Derivation of hypermethylated pluripotent embryonic stem cells with high potency

Naive hypomethylated embryonic pluripotent stem cells (ESCs) are developmentally closest to the preimplantation epiblast of blastocysts, with the potential to contribute to all embryonic tissues and the germline, excepting the extra-embryonic tissues in chimeric embryos. By contrast, epiblast stem cells (EpiSCs) resembling postimplantation epiblast are relatively more methylated and show a limited potential for chimerism. Here, for the first time, we reveal advanced pluripotent stem cells (ASCs), which are developmentally beyond the pluripotent cells in the inner cell mass but with higher potency than EpiSCs. Accordingly, a single ASC contributes very efficiently to the fetus, germline, yolk sac and the placental labyrinth in chimeras. Since they are developmentally more advanced, ASCs do not contribute to the trophoblast. ASCs were derived from blastocysts in two steps in a chemically defined medium supplemented with Activin A and basic fibroblast growth factor, followed by culturing in ABCL medium containing ActA, BMP4, CHIR99021 and leukemia inhibitory factor. Notably, ASCs exhibit a distinct transcriptome with the expression of both naive pluripotency genes, as well as mesodermal somatic genes; Eomes, Eras, Tdgf1, Evx1, hand1, Wnt5a and distinct repetitive elements. Conversion of established ESCs to ASCs is also achievable. Importantly, ASCs exhibit a stable hypermethylated epigenome and mostly intact imprints as compared to the hypomethylated inner cell mass of blastocysts and naive ESCs. Properties of ASCs suggest that they represent cells at an intermediate cellular state between the naive and primed states of pluripotency.


Introduction
Development of zygotes to blastocysts is accompanied by a sequential change from totipotency to pluripotency, with differentiation of the outer trophectoderm cells and the pluripotent cells in the inner cell mass (ICM). The ICM gives rise to all the somatic tissues, as well as germ cells. Pluripotent embryonic stem cells (ESCs) isolated from the ICM are in vitro counterparts [1][2][3], which when introduced into the blastocysts can differentiate into all fetal tissues and germ cells, albeit making an insignificant contribution to the extraembryonic tissues [4].
Following blastocyst implantation, epiblast cells in postimplantation embryos retain some characteristics of pluripotency while becoming progressively more hypermethylated before the initiation of somatic cell fates [5,6]. Epiblast stem cells (EpiSCs) from postimplantation epiblast cells are hypermethylated but retain some features of pluripotency [7,8], where female EpiSCs have an inactive X chromosome [9]. A variety of self-renewing pluripotent stem cells have been reported [10][11][12][13][14], with variable potencies, epigenetic states and transcriptional regulators. These cells have been generated in vitro using empirical combinations of cytokines, fetal calf serum and hormones, including mouse embryonic fibroblasts (MEF) as feeders.
Here, ASCs showed unique properties with the expression of naive pluripotency and some mesodermal somatic genes in chemically defined medium without serum. Importantly, these cells were also hypermethylated relative to naive ESCs. Altogether, ASCs represent an early postimplantation epiblast-like state of development. However, when injected into host blastocysts, single ASC contributed to embryonic tissues, germline, yolk sac and placental labyrinth that originate from the postimplantation epiblast, but did not contribute to the trophoblast. ASCs exhibit unique features and provide a new model for investigations on the epigenetic and transcriptional states that regulate cell potency.

