according to k-means on differentially expressed (>two-fold) genes. conventional EpiSCs relative to ES cells, but higher relative to MEFs at ESC-defined super enhancers (Fig. 3J). PCA analysis confirmed that H3K4me2 levels were similar between maEpiSCs and conventional EpiSCs relative to ES cells or MEFs (Fig. 3K) where maEpiSCs and conventional EpiSCs clustered closer to one another within the two-dimensional space relative to ES cells or MEFs. Heat maps showed that conventional EpiSCs and maEpiSCs exhibit similar global profiles of H3K4me2 relative to ES cells or MEFs (Fig. 3L), where H3K4me2 levels were higher in maEpiSCs and conventional EpiSCs relative to ESCs and MEFs, at genes whose expression is elevated in EpiSCs. Custom genome browser views demonstrate that levels of H3K4me2 and H3K27ac are enriched at pluripotency-associated (Sall4) and EpiSC-enriched genes (Fgf5) in maEpiSCs (Fig. 3M). Altogether, our results demonstrate that the enhancer landscape of maEpiSCs is similar to conventional EpiSCs.   Norm. tag density (log2 RPBM) -14 -12 -10 -8 -6 Norm. tag density (log2 RPBM)  germ layers such as endoderm (glandular endoderm), mesoderm (muscle, fat, osteoclasts, and osteoblasts), and ectoderm (epidermis) (Fig. 4A, middle and right). Likewise, conventional EpiSCs also gave rise to cells found in the three germ layers such as endoderm (glandular endoderm), mesoderm (fat, muscle), and ectoderm (neuroectoderm) (Fig. 4A, left). The ability to generate teratomas demonstrates the functionality of maEpiSCs. Moreover, immunofluorescence analysis showed that in vitro differentiated maEpiSCs and conventional EpiSCs express OTX2 and SOX1 (ectoderm), SMA (mesoderm), and GATA4 (endoderm) (Fig. 4B,C). Overall, our findings demonstrate that maEpiSCs are able to differentiate in vivo and in vitro into cells found in the three germ layers.

Discussion
Previous studies have demonstrated that ES cells can be derived from haploid embryos 12 . However, EpiSCs have not been directly generated from haploid embryos. Here, we demonstrate that EpiSCs can be generated from haploid embryos. Although we derived maEpiSCs in FGF4-containing media, it is possible that conventional EpiSC media containing bFGF may also support the generation of EpiSCs from haploid embryos. In addition, our results show that established maEpiSCs described in this study, and conventional EpiSCs, contain a diploid genome with a normal number of chromosomes. However, because ploidy analysis was performed after establishment of maEpiSC lines, and not during the initial derivation, it is unclear whether maEpiSCs exhibited a decrease in the percentage of haploid cells during culture, or whether diploidization of cells is required for derivation of maEpiSCs under these conditions. Therefore, results described here do not rule out the possibility that maEpiSCs could be generated and maintained in a haploid state using culture conditions described in this study, or under alternative culture conditions, as it was previously shown that haploid ES cells could be differentiated into haploid EpiSCs 27 . In addition, while we derived maEpiSCs from blastocyst-stage embryos, conventional EpiSCs are derived from E5.5-E6.5 embryos 1,4 . Therefore, it is unclear whether haploid EpiSCs could be derived from E5.5-E6.5 embryos, as previous results demonstrate that haploid ES cells exhibit diploidization upon differentiation 12 . Moreover, while FGF4 has previously been shown to support the self-renewal of both mouse TS cells and EpiSCs 7,10 , our results demonstrate that the haploid genome of blastocysts favors the generation of EpiSCs over TS cells in FGF4-conditions. Interestingly, while we observed trophoblast giant cells from the outgrowths of haploid blastocysts (Fig. 1D), we did not observe the formation of TS cell colonies, suggesting that formation of TS cells may require a diploid genome.
Our results also show that pluripotent maternal (parthenogenetic/gynogenetic) EpiSCs (maEpiSCs) and conventional EpiSCs exhibit a similarly primed transcriptional profile and epigenetic landscape relative to pluripotent ESCs and differentiated MEFs. We further reveal that the transcriptional profile of maEpiSCs is more similar to conventional EpiSCs than ESCs or differentiated cells such as MEFs or EBs. The transcriptional network intersection between maEpiSCs and conventional EpiSCs may represent a primed-pluripotent gene network. Moreover, the variation in culture conditions between maEpiSCs and conventional EpiSCs represent a resource to identity genes that define the primed-pluripotent network. In this case, alterations in external signals or culture media components, which do not alter self-renewal, may produce transcriptional noise of genes not required for pluripotency. A subset of the stably expressed genes may represent the minimal network of primed-pluripotent EpiSCs.
We also demonstrate that the epigenetic profile of maEpiSCs is more similar to conventional EpiSCs than MEFs or ESCs. These global epigenetic analyses demonstrate similar genome-wide landscapes of active enhancers and histone modifications between maEpiSCs and conventional EpiSCs. In contrast, MEFs and ES cells displayed less similar histone modification profiles. Moreover, enhancer activity, as measured by H3K27ac and H3K4me2 levels, was similar between maEpiSCs and conventional EpiSCs relative to ES cells or MEFs. Our results also demonstrate that super-enhancers are similarly active in maEpiSCs and conventional EpiSCs, and we show that maEpiSCs are fully capable of differentiating in vitro and in vivo. Altogether, our findings provide a novel method to derive EpiSCs from haploid embryos, and lend epigenetic insight into mechanisms of EpiSC self-renewal.

Derivation of maEpiSCs.
Mature oocytes were isolated from the oviducts of C57Bl/6 J female mice that had been superovulated with 5 IU pregnant mare serum gonadotropin (PMSG) and subsequently 5 IU of human chorionic gonadotropin (hCG) 48 hours later. Cumulus-oocyte complexes (COCs) collected from oviducts were treated with 0.1% hyaluronidase to disperse cumulus cells. Denuded oocytes were activated in M16 media using 5 mM strontium chloride and 2 mM EGTA as previously described 11 , and subsequently cultured to the blastocyst-stage in M16 or KSOM media in microdrops covered by mineral oil. Blastocysts were then transferred to 24-well dishes containing inactivated MEFs and media containing RPMI 1640, 20% FBS, sodium pyruvate, L-glutamine, β-mercaptoethanol (100 µM), pen/strep, FGF4 (25 ng/mL), and heparin (1 µg/mL). maEpiSC colonies were pick and expanded in the same media. Experiments were performed at least three times.