TET proteins are critical for hPGCLC differentiation

To explore the role of TET-mediated demethylation in hPGC specification, we used the CRISPR-Cas9 system to generate a set of TET1, TET2 and TET3 knockout H1 hESC lines bearing a BLIMP1-2A-mKate2 reporter, which we established before to optimize hPGCLCs induction methods (Additional file 1: Fig. S1A). The BLIMP1-2A-mKate reporter activates the expression of mKate2 upon BLIMP1 expression [26]. The knockout cells were analyzed for targeted mutations of the relevant loci by DNA sequencing, and the efficiency of knockout were varied with different sgRNAs (Additional file 1: Fig. S1B–D). TET1 single KO cells and TKO cells exhibit separate mutation sites at the TET1 gene (Additional file 1: Fig. S1C), which reduces the chance of off-target effects causing the following phenotype.

Our results showed that the 5-hydroxymethylcytosine (5hmC) levels were dramatically decreased when TET1 was inactivated, and no 5hmC signal was detected by dot blot in TKO hESCs. But all the cell lines exhibited no difference in 5mC levels (Additional file 2: Fig. S2A, B). Meanwhile, TKO hESCs still maintained normal morphology, and expressed pluripotency markers, such as NANOG, SOX2 and POU5F1 (Fig. 1A). And TKO hESCs showed no difference in proliferation ability compared with WT hESCs (Additional file 2: Fig. S2C). We applied our optimized method to generate hPGCLCs (Fig. 1B), however, TKO hESCs displayed a complete inability to form TNAP (tissue-nonspecific alkaline phosphatase, a PGC and pluripotency marker in humans and mice)/BLIMP1 double-positive hPGCLCs from day 2 to day 8 of induction as determined by FACS (Fig. 1C, D). And activation of key PGC genes like SOX17, TFAP2C, NANOS3, BLIMP1 and POU5F1 (also known as OCT4) was repressed upon hPGCs induction from TKO hESCs, suggesting that the TET proteins are necessary for the initiation of hPGC specification (Fig. 1F). But the expression of SOX2, which is upregulated in mPGCs, was downregulated in both TKO and WT hESCs upon hPGCLC induction (Fig. 1F). During the hPGCLC induction process, the percentage of BLIMP1-mKate2 positive cells increased progressively until day 4, resulting in 37% ~ 50% of TNAP/BLIMP1 double-positive putative hPGCLCs. Similar to other studies, hPGCLCs did not proliferate significantly after day 4 of induction (Fig. 1D). Immunofluorescence confirmed that BLIMP1-mKate2 expression coincided with POU5F1, SOX17 and TFAP2C in day 4 embryoids (Fig. 1E).

Fig. 1

TET TKO hESCs Exhibit hPGC Differentiation Defects. A Left, A phase-contrast image of TKO hESCs. Scale bar = 100 μm. Right, FACS analysis for POU5F1, SOX2, NANOG, TRA-1-60 and SSEA-4 expression in TKO hESCs; B Scheme of hPGC differentiation through iMeLCs in vitro; C FACS analysis of WT and TKO hESCs on hPGCLCs induction for 8 days. Boxed areas indicate TNAP/BLIMP1 (+) cells with their percentages. D Quantification of TNAP/BLIMP1 (+) cells at day 2, 4, 6 and 8 of hPGC induction; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided), ***represent compared to WT group p < 0.001; E Immunofluorescence of SOX17, TFAP2C, POU5F1, BLIMP1 and SOX2 at the day4 embryoids for WT and TKO cells. Scale bar = 50 μm; F RT-qPCR analysis for SOX17, BLIMP1, TFAP2C, NANOS3, POU5F1 and SOX2 during hPGC differentiation in day 2 ~ 8 embryoids; n = 3 independent experiments. Data are presented as means ± s.d

Interestingly, the efficiency of hPGCLC induction from TET1-inactivated cell lines was significantly lower than that of WT hESCs, but TET2 and/or TET3 knockout did not affect hPGCLCs formation in vitro (Additional file 3: Fig. S3A–D, Additional file 4: Fig. S4). Meanwhile, there is no difference in morphology and key gene expression between TKO and WT hESCs after iMeLCs induction (Additional file 5: Fig. S5A, B). Considering TKO hESCs completely abolished hPGCLC formation, we infer that hPGCLC induction from hESCs is dominantly regulated by TET1-mediated DNA demethylation, and TET proteins act complementarily to orchestrate the epigenetic regulation in the hPGC differentiation process.