Derivation and characterization of AFSCs
First, we attempted to derive stem cells from blastocysts in a chemically defined medium with N2B27 in the presence of ActA and basic fibroblast growth factor (bFGF) (but excluding KSR and feeder cells). Most of the E3.5 blastocysts attached within 5-6 days, forming ICM outgrowth, of which ~20% resulted in stable lines from diverse genetic backgrounds; we designated these as ActA/bFGF stem cells (AFSCs) ( Figure 1A and 1B). The rate of AFSC derivation was similar to that reported previously for naive ESCs in 2i/LIF [15].
Immunofluorescence (IF) and RT-qPCR analysis of AFSCs revealed expression of OCT4, SOX2, NANOG, E-cadherin; we also detected H3K27me3 which is a distinctive mark of an inactive X chromosome ( Figure  1C; Supplementary information, Figure S1A). AFSCs share similarities with EpiSCs [7,8], where expression of some genes of the ICM and ESCs were significantly reduced, including Prdm14, Tbx3, Tcl1, Klf2, Esrrb, Klf4 and Rex1. By contrast, we observed higher expression of early germ layer genes Eomes, Lefty1 and brachyury (T) ( Figure 1C), together with the strong expression of Sox17 and Blimp1. AFSCs also uniquely displayed expression of Gata6 and Gata4, which are a part of the primitive en-doderm transcriptome ( Figure 1C). Overall, AFSCs were clearly at a stage beyond the pluripotent state of epiblast cells in blastocysts but as teratomas, they were still able to differentiate into cells of all the three germ layers (Supplementary information, Figure S1B).

Derivation and characterization of ASCs
Next, we explored the impact of Wnt and BMP4 signaling on AFSCs, which can substitute for serum in ESC cultures [16,17], and induce differentiation of EpiSCs [18,19]. Using CHIR99021 (CH) for simulating canonical Wnt signaling [20], we started with two established AFSCs cell lines with GOF-GFP or Stella-GFP reporters. AFSCs were dissociated and cultured in N2B27 medium with ActA, BMP4, CH and LIF (henceforth called ABCL medium) ( Figure 1D). After 10 days, we detected GOF-GFP-positive clones, which were able to self-renew for more than 30 passages. We designated these cells as advanced pluripotent stem cells (ASCs) and compared them with EpiSCs (Supplementary information, Table  S1B).
To determine the key features of ASCs relative to ESCs, we examined expression by IF for cMYC, KLF4, E-cadherin, ESRRB and PRDM14. Excluding the higher expression of cMYC in ASCs than in ESCs, expression of other genes was equivalent to that in ESCs ( Figure 2A). Notably, high expression of cMYC in ESCs promotes rapid transition through the G1 phase of the cell cycle, suppresses differentiation and controls pluripotency [21].
Next, we assessed the differentiation potential of ASCs with GOF-GFP reporter in chimeric embryos by injecting 15 cells into individual blastocysts. We found that ASC contributed to the embryo proper and germline of E12.5 embryo, as well as to the full-term pups (Figure 2B and 2C; Supplementary information, Table S1C and S1D). Importantly, ASCs contributed to the placenta in chimeric conceptuses (Supplementary information, Figure S2A), which was also the case with the other ASC line harboring Stella-GFP reporter (Supplementary information, Figure S2B). Furthermore, using ASCs with GOF-GFP and a constitutive tdTomato reporter, we detected these cells in the placental labyrinth ( Figure  2D; Supplementary information, Figure S2C), yolk sac ( Figure 2E and 2F), embryo proper and the germline at E10.5-E13.5 (Supplementary information, Figure S2D, S2E and Table S1E). Notably, ASCs contributed to more than 50% of the extra-embryonic mesoderm of yolk sac, but there was no detectable contribution to the extra-embryonic endoderm of yolk sac ( Figure 2E and 2F). This result indicates that ASCs are in a stable state between pluripotency and the onset of differentiation, with a potential to contribute to embryos, yolk sac and placental labyrinth.
For a further stringent test of the potency of ASCs, we introduced them into tetraploid host blastocysts, where the donor cells contribute predominantly if not exclusively to the embryo [22]. Using 15 donor ASCs, we obtained live pups with normal fertility at a higher rate than reported previously [23,24] (Figure 2G and 2H; Supplementary information, Table S1F). The derivation efficiency of clonal lines from single ASC was equivalent to that of naive ESCs and much higher than that of EpiSCs or AFSCs ( Figure 2I). Thus WNT and BMP4 signaling pathways increase the potency of AFSCs in the course of ASC derivation, and this potency exceeds that shown by EpiSCs in vivo.
Finally, we tested the potency of ASCs more stringently by introducing single ASCs into 8-cell stage embryos, which were cultured overnight and then transferred to recipient females. We found that 24%-36% of the embryos at E6.5-E7.5 and 31%-50% of the embryos at E8.5-E10.5 contained derivatives of ASCs with tdTomato reporter ( Figure 2J; Supplementary information, Figure S2F and Table S2A). We also observed germline transmission and full-term chimeras from single ASCs (Supplementary information, Table S2B and S2C). Additional analysis of 18 conceptuses at E8.5-E10.5 resulting from single donor ASCs showed 14 chimeras (78%) with contributions to the embryo, yolk sac and the placenta ( Figure  2J; Supplementary information, Figure S2F and Table  S2A). In contrast, single naive donor ESCs cultured in 2i/LIF resulted in just 2/20 chimeras, with a contribution to the embryo, and no contribution to the extraembryonic tissues. When the number of donor cells was increased to 15, we obtained 18 chimeras, in which donor cell contributed predominantly to the fetus. However, in only one of these chimeras, ESC cells contributed to the fetus, yolk sac and the placenta (Supplementary information, Table S2D). Altogether, ASCs exhibit a unique pluripotent state, with high clonogenicity and potency, and a single ASC contributed to the embryo, placental labyrinth and the germline.
To test for the potential for the conversion of ESCs (2i/LIF) to ASCs, we cultured two ESC cell lines (SQ3.3 and E14). Both cell lines demonstrated normal growth and were passaged over 20 times in ABCL medium; and the resulting cells was designated as ESC-ASCs. We tested the in vivo developmental potential of ESC-ASCs in chimeras using single or multiple ESC-ASCs. We found that both single and multiple ESC-ASCs contributed to the embryo, yolk sac and placenta labyrinth (Supplementary information, Figure S2G; Supplementary information, Table S2E). Thus ESCs can be converted to ASCs by culturing in ABCL, exhibiting increased potency and characteristics resembling ASCs.

