Long‐term expansion with germline potential of human primordial germ cell‐like cells in vitro

Abstract

Human germ cells perpetuate human genetic and epigenetic information. However, the underlying mechanism remains elusive, due to a lack of appropriate experimental systems. Here, we show that human primordial germ cell‐like cells (hPGCLCs) derived from human‐induced pluripotent stem cells (hiPSCs) can be propagated to at least ~10⁶‐fold over a period of 4 months under a defined condition in vitro. During expansion, hPGCLCs maintain an early hPGC‐like transcriptome and preserve their genome‐wide DNA methylation profiles, most likely due to retention of maintenance DNA methyltransferase activity. These characteristics contrast starkly with those of mouse PGCLCs, which, under an analogous condition, show a limited propagation (up to ~50‐fold) and persist only around 1 week, yet undergo cell‐autonomous genome‐wide DNA demethylation. Importantly, upon aggregation culture with mouse embryonic ovarian somatic cells in xenogeneic‐reconstituted ovaries, expanded hPGCLCs initiate genome‐wide DNA demethylation and differentiate into oogonia/gonocyte‐like cells, demonstrating their germline potential. By creating a paradigm for hPGCLC expansion, our study uncovers critical divergences in expansion potential and the mechanism for epigenetic reprogramming between the human and mouse germ cell lineage.

New resource work defines culture conditions for propagation of human germ‐cell precursors.

Introduction

The germ cell lineage ensures the continuity and the generation of diversity of genetic as well as epigenetic information across the generations, thereby serving as a foundation for the perpetuation and evolution of a given species. In humans, anomalies in germ cell development result in critical diseased states, including infertility and genetic and epigenetic disorders of offspring. Thus, investigations into the mechanism of germ cell development constitute a fundamental theme both in biology and in medicine. Accordingly, the mechanism of mammalian germ cell development has been a subject of intensive study using the mouse as a model organism (Lesch & Page, 2012; Saitou & Miyauchi, 2016). On the other hand, until now, the mechanism for human germ cell development has been largely elusive, due to a difficulty in analyzing human germ cell development and a lack of appropriate experimental systems (Saitou & Miyauchi, 2016; Tang et al, 2016).

Recent studies have demonstrated an in vitro reconstitution of germ cell development in mice. In this in vitro system, mouse pluripotent stem cells (mPSCs), including embryonic stem cells (mESCs) and induced pluripotent stem cells (miPSCs), cultured under a “naïve” pluripotent state, are induced into epiblast‐like cells (EpiLCs) and then into primordial germ cell‐like cells (mPGCLCs), which contribute to spermatogenesis upon transplantation into the testes of neonatal mice (Hayashi et al, 2011) or upon aggregation with embryonic testicular somatic cells (reconstituted testes) followed by spermatogonial stem cell‐like cell derivation and transplantation into the testes of adult mice (Ishikura et al, 2016). Moreover, mPGCLCs contribute to oogenesis upon aggregation with embryonic ovarian somatic cells (reconstituted ovaries: rOvaries) followed by transplantation under the ovarian bursa or an appropriate in vitro culture (Hayashi et al, 2012). The resultant spermatozoa and oocytes generate healthy offspring (Hayashi et al, 2011, 2012). In addition, a methodology for the in vitro propagation of mPGCLCs with the use of forskolin and rolipram, which stimulate the intracellular production of cyclic AMPs (cAMPs), has been developed (Ohta et al, 2017). Under this condition, mPGCLCs undergo epigenetic reprogramming, including genome‐wide DNA demethylation, most likely in a replication‐coupled, passive manner, and acquire an epigenetic “blank slate”, an immediately precursory state for sexually dimorphic development (Seisenberger et al, 2012; Ohta et al, 2017). Consequently, based on this in vitro system, the key signals and the underlying mechanisms for the female sex determination of germ cells have been clarified (Miyauchi et al, 2017; Nagaoka et al, 2020).

These advances have been serving as the basis for an in vitro reconstitution of human germ cell development. In this in vitro system, human iPSCs (hiPSCs) cultured under a primed pluripotent state are induced into incipient mesoderm‐like cells (iMeLCs) and then into hPGCLCs bearing a property of early hPGCs (Sasaki et al, 2015; Yokobayashi et al, 2017), or alternatively, hPSCs cultured under a “distinct” state of pluripotency are induced directly into hPGCLCs (Irie et al, 2015). The hPGCLC induction system was instrumental in clarifying the mechanism of hPGC specification, which turned out to involve a network and hierarchy of transcription factors that were evolutionarily divergent from those of mPGC specification (Irie et al, 2015; Kojima et al, 2017). More recently, using a culture system involving the aggregation of hPGCLCs with mouse embryonic ovarian somatic cells (xenogeneic rOvaries: xrOvaries), hPGCLCs have been induced into oogonia/gonocyte‐like cells with appropriate epigenetic reprogramming, acquiring a state immediately precursory to entry into the meiotic prophase (Yamashiro et al, 2018). Combined with rapid progress in the analysis of transcriptomes and epigenetic properties of human germ cells at a single‐cell resolution (Gkountela et al, 2015; Guo et al, 2015; Tang et al, 2015; Li et al, 2017), these studies have paved the way to an improved understanding of the mechanism of human germ cell development.