Inactivation of TET proteins impairs LEFTY-NODAL signaling pathway during hPGCLC specification

Owing to difficulties in studying human embryos without TET proteins, we compared the RNA-seq data of mouse E6.5 Tet-null epiblasts (where and when mice PGCs are first specified) and that of TKO day4 embryoids [27]. Gene set enrichment analysis (GSEA) highlights the similarities between mouse E6.5 Tet-null epiblasts and TKO day4 embryoids (Fig. 2A, B). Genes that were up- or down-regulated in mouse E6.5 Tet-null epiblasts were highly biased to be up- or down-regulated in TKO day4 embryoids, respectively. Lefty1 and Lefty2 are members of the TGF-β superfamily and antagonize the Nodal signaling that is essential for primitive streak and mesoderm development in mice [28, 29]. Interestingly, they were among the significantly downregulated genes in TKO day4 embryoids (Fig. 2B), but NODAL was also decreased during hPGCLC differentiation (Fig. 2C, Additional file 6: Fig. S6I), and the promoters of NODAL and LEFTY1/2 were hypermethylated (Fig. 2D, E). This is different from Tet-null mouse epiblasts in which increased Nodal signaling was observed, probably due to diminished expression of Lefty1 and Lefty2 genes (Fig. 2F). But in pluripotent cells, inactivation of TET proteins led to the upregulation of NODAL, LEFTY1, LEFTY2 and their downstream gene NANOG, whereas no difference was shown in their methylation levels (Figs. 2–E, 5G). However, NODAL, LEFTY1, LEFTY2 and NANOG were rarely expressed in TKO day4 embryoids compared to WT day4 hPGCLCs, and the expression of p-SMAD2/3 was also decreased in TKO day4 embryoids, which demonstrated NODAL signaling is critical for hPGCs specification (Fig. 2C, G, Additional file 6: Fig. S6I, J).

Fig. 2

Impaired LEFTY-NODAL Signaling Pathway in TKO hESCs. A Gene set enrichment analysis (GSEA) highlights the similarities between TKO day4 embryoids and mouse Tet-TKO E6.5 epiblasts. Genes upregulated in Tet-TKO E6.5 epiblasts are highly biased to be upregulated in TKO day4 embryoids; B Upregulated or downregulated genes between TKO day4 embryoids and mouse Tet-TKO E6.5 epiblasts by GSEA. LEFTY1, LEFTY2 and NANOG were highlighted by red box; C RNA-seq TPM of NODAL, LEFTY1 and LEFTY2 in each cell line; D Heat map showing DNA demethylation dynamics of NODAL, LEFTY1, LEFTY2 promoters in each sample, black boxes indicate the differential methylation level, * represent compared to WT_d4 group p < 0.05; E Analysis of the percentage of 5mC at the NODAL, LEFTY1, LEFTY2 promoters by Epimark in hESCs and day 4 embryoids upon hPGCs induction; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided): * represent compared to WT_d4 group p < 0.05. F RNA-seq counts of Nodal, Lefty1, Lefty2 in mice WT E6.25 epiblast and TKO E6.25 epiblast; G RNA-seq TPM of NANOG, POU5F1, SOX2 in hESC, WT day4 hPGCLCs and TKO day4 embryoids