Molecular features of ASCs
To investigate the molecular features of ASCs, we profiled their transcriptome by RNA sequencing (RNA-seq), and compared it with naive ESCs (cultured in 2i/LIF), EpiSCs and EpiLCs (epiblast-like cells) (Supplementary information, Table S3A). Unsupervised hierarchical clustering (UHC) showed that ASCs and naive ESCs cluster together, whereas AFSCs are close to EpiSCs ( Figure  3A; Supplementary information, Figure S3A). Moreover, the t-distributed stochastic neighbor embedding (t-SNE) analysis indicated a trajectory representing naive to primed cell states from ESCs and ASCs toward EpiLCs, and to more advanced AFSCs and EpiSCs ( Figure 3B).
We examined the similarities and differences among the five cell types by analyzing a total of 2 588 differentially expressed genes. These genes were segregated into five distinct classes by UHC ( Figure 3C; Supplementary information, Table S3B and S3C). All the five pluripotent cell types expressed core pluripotency factors Oct4, Sox2 and Nanog ( Figure 3D), but the expression of naive pluripotency genes (e.g., Esrrb, Tfcp2l1, Klf4, Tbx3 and Prdm14) was detected exclusively in ASCs and ESCs (cluster 2, Figure 3C and 3D), which was confirmed by IF analysis (Figure 2A). On the other hand, AFSCs and EpiSCs showed enrichment for developmental genes, including Otx2, Zic2, T, Eomes and Foxa2 (cluster 4, Figure 3D).
Notably, we found that 421 genes were exclusively expressed in ASCs (cluster 3, Figure 3C), including genes that encode preferentially expressed antigen of melanoma (PRAME) family and translation initiation factor 1A (see below). Principal component analysis (PCA) of transposable element expression also showed a similar trajectory from naive to primed cell states as seen by analyzing the overall gene expression profiles (Supplementary information, Figure S3B). Notably, both ESCs and ASCs showed high expression of RLTR45, RMER16 and LTRIS2 (Supplementary information, Figure S3C), which might distinguish them from EpiSCs.
To this end, we injected a single ASC cell with td-Tomato and GOF-GFP into an 8-cell stage embryo and examined its chimeric contribution after 48 h of in vitro culturing. Consistently, ASC contributed only to the ICM, and not to the trophoblast in blastocysts (0/42 re-  Figure  S4E).