Here, to establish a further basis for promoting human germ cell biology, we set out to explore a methodology for the in vitro expansion of hPGCLCs. By screening a number of chemicals/cytokines known to have a positive effect on mPGC(LC) expansion, we determined a culture condition for hPGCLCs, under which they proliferated ~10⁶‐fold during a period of 4 months by essentially retaining their original properties, including the transcriptome, the DNA methylation profiles, and the capacity to differentiate into oogonia/gonocyte‐like cells. The hPGCLC expansion system will be instrumental in exploring the mechanism for hPGC specification, survival/propagation, and epigenetic reprogramming.

Results

Exploration of a condition for the in vitro expansion of hPGCLCs

To establish a condition for the in vitro expansion of hPGCLCs, we set out to explore whether hPGCLCs might propagate under a condition similar to that for the in vitro expansion of mPGCLCs (Ohta et al, 2017). For this purpose, hiPSCs bearing the B LIMP1‐td T omato (BT) and TF A P2C‐E G FP (AG) alleles [585B1 BTAG hiPSCs (XY)] (Sasaki et al, 2015; Yokobayashi et al, 2017) were induced into iMeLCs by activin A (ActA) and an activator of WNT signaling (CHIR99021), and then into BTAG‐positive (BT⁺AG⁺) hPGCLCs by bone morphogenetic protein 4 (BMP4), leukemia inhibitory factor (LIF), stem cell factor (SCF), and epidermal growth factor (EGF; Fig 1A). The BT⁺AG⁺ hPGCLCs induced for 6 days (d6 hPGCLCs) were isolated by fluorescence‐activated cell sorting (FACS), and ~5.0 × 10³ such cells were seeded onto a subline of m220 feeders in GMEM with 15% knockout serum replacement (KSR), 2.5% fetal bovine serum (FBS), 100 ng/ml SCF, and various combinations of the chemicals [forskolin (10 μM), rolipram (10 μM), and cyclosporin A (5 μM)] and the cytokines [LIF (10 ng/ml), EGF (50 ng/ml), and basic fibroblast growth factor (bFGF: 20 ng/ml)] (32 combinations in total) that are known to have a positive influence on mPGC(LC) expansion (De Felici et al, 1993; Dolci et al, 1991; Matsui et al, 1991; Ohta et al, 2017; Fig 1A and B).

Figure 1. Exploration of a condition that expands hPGCLCs in vitro .

• A
  A scheme for hPGCLC induction and culture on the m220‐5 feeders. hPGCLCs were sorted by FACS as B LIMP1‐td T omato (BT) and TF A P2C‐E G FP (AG)‐positive (BT⁺AG⁺) cells. Scale bar, 200 μm.
• B
  (left) A scheme for the exploration of a condition for BT⁺AG⁺ cell expansion. (right) A plot showing the numbers of BT⁺AG⁺ cells after a 10‐day culture under the indicated conditions as determined by FACS analyses. The combination of chemicals (blue) [forskolin (10 μM), rolipram (10 μM), and cyclosporin A (5 μM)] and cytokines (yellow) [LIF (10 ng/ml), EGF (50 ng/ml), and basic fibroblast growth factor (bFGF: 20 ng/ml)] used for BT⁺AG⁺ cell expansion is shown at the bottom. Around 5,000 hPGCLCs were used as a starting cell population. Since cyclosporin A had a negative effect on BT⁺AG⁺ cell expansion, we performed experimental replicates excluding the conditions with cyclosporin A. Mean fold changes are shown by horizontal bars. The condition for mPGCLC expansion: green; and the condition without additional chemicals/cytokines: gray.
• C
  A scheme for the passage by FACS of BT⁺AG⁺ cells. BT⁺ and AG⁺ cells were defined as the cells within the indicated FACS gates.
• D, E
  Results of two experimental replicates for BT⁺AG⁺ cell expansion in the indicated basal media with 15% KSR, 2.5% FBS, 100 ng/ml SCF, 10 μM forskolin, and 20 ng/ml bFGF. Approximately 5,000 hPGCLCs were used as a starting cell population (culture day 0: c0), and their increases in number measured by FACS upon passages at c10, c20, and c30 were indicated as fold changes from c0.
• F
  A typical example of relief‐contrast and fluorescence (BT) images of BT⁺AG⁺ cell expansion culture from c21 to c30. Images at c21, 23, 25, 29, 29, and 30 are shown. A boxed area in the middle panel for c30 cells is magnified in the rightmost panel. Arrowheads indicate BT⁺ cells with a typical morphology. Scale bars: left, 500 μm; second right, 200 μm; right, 50 μm.