Next, we used NODAL signaling inhibitor SB431542 and stimulator Activin A to investigate the role of NODAL signaling in hPGCLC induction (Fig. 3A). Our results showed that NODAL inhibitor completely blocked hPGCLC differentiation and small embryoids aggregated in day4 (Fig. 3B, C). Interestingly, adding SB431542 in the iMeLCs stage also inhibited hPGCLCs formation, because inhibited NODAL signaling in the iMeLCs stage impaired mesoderm-like cell differentiation and blocked hPGCLCs induction. This result confirms the significance of NODAL signaling in the iMeLCs phase. Both NANOG and SOX17 are downstream genes of the ACTIVIN/NODAL signaling pathway and are vital for the self-renew of pluripotent cells and endoderm specification, respectively [30, 31]. Moreover, NANOG and SOX17 are essential for hPGC differentiation. When the different concentration of Activin A was added in PGCLCs induction stage, there were no TNAP/BLIMP1 double-positive cells in embryoids derived from either TKO or WT pluripotent cells (Fig. 3D), but a small number of TNAP/BLIMP1 double-positive cells were detected in embryoids induced by Activin A without other cytokines (BMP4, SCF, EGF and LIF) (Fig. 3E). However, the RT-qPCR results showed that SOX17 and BLIMP1 were upregulated, but other key hPGC genes such as TFAP2C, NANOS3, NANOG and POU5F1 were barely expressed compared to the day4 hPGCLCs group. Notably, the embryoids also expressed endoderm, trophectoderm and ectoderm markers like GATA4, EOMES, and PAX6 after Activin A treatment (Fig. 3F). These results suggest that those Activin A-induced TNAP/BLIMP1 double-positive cells were not hPGCLCs, and that ACTIVIN/NODAL signaling can activate SOX17 only, but fails to maintain the expression of the pluripotency factor NANOG in hPGCLCs differentiation. Hence, we assume that TET proteins may regulate the NANOG expression in hPGC specification.

Fig. 3

LEFTY-NODAL Signaling Pathway is Essential for hPGCLCs Differentiation. A Scheme of hPGC differentiation through iMeLCs in vitro; SB431524: NODAL signaling inhibitor; ACTA (Activin A): NODAL signaling stimulator; B Bright field (BF) and fluorescence images of day4 embryoids stimulated by cytokines and with or without SB431542 in iMeLCs or hPGCLCs induction. Scale bar = 100 μm; C FACS patterns show the induction efficiency of hPGCs with or without SB431542 at iMeLCs or hPGCLCs induction; D FACS analysis of WT and TKO hESCs in day4 hPGCLCs induction with cytokines and different concentration of Activin A; E FACS analysis of WT and TKO hESCs in day4 hPGCLCs induction with different concentration of Activin A and without cytokines; F mRNA expression was assayed in day4 embryoids treated with different concentration of Activin A without cytokines using qRT-PCR assay, the WT day4 hPGCLCs were used as positive control; n = 3 independent experiments. Data are presented as means ± s.d

TET proteins mediated balance between NANOG and SOX17 is critical for hPGC induction

Previous studies have demonstrated ACTIVIN/NODAL signaling and NANOG orchestrate human embryonic stem cell fate decisions and SOX17 is critical for hPGC specification [22, 32, 33]. To further distinguish the relationship between ACTIVIN/NODAL signaling and NANOG transcription in hPGCs induction, we overexpressed inducible SOX17 and NANOG transgenes under the control of trimethoprim and doxycycline (TD) in hPGCLCs induction alone or together (Fig. 4A, Additional file 5: Fig. S5C). However, overexpression of NANOG individually elicited an inappreciable response after 4 days of differentiation in both WT and TKO groups, whereas SOX17 alone produces a modest response in the WT group, which is consistency with a previous study [34], but a low response in the TKO group (Fig. 4B, C). Notably, overexpression of NANOG and SOX17 together induced a strong response in the WT group with a large proportion of TNAP/BLIMP1 double-positive hPGCLCs cells and a moderate response in the TKO group, suggesting that NANOG and SOX17 act synergistically and rapidly to induce a similar response to the WT pluripotent cells treated with cytokines for hPGCLC specification. This response is preceded by downregulation of SOX2 and upregulation of the PGC markers NANOS3, TFAP2C, and ‘naïve’ pluripotency genes including KLF4 and TFCP2L1, as well as in vivo hPGCs. Nevertheless, endogenous expression of NANOG and SOX17 was not restored as measured by RT-qPCR with primers targeting their 5’ UTR region (Fig. 4D). Global RNA-sequencing data also demonstrated that the response for hPGCLC specification induced by NANOG and SOX17 co-overexpression is similar to that induced by cytokines (Fig. 6A).