ASCs exhibit a hypermethylated stable methylome
Naive ESC pluripotent state displays global DNA hypomethylation, where both X chromosomes are active in female cells [15,35,36]. Considering the enhanced potency of ASCs described above, we examined the epigenome of ASCs. Our initial IF observations showed bright H3K27me3 foci in female AFSCs, but not in female ASCs, suggesting reactivation of the X-chromosome in ASCs (Supplementary information, Figure S5A).
We further performed DNA methylation analysis of female ASCs, ESCs and EpiSCs by whole-genome bisulfite sequencing (BS-seq) (Supplementary information, Table S3F). Although naive ESCs (2i/LIF) were globally hypomethylated (median CpG methylation of ~20%), ASCs showed a hypermethylated genome (median ~70%) ( Figure 4A). Indeed, UHC analysis revealed that the methylome of ASCs was more similar to the hypermethylated epigenome of EpiSCs (median 90%) ( Figure  4B; Supplementary information, Figure S5B). Most genomic regions, including promoters, enhancers, exons, introns and intergenic regions, were highly methylated in ASCs (Supplementary information, Figure S5C). In line with a previous report [35], all genomic imprints were erased in female 2i/LIF ESCs, whereas most imprints were retained in both ASCs and EpiSCs ( Figure 4C). The retention of imprints, which are essential for development, would make ASCs more robust in culture, and during subsequent differentiation.
To characterize the methylome further, we calculated hypermethylated regions (HyperMRs) in ESCs and hypomethylated regions (HypoMRs) in ASCs and EpiSCs using a hidden Markov model [37]. We found very few HyperMRs in 2i/LIF ESCs (1 910); and most of these regions also showed high methylation levels in ASCs and EpiSCs ( Figure 4D). We then focused on comparing the methylomes of ASCs and EpiSCs. Among the 51 126 HypoMRs in ASCs, 27 217 (53%) were not present in EpiSCs (Figure 4E and 4F; Supplementary information,  Table S3G). Analysis using Genomic Regions Enrichment of Annotations Tool (GREAT) [38] suggested that the ASC-specific HypoMRs (CpG coverage > 40, methylation difference > 40%) were in the vicinity of genes that function in blastocyst formation (e.g., Esrrb and Oct4), trophectodermal cell differentiation (e.g., Tead4, Tfap2c, Eomes) and dorsal/ventral axis specification (e.g., Wnt3, Pax6) ( Figure 4G). Promoter methylation and gene expression analysis between ASCs and EpiSCs suggested an overall weak anti-correlation ( Figure 4H). However, we found that a small fraction of genes (551) were upregulated in ASCs and exhibited promoter demethylation. They were enriched for genes involved in regulating Meiotic cell cycle (e.g., Mael, Sycp3 and Stra8) ( Figure 4H; Supplementary information, Table  S3H and S3I). These germline genes are promoter methylation-sensitive and are expressed in naive ESCs, where their promoters are demethylated. In addition, we found that naive pluripotency genes, such as Dppa3, Zfp42 (also known as Rex1), Tcl1, Esrrb, Gdf3 and Fbxo15, were also promoter demethylated and upregulated in ASCs ( Figure 4H and 4I; Supplementary information, Figure  S5D). Thus, ASCs exhibited an overall hypermethylated genome and maintained imprints, but showed targeted demethylation at loci associated with the preimplantation development and naive pluripotency. This is consistent with their high contribution in chimeras described above.