After 10 days of culture with a medium change every 2 days, we examined whether d6 hPGCLCs were expanded in number by a FACS analysis under each condition. With regard to the chemicals, we found that the addition of forskolin alone contributed to the expansion of the BT⁺AG⁺ cell number more than the addition of either forskolin plus rolipram or forskolin plus cyclosporin A (cyclosporin A had a negative effect on the BT⁺AG⁺ cell expansion; Fig 1B), and with regard to the cytokines, the addition of LIF and EGF, FGF alone, or all three cytokines had a positive effect on the expansion of BT⁺AG⁺ cells (Fig 1B). Consequently, the addition of forskolin/LIF/EGF or forskolin/bFGF resulted in a more than ~4‐fold expansion of BT⁺AG⁺ cells after 10 days of culture (Fig 1B). We decided to use the condition with forskolin/bFGF in the following experiments, since it represents a simpler strategy.

We next examined whether the composition of the basal media affected the expansion of hPGCLCs. We first examined the effect of four different basal media [GMEM and DMEM with three different concentrations of glucose (4.5, 1.0, and 0.22 g/l)] (Table EV1). We seeded ~5.0 × 10³ BT⁺AG⁺ d6 hPGCLCs on the m220 feeders in the four media, each with 15% KSR, 2.5% FBS, 100 ng/ml SCF, 10 μM forskolin, and 20 ng/ml bFGF. After 10 days of culture, we evaluated the number of BT⁺AG⁺ cells by FACS and passaged the ~5.0 × 10³ BT⁺AG⁺ cells to repeat the analysis until culture day 30 (c30) under each condition (Fig 1C). As shown in Fig 1D, in two independent experiments, although the expansion rate under each condition was somewhat variable between the experiments, DMEM with 1 g/l of glucose was the most effective in expanding the BT⁺AG⁺ cells during the 30‐day culture (up to ~80‐fold expansion). We then compared the effect of DMEM (1 g/l of glucose), GMEM, and seven different basal media (DMEM/F12, F12, Temin's MEM, αMEM, IMDM, RPMI, and CMRL) in the same manner (Table EV1) and found that DMEM (1 g/l of glucose) showed the most prominent effect on the expansion of BT⁺AG⁺ cells, while basal media such as F12 and DMEM/F12 exhibited a poor influence (Fig 1E). A superior effect of DMEM (1 g/l of glucose) on BT⁺AG⁺ cell expansion was validated in another experimental replicate comparing the effects of DMEM (1 g/l of glucose), GMEM, RPMI, and CMRL (Fig 1E).

Figure 1F shows a typical example of the culture of BT⁺AG⁺ cells (from c21 to c30) on the m220 feeders in DMEM (1 g/l of glucose) with the selected supplements (15% KSR, 2.5% FBS, 100 ng/ml SCF, 10 μM forskolin, and 20 ng/ml bFGF). During the culture, albeit at a slow rate, the BT⁺ cells (AG fluorescence was relatively weak under an inverted fluorescence microscope), which were characterized by lipid droplet‐like vesicles around the nucleus, exhibited a progressive propagation, forming colonies of distinct shape after 10 days of culture (Fig 1F). These findings indicate that d6 hPGCLCs can be propagated in culture as BT⁺AG⁺ cells and that their requirements for efficient expansion as well as their manner of expansion appear to be somewhat different from those of mPGCLCs (Ohta et al, 2017).

Persistent expansion of hPGCLCs under a defined condition

We went on to explore the extent to which hPGCLCs can be expanded as BT⁺AG⁺ cells under the selected condition [on m220 feeders in DMEM (1 g/l of glucose) with 15% KSR, 2.5% FBS, 100 ng/ml SCF, 10 μM forskolin, and 20 ng/ml bFGF]. We used d4 and d6 hPGCLCs as starting cell populations and cultured them with a passage of BT⁺AG⁺ cells every 10 days. As shown in Fig 2A and EV1A, both d4 and d6 hPGCLCs had been successfully cultured and passaged up to at least c120 as BT⁺AG⁺ cells. When we used d4 hPGCLCs as a starting cell population, the percentage of BT⁺AG⁺ cells upon passage (measured on day 10 of culture after each passage) was around ~20% for the earlier passages (~c20) and ~10% for the later passages (~c30–c120), with a progressive increase in the percentage of non‐BT⁺AG⁺ cells (up to ~95%), including BT