Fig. 4

Hypermethylation of NANOG and SOX17 Promoters in TET TKO hESCs Leads to a Failure of NANOG and SOX17 Activation Upon hPGC Differentiation. A Scheme of hPGC differentiation rescue in TKO hESCs through NANOG and SOX17 overexpression; B FACS analysis for induction of hPGCs by overexpression of NANOG or/and SOX17 in WT and TKO hESCs without cytokines; C Bright field and fluorescence images of day 4 embryoids upon hPGC induction in B. Scale bar = 200 μm; D RT-qPCR analysis for gene expression during hPGC differentiation in day 4 embryoids by overexpression of NANOG or/and SOX17 in hPGCs induction, the WT day4 hPGCLCs were used as positive control; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided): * represent compared to WT group p < 0.05; E ChIP–qPCR for TET1 in WT and TKO hESCs in NANOG and SOX17 promoters, RPL30 as positive control; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided): *p < 0.05; F Analysis of the percentage of 5mC at the NANOG and SOX17 promoters by Epimark with or without trimethoprim and doxycycline (TD) in day 4 embryoids upon hPGCs induction; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided): * represent compared to WT group p < 0.05

Interestingly, NANOG overexpression upregulated mesoderm marker HAND1 and TE marker CDX2, while SOX17 overexpression resulted in the upregulation of endoderm marker GATA4 (Fig. 4D). These results suggest that overexpression of NANOG and SOX17 separately promoted hESCs to diverging germ layers. As reported before, NANOG could bind to the SOX17 promoter to restrain SOX17 expression in pluripotent cells and is highly expressed in hPGCs [32]. Therefore, we deduce that NANOG regulates the expression of SOX17 by binding to its promoter in hPGCs, and that the balance of NANOG and SOX17 guards the initiation of hPGC specification.

To precisely characterize the defect in differentiation potency of TKO hESCs, we also verified that TET1 could bind to the NANOG and SOX17 promoters in WT hESCs by ChIP-qPCR (Fig. 4E) and further analyzed the 5mC levels of NANOG and SOX17 promoters by Epimark 5mC analysis. In comparison to WT pluripotent cells, the promoters of NODAL signaling and NANOG/SOX17 in TKO day4 embryoids showed much higher methylation levels (Fig. 4F, Additional file 8: S8A). Thus, the inactivation of TET proteins in hESCs causes aberrant hypermethylation of NODAL signaling genes and fails to activate NANOG and SOX17 expression, resulting in the consequent defects in hPGC differentiation.

De novo methylation by DNMT3B causes hypermethylation of NANOG and SOX17 promoters

Three DNA methyltransferases, DNMT1, DNMT3A and DNMT3B, are responsible for cytosine methylation in mammals. Although SOX17 promoter was hypermethylated in TKO hESCs, there were no differences in the expression of DNMT genes between WT and TKO pluripotent cells, but DNMT1, DNMT3A and DNMT3B were downregulated in WT d4 hPGCLCs (Additional file 6: Fig. S6A). However, ChIP-qPCR analysis showed increased binding of DNMT3B, but not DNMT1 or DNMT3A, at the NANOG and SOX17 promoters in TKO hESCs as compared to WT hESCs (Fig. 5A, Additional file 6: Fig. S6B). Therefore, we inactivated the DNMT3B gene in TKO hESCs to further investigate whether DNMT3B is responsible for the hypermethylation of NANOG and SOX17 promoters. Using the CRISPR-Cas9 technique, we generated a TET1, TET2, TET3 and DNMT3B quadruple-knockout (QKO) cell line (Additional file 6: Fig. S6C–F). Like TKO hESCs, QKO hESCs still maintained normal morphology and expressed pluripotency markers, such as NANOG, SOX2, and POU5F1 (Additional file 6: Fig. S6G). There were no detectable 5hmC in QKO hESCs and no difference in 5mC levels was observed among QKO, TKO and WT hESCs (Additional file 6: Fig. S6H).

Fig. 5

Genetic Inactivation of DNMT3B Partially Rescues the hPGCLC Differentiation Defect of TKO hESCs. A ChIP–qPCR for DNMT3B at the NANOG and SOX17 promoters in WT and TKO hESCs; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided), *p < 0.05; B Bright field and fluorescence images of Day 4 embryoid with BLIMP1-mKste2 reporter in WT, TKO, QKO hESCs, Scale bar = 100 μm; C FACS analysis for induction of hPGCs in WT, TKO, QKO hESCs; D Quantification of FACS at day 4 of hPGCLC induction in WT, TKO, QKO hESCs; n = 4 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided), * represent compared to WT group p < 0.05, # represent compared to TKO group p < 0.05; E Immunofluorescence of SOX17, TFAP2C, POU5F1, BLIMP1 and SOX2 at the day4 embryoid for WT, TKO and QKO cells. Scale bar = 50 μm; F RT-qPCR analysis for gene expression during hPGC differentiation in day 4 embryoid; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.001; G Methylation analysis of the NANOG and SOX17 promoters in hESCs and day 4 embryoids by Epimark; n = 3 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by one-way ANOVA: * represent compared to WT or WT_d4 group p < 0.05, # represent compared to TKO or TKO_d4 group p < 0.05