Discussion
We describe the derivation of two novel cell types, AFSCs and ASCs from mouse blastocysts, which provide new insights on pluripotent stem cells. In particular, the ASCs (advanced pluripotent stem cells) show higher DNA methylation levels than naive ESCs ( Figure 4A and 4B), and with the expression of both pluripotent and somatic genes ( Figure 3C; Supplementary information,  [25], based on 1 685 dynamically expressed genes [25]. Arrow indicates developmental progression from E2.5 morula to E5.5 postimplantation epiblast. (F) Pair-wise gene expression comparison between ASCs and ESCs (2i/LIF). Upregulated (red) and downregulated (blue) genes in ASCs were highlighted (mean log 2 (normalized read counts) > 3 in either samples, log 2 (fold change) > 2, adjusted P-value < 0.05). Gene ontology analysis (by DAVID) showed ASCs upregulated genes that are involved in multicellular organism development and cell differentiation.  Figure S3E). We also find that genomic imprints are more stable in ASCs relative to naive ESCs ( Figure 4C). On the basis of the transcriptome and DNA methylome analysis, ASCs appeared to be at an intermediate state between naive ESCs and primed EpiSCs, and apparently represent the cells with the characteristics of the early postimplantation epiblast. Remarkably, ASC has high developmental potency in chimeras, since a single ASC contributes extensively to the developing embryo, germline and the extraembryonic mesoderm ( Figure  2J). Indeed it is possible to obtain a live pup consisting exclusively of ASCs in "tetraploid rescue experiments". The high potency of individual ASCs may in part be due to the shorter cell cycle time (data not shown) and higher expression of cMYC in ASCs as compared to that in ESCs (Figures 2A and 3F). Accordingly, cMYC might promote rapid cell transition through the G1 phase of the cell cycle while suppressing cell differentiation by activating Dusp2 and Dusp7 phosphatases, which in turn represses FGF/ERK signaling [21]. The relevance and significance of the cMyc-Dusp2-FGF/ERK pathway in ASCs merit further investigation.
Concerning the extraembryonic tissues, individual ASCs do not contribute to the trophectoderm cells when introduced into 8-cell stage embryos (Supplementary information, Figure S4E), unlike some recently described pluripotent stem cells [11][12][13][14]. This observation is consistent with ASCs being equivalent to early postimplantation epiblast cells, and therefore developmentally they represent cells at a later stage when we would not expect their contribution to the trophoblast cells, which also accounts for our observation that ASCs did not contribute to the extraembryonic endoderm. Nevertheless, single ASCs contribute significantly to some extraembryonic tissues that originate from the postimplantation epiblast, especially the yolk sac mesoderm and placental labyrinth in chimeric conceptuses as compared to ESCs; this is also a mark of higher potency of ASC [39]. Altogether, ASC contribute extensively to the tissues that normally originate from postimplantation epiblast cells, which is consistent with their overall molecular state and characteristics.
Another feature of interest concerning ASC involves the Hoxa family genes. We found that the lncRNA Haunt/Halr1 located ~40 kb upstream of the Hoxa cluster and showed higher expression in ASCs than in ESCs (Supplementary information, Figure S3G). The Huant/ Halr1 genomic locus has been shown to be bound by pluripotency regulators, including Oct4, Nanog and Sox2, with enhancer function-related epigenetic marks, such as H3K4me1, DNase I hypersensitivity, H3K4me3 and H3K36me3 [29,30]. The Haunt lncRNA and its ge-nomic locus regulate the Hoxa gene cluster during differentiation of pluripotent cells [31]. Our data suggest that high expression of Haunt/Halr1 in ASCs may promote downregulation of Hoxa cluster to maintain a stable pluripotent state.
How precisely the chemically defined ABCL medium induces and maintains the hypermethylated state in ASC deserves further investigation in the future. The relatively hypermethylated state of ASCs with intact genomic imprints compared to naive ESCs is likely to provide greater stability during self-renewal in vitro, which is consistent with the findings of two recent studies on ESCs [40,41].
The AFSCs we describe in this study are quite similar to EpiSCs, which differentiate in the presence of Wnt and BMP4 [18,19]. On the other hand, Activin A and LIF support self-renewal of AFSCs and ESC [17], respectively. Thus, during the derivation of ASCs, BMP4 and CH would likely promote AFSCs toward differentiation, which would be counteracted by Act A and LIF that supports self-renewal. Under these signaling conditions, a small proportion of surviving cells acquire new phenotypes and we describe these cells as ASCs. Further detailed analyses, including analysis of epigenetic resetting, determining how precisely the ASCs become established in culture, need to be explored in the future studies. Notably, ESCs can also be readily converted to ASCs, referred to here as ESC-ASCs, with similar properties.
In conclusion, ASCs are a novel and distinct self-renewing type of pluripotent stem cells, which exhibit intermediate features between ESCs and EpiSCs. Transcriptionally, ASCs are at a cellular state that beyond the epiblast state of the blastocyst. In this respect, ASCs are like EpiSCs, albeit some loci remain hypomethylated in ASCs, and importantly, female ASCs do not exhibit X-inactivation. Thus, ASC represents a unique pluripotent state with distinct transcriptional and epigenetic features, which confers greater stability and potency.