Fig. 6

Transcriptome by RNA-seq and DNA Demethylation by Base-Resolution BS-Seq Analysis of Each Group and in vivo Data Sets. A tSNE plot of RNA-seq data. Color codes indicate the cell types, and shapes for cell states; B PCA plot of WGBS data, color codes for the cell types are indicated; C Pseudotime trajectory (Monocle analysis) of the cells. Cells are colored based on the predicted pseudotime; D Heat map of top 20 DEGs in each subpopulation estimated by ROTS. The GO functional terms and representative genes included are shown for each gene cluster; E Violin plots showing the distribution of CpG methylation levels in overlapped 1 kb genomic tiles of each sample, the white point indicates median; F Averaged CpG methylation level profiles of all genes from -5 kb upstream of the transcription start sites (TSS), through scaled gene bodies to + 5 kb downstream of transcription end sites (TES); (G) Methylation distribution according to the different segments defined by the MethylSeekR approach; n = 2 independent experiments. Data are presented as means ± s.d. Statistical analysis was performed by Student’s t-test (two-sided): * represent compared to hESCs group p < 0.05

Furthermore, there was a partial rescue of the hPGC induction from QKO pluripotent cells, with about 30% of TNAP/BLIMP1 double-positive hPGCLCs, as compared to TKO pluripotent cells, and the hPGCLCs separate clearly into two populations even though the induction efficiency was still lower than WT hESCs (Fig. 5B–D). Immunofluorescence and RT-qPCR analysis also showed rescue in the expression of hPGC markers SOX17, TFAP2C, NANOS3 and BLIMP1, and proper inhibition of pluripotency marker SOX2 (Fig. 5E, F). And NODAL, LEFTY1/2, NANOG mRNA expression levels were rescued in QKO day4 hPGCLCs (Additional file 6: Fig. S6I). In addition, methylation levels of NANOG and SOX17 promoters presented a significant reduction in QKO day4 embryoids detected by Epimark 5mC analysis and WGBS (Figs. 5G, 7A). However, the methylation level of the NANOG promoter was unchanged in each of the hESCs groups, implying that other epigenetic modifications or mechanisms may play a role in NANOG regulation. And the methylation levels of the NANOG promoter was increased in the WT_d4 group because the whole embryoid was used. Meanwhile, as a bivalent promoter (marked by H3K4me3 and H3K27me3), SOX17 might be more sensitive to TET and DNMT3B regulation in pluripotent cells. Altogether, these results suggest that DNMT3B is the primary actor in hypermethylation of NANOG and SOX17 promoters and plays a major role in the impaired hPGC differentiation in TKO hESCs. In WT hESCs, TET proteins counteract with DNMT3B to maintain the expression of NANOG and SOX17 and facilitate hPGC differentiation.

Fig. 7

DNA methylation Controls the Balance of NANOG and SOX17 in hPGC Specification. A NANOG, SMAD2/3, TET1 ChIP-seq binding sites and methylation profile for the SOX17 and NANOG locus, red area indicated promoter region; B Density-scatterplot showing differentially methylated promoters in TKO day4 embryoids and WT day4 hPGCLCs; C GO analysis of hypermethylation and hypomethylation promoters in B; D ChIP-qPCR for TET1 in WT day4 embryoids in NANOG and SOX17 promoters; E ChIP-qPCR for NANOG in WT day4 embryoids in SOX17 promoters; F A hypothesis for epigenetic regulation of hPGC fate. The balance between NANOG and SOX17 mediated by TETs and DNMT3B guarantees hPGC specification from pluripotent cells. Overexpression of SOX17 or NANOG would compel cells to endoderm or meso/ectoderm germ layers, respectively; G A model illustrating the different functions of NANOG for PGC specification in mice (top) and humans (bottom). In mice, NANOG activates the expression of Blimp1 and Prdm14 by binding to their enhancers, which could be repressed by Sox2 during mPGC induction in vitro. In humans, TET1 and DNMT3B work oppositely to regulate NANOG expression during hPGC induction in vitro. And NANOG further regulates SOX17 expression by binding to its promoter

Comparison of hPGCs and hPGCLCs by RNA-seq and WGBS analysis

In order to build a comprehensive understanding of the characteristics of pluripotent cells and induced hPGCLCs in each cell line, we determined global mRNA transcription and DNA methylation profiles of each cell type during hPGCLC induction by RNA-seq and WGBS technology, and published in vivo datasets were also included in our analysis. We compared the data with gonadal hPGCs from week 7 male human embryos (Carnegie stage 18/19), which not only retain key characteristics of early hPGCs but also express later germ cell markers such as VASA and DAZL (Additional file 7: Fig. S7C).

t-Distributed Stochastic Neighbor Embedding (tSNE) analysis showed that the pluripotent cells, hPGCs_week7, hSoma_week7, day4 hPGCLCs and TKO day4 embryoids settled at five discrete positions (Fig. 6A). In particular, the day4 hPGCLCs, including the QKO day4 hPGCLCs and the NANOG/SOX17 overexpressed TKO day4 hPGCLCs were clustered together. Unsupervised hierarchical clustering of RNA-seq transcription profiles showed that induced hPGCLCs, gonadal samples and pluripotent cells formed distinct branches. Notably, day4 TKO embryoids formed a sub-cluster with hESCs, suggesting that day4 TKO embryoids impaired hPGC differentiation capacity to form another germ layer (Additional file 7: Fig. S7A). All of the pluripotent cells were distributed together, and WT, TKO and QKO hESCs showed relatively fewer transcriptional changes compared with each other (Additional file 7: Fig. S7F–H). The WGBS PCA plot showed that the TKO and QKO hESCs were clustered together, but away from the WT hESCs. Similarly, QKO_day4 hPGCLCs distributed differentially with WT_day4 hPGCLCs, and closer with TKO_day4 embryoids (Fig. 6B). The clustered heat map of the methylation values according to the first clustered 1000 promoters showed that QKO and WT hESCs, TKO hESCs and TKO day4 embryoids formed two distinct branches (Additional file 7: Fig. S7B). These results demonstrate that even though mRNA expression patterns are consistent, the DNA methylation levels are still in diversity.

An orthogonal pseudotime analysis using the Monocle package further supported that day4 hPGCLCs and gonadal hPGCs were in different developmental branches [35], and day4 TKO embryoids and hSoma_week7 clustered together at the end of the branch (Fig. 6C). To detect DEGs that were specifically distinguished in each cell type, we performed optimized test statistic (ROTS) for the defined populations [36]. Each population was compared to the other pooled populations to find unique gene signatures and upregulated genes with an FDR < 0.001 were considered significantly differentially expressed (SDE). The top 20 SDE genes for each cell type are represented in the heatmap depicted in Fig. 6D. Next, we used the top 100 SDE genes of each population to define the gene signatures by GO analysis and described the GO terms in each subpopulation in Fig. 6D.

The global DNA methylation levels of each sample were slightly changed, but week7 gonad PGCs were going through genome-wide DNA methylation reprogramming (Fig. 6E). However, DNA methylation levels in promoters and CpG island regions were notably different, especially QKO hESCs showed a global decrease in methylation, both in promoters and CpG island (Additional file 7: Fig. S7D). The methylation patterns over genes, with low methylation at the transcription start sites (TSSs) and slightly increased levels over gene bodies, showed no obvious difference between hESCs and induced cells (Fig. 6F). Using MethylSeekR [37], we classified genome methylation regions into highly methylated regions (HMRs) and partially methylated domains (PMDs), which are in a transcriptionally repressed state, unmethylated regions (UMRs) and lowly methylated regions (LMRs), corresponding to proximal and distal regulatory sites, respectively [38, 39]. Our results showed the average methylation range of PMDs, LMRs and UMRs was increased during hPGCLCs induction from WT and QKO pluripotent cells while HMRs were decreased modestly (Fig. 6G). But in the TKO cell line PMDs and HMRs were increased during hPGCLCs induction, and LMRs were modestly decreased. This suggests that the TET and DNMT proteins dynamically regulate the methylation states of PMDs, LMRs and UMRs during hPGC specification.