Derivation of AFSC cell lines
Mouse embryos at E3.5 were isolated from 129/sv females mated with Oct4-ΔPE-GFP (GOF-GFP), Stella-GFP, Rosa129/Sv and Rosa26R-mT/mG transgenic males with a mixed background of MF1, 129/sv and C57BL/6J strains [42][43][44]. The GOF-GFP expression of the reporter that lacked the proximal enhancer and was under the control of the Oct4 promoter and distal enhancer. This GFP transgene showed expression in the ICM of blastocysts, PGC in vivo and in ESC. Rosa129 is a knockin mouse line with a LacZ reporter in the Rosa locus, where LacZ is ubiquitously expressed. Standard AF medium consists of N2B27 medium (Life technology) supplemented with Activin A (20 ng/ml, R&D Systems) and bFGF (12 ng/ml, R&D systems). E3.5 blastocysts were cultured in M16 medium (Sigma-Aldrich) for 2 days. Most of them were zone pellucida free. If blastocysts had zona pellucida, they were removed by Acidic Tyrode's Solution (Sigma-Aldrich). They were then placed in 24-well fibronectin-coated (16.7 µg/ml, Millipore) plate with AF medium. Epiblast cultures grew efficiently and formed "flat" epithelial-like colonies after 5-6 days. The resulting colonies were further cut into smaller pieces by glass needles after 7-9 days of culturing, and then the colonies passaged by Accutase (Life technology) regularly every 2-3 days. These cells, referred to as AFSCs, could be passaged for over 30 passages in AF medium.

Derivation of ASC cell lines
We dissociated GOF-GFP AFSCs using Accutase and placed AFSCs (3 × 10 5 single cells) in fibronectin-coated 24-well plates with 1 ml ABCL medium. The ABCL medium was N2B27 medium with Activin A (20 ng/ml, R&D systems), BMP4 (50 ng/ ml, R&D systems), CHIR9902 (3 µM, Miltenyi Biotech) and leukemia inhibitory factor (1 000 IU/ml, Millipore). Dependent on cell growth, AFSCs were then passaged after 2-4 days. After 6-7 days of ABCL treatment, we noted cell dying. GOF-GFP-positive clones started to appear around day 10-14. When these GFP-positive colonies grew to around 200 µm in diameter, they were cut into smaller pieces using glass needles and then transferred. When these colonies had grown for 6-7 days, they were treated with Accutase, and the resulting cells were cultured to produce GFP-positive colonies, which were capable of self-renewal for over 30 passages. We called these cells ASCs.

Teratoma formation
The AFSCs were disaggregated using Accutase, and 1 × 10 6 cells were injected under the epithelium of NOD-SCID mice. Three to five weeks after transplantation, tumor(s) were collected and fixed with 4% paraformaldehyde, and processed for paraffin sectioning. Sections were observed following hematoxylin and eosin staining.