Since retrotransposons take up about half of the human genome and are mainly repressed by DNA methylation [40], we also evaluated the methylation levels of major human retrotransposon classes. During the hPGCLC induction process, methylation levels of most retrotransposon loci were increased, except for Alu, SVA, and ERVK elements, which were also resisted demethylation in gonad hPGCs. Interestingly, methylation levels of all retrotransposon loci in TKO hESCs were lower than WT hESCs but were upregulated when DNMT3B was knocked out (Additional file 7: Fig. S7E). Therefore, we analyzed the expression of the Krüppel-associated box zinc finger protein (KRAB-ZFP) family which plays a role in restricting transposable elements activity [41, 42]. Unsupervised hierarchical clustering of all annotated KRAB-ZFPs showed the expression of many genes was changed in TKO hESCs, and DEG analysis showed ZNF248 was downregulated in TKO versus WT hESCs, but upregulated in QKO versus TKO hESCs (Additional file 7: Fig. S7I, J). Thus, ZNF248 may confer demethylation resistance in these retrotransposon families and be regulated by TET proteins.

Taken together, the derived hPGCLCs exhibit early-stage germ cell characteristics that are apparently en route to hPGCs, and this in vitro differentiation model provides a method to explore the epigenetic regulation mechanism at the initial stage of hPGC specification, which is otherwise not possible in vivo because postimplantation human embryos before week 4 are inaccessible to be investigated.

Epigenetic regulation of the balance of NANOG and SOX17 during hPGC specification

A previous study reported that the ACTIVIN/NODAL signaling controls the expression of NANOG, which in turn interacts with SMAD2/3 to maintain the expression of pluripotency genes, and binds to endoderm genes to inhibit their expression [32]. We analyzed published NANOG, SMAD2/3, and TET1 ChIP-seq datasets, and found NANOG, SMAD2/3, and TET1 bind to the vicinity of PGC marker genes such as SOX17, NANOG, BLIMP1, TFAP2C and LETFY/NODAL signaling genes (Fig. 7A, Additional file 8: Fig. S8A). Gene-specific inspection of the methylome data revealed a significant increase in DNA methylation at the SOX17, NANOG, BLIMP1, TFAP2C and LETFY/NODAL signaling genes in TKO day4 embryoids (Fig. 7A, Additional file 8: Fig. S8A). Comparing TKO day4 embryoids with WT hPGCLCs, we identified 1770 differentially methylated promoters across the genome and 1192 of the promoters gained methylation after TET proteins inactivation (1192 hypermethylated promoters versus 578 hypomethylated promoters; Fig. 7B). Notably, hypermethylated promoters in TKO day4 embryoids were enriched for Sertoli cell development, anterior/posterior pattern specification and spermatogenesis; while 578 hypomethylated promoters were enriched for detection of chemical stimulus, nervous system process and RNA processing, indicating that DNA demethylation in some regions is primarily required for hPGC specification (Fig. 7C). Compared to RNA-seq results, much more changes were detected in methylation levels of promoters in different types of hESCs, and GO analysis indicated different biological processes were associated with the hypomethylated promoters (Additional file 8: Fig. S8B, C).

Further, ChIP-qPCR in day 4 hPGCLCs demonstrated that TET1 could bind to the NANOG and SOX17 promoters, and NANOG could also bind to the SOX17 promoter, which may be involved in the regulation of SOX17 expression during hPGC specification (Fig. 7D, E). Interestingly, overexpression of NANOG alone in WT hESCs is unable to promote hPGCLC differentiation. On the contrary, overexpression of SOX17 alone can induce hPGCLC differentiation in WT hESCs but not TKO hESCs (Fig. 4B). Moreover, recent research found that SOX17-TFAP2C cooperated to directly upregulate/sustain the expression of core pluripotency factors NANOG and POU5F1 during hPGCLC induction [21, 43]. But our results found that TET proteins were necessary to keep the NANOG promoter at a low methylation levels and that NANOG can bind to the SOX17 promoter to regulate its expression in hPGCLCs specification.