Production of chimeras
Approximately, 15-20 cells were injected gently into ICR mice blastocoel cavity of blastocysts (E3.5) using a piezo-assisted micromanipulator attached to an inverted microscope. The injected embryos were cultured in KSOM medium (Millipore) to enable re-expansion of the blastocoel cavity and then transferred to the uteri of pseudopregnant ICR mice at 2.5 days post coitus (dpc). In a similar manner to the injection of blastocysts, 8-cell embryos (E2.5) were injected with a single cell placed carefully into the perivitelline space under the zona pellucida. The embryos were cultured overnight in KSOM medium (Millipore) at 37 °C in a 5% CO 2 atmosphere and transferred to the uteri of pseudopregnant ICR mice at 2.5 dpc. The embryos were isolated at embryonic stage E6.5-E13.5 with chimeric contribution checked. Full-term chimeras were confirmed by the coat color pattern of the pups at birth.

Production of full-term pups in tetraploid host blastocysts
Two-cell stage embryos (E1.5) from ICR mice were collected by flushing oviducts; they were subjected to electrofusion to create tetraploid (4N) host blastocysts. Typically 15-20 ASCs were injected into tetraploid host blastocyts, which were transferred to pseudopregnant recipients at 2.5 dpc.

Immunofluorescence
Cells were briefly washed with PBS and fixed in 4% parafor-maldehyde in PBS for 15 min at room temperature. Cells were permeabilized for 30 min with 1% BSA and 0.1% Triton X-100 in PBS. Antibody staining was carried out in the same buffer at 4 °C overnight. The slides were subsequently washed three times in PBS with 1% BSA, 0.1% Triton X-100 (5 min each wash), then incubated with secondary antibody for 1 h at room temperature in the dark, washed once for 5 min in 1% BSA, 0.1% Triton X-100 in PBS and twice for 5 min in PBS. The slides were then mounted in Vectashield with DAPI (Vector Laboratories) and imaged using a BioRad Radiance 2100 confocal microscope. Primary antibodies used were: mouse monoclonal Oct4 (BD Biosciences, 1:200), rat monoclonal Nanog (eBioscience, 1:500), goat polyclonal Sox2

RNA-seq library preparation
The RNA-seq libraries for this paper were constructed by using NEBNext Ultra Directional RNA Library Prep Kit from Illumina. After isolation of total RNA from bulk amount of mouse stem cells using RNeasy Micro Kit (Qiagen), ~500 ng total RNA was used to enrich mRNA with polyA tails by using NEBNext Magnetic Oligo d(T)25 beads. After RNA fragmentation and random priming, the first strand cDNAs and second strand cDNAs were synthesized, which was followed by end repair, dA-tailing, adapter ligation and PCR enrichment. The RNA-seq libraries were sequenced on HiS-eq4000 platform.

RNA-seq analysis
Paired-end RNA-seq reads were mapped to the mouse reference genome (GRCm38/mm10) with TopHat2 (https://ccb.jhu.edu/ software/tophat), and one mismatch was allowed. The alignments were guided by Ensembl gene models (Ensembl release 86). Read counts for either the full transcript or only the first exon were obtained with featureCounts (http://bioinf.wehi.edu.au/feature-Counts). Transcript or first exon counts were normalized, and the statistical significance of differential expression between samples was assessed using the R Bioconductor DESeq2 package (https:// bioconductor.org/packages/release/bioc/html/DESeq2.html). Transcript and exon counts normalized by DESeq2 size factors were subsequently normalized by their length/1 000. RNA-seq data [25] from early mouse embryos were obtained from the European Nucleotide Archive (accession number: ERP007120), and identically processed.
Paired-end reads were further aligned to the mouse reference genome (GRCm38/mm10) with bowtie (options: "-M 1 -v2 -best -strata"). RepeatMasker annotations of transposable elements on the mouse reference genome were obtained from the UCSC genome browser table (GRCm38/mm10). Counts for all TE copies and Ensembl transcripts were obtained using featureCounts, and TE counts were subsequently processed by DESeq2 as described above, with the exception that TE counts were normalized by the total number of transcript counts instead of DESeq2 size factors.
The top 90% expressed genes (log 2 (normalized counts)) in the whole RNA-seq data set were used for clustering analysis. Pearson correlation analysis was performed by the R cor function with "1-Pearson correlation coefficient" as a distance metric. UHC analysis was performed by the R hclust function with the "ward" method and Euclidean distance metric. t-SNE analysis was performed using the R Rtsne function omitting an initial PCA step. PCA analysis was performed with the R prcomp function on centered and scaled log 2 (normalized counts) expression values. GSEA analysis was performed by the javaGSEA programme developed by the Broad Institute.

Real-time PCR
The cDNAs from bulk amount (2 µg) of total RNAs were diluted 10 to 40-fold. Then 2 µl of diluted cDNAs was used for each 20 µl real-time PCR reaction (1× Syber Green PCR Master Mix, 0.2 µM of each primer). All reactions were duplicated. The PCR was carried out as follows: first, 95 °C for 10 min to activate the Taq polymerase, followed by 40 cycles at 95 °C for 15 s (for denaturation) and for 1 min at 60 °C for annealing and extension. Finally a dissociation step was run to exclude the possibility of non-specific amplification.

Alkaline phosphatase staining
Alkaline phosphatase (AP) staining was carried out using AP Staining Kit from Sigma (86R-1KT) according to manufacturer's instructions. Briefly, the cells were fixed by 4% paraformaldehyde for 10 min, and then were stained overnight by AP staining solution at room temperature in the dark.

BS-seq library preparation
Genomic DNA (two biological replicates per condition) was extracted by QIAamp DNA Micro Kit (Qiagen). Subsequently, 100 ng of genomic DNA was subjected to 3.5 h bisulfite treatment using the Methylcode Bisulfite Conversion Kit (Invitrogen). To monitor the bisulfite conversion efficiency, unmethylated Lambda DNA (0.5 ng, Promega) was spiked-in before conversion. Single-stranded BS-converted DNA was subjected to the post-bisulfite adaptor tagging (PBAT) protocol as described previously [45,46]. Libraries were subjected to single-end 100 bp sequencing on HiSeq 4000 sequencing system (Illumina). Coverage information was summarized in Supplementary information, Table S3.
Post-alignment analysis was performed by using the MethPipe package [37]. First, aligned sequences were converted to MethPipe format using the to-mr command. Methylation levels and coverage for each symmetric CpG site were calculated by the methcounts and symmetric-cpgs commands. Average CpG methylation levels of annotated genomic regions, i.e., promoters, exons, repeats, imprint control regions, CGIs, enhancers and 2 kb genomic tiles, were calculated with the MethPipe roimethstat program. Only high-confidence genomic regions with at least 40 CpG observations from reads in the region were used in further analyses.
About 2 kb tiles were calculated for all chromosomes with a 1 kb offset. Coordinates of promoters (−1 000 bp and +100 bp from the transcription start site (TSS)), exons, introns and intergenic regions were retrieved from Ensembl (Mouse Release 86). CGI annotations were obtained from the UCSC Table Browser. Imprint control region annotations were obtained from Tomizawa et al [48]. Tissue-specific enhancer coordinates (± 1 000 bp from peak; mm9 coordinates were converted to mm10 by the UCSC liftOver tool) were retrieved from the Mouse Encode Project [49]. HyperMRs and hypomethylated regions were calculated by the hmr command from the MethPipe package.
UHC of 2 kb tile methylation was performed using the R hclust command with the "ward" method. Correlation matrix was calculated by the R corrplot package. Profile plot was created by the R seqplots package. Gene ontology analysis was performed by GREAT [50] and DAVID Bioinformatics Resource 6.8 [50].

Data accession
All the sequencing data were deposited in the NCBI, Gene Expression Omnibus (GEO) under accession number GSE99494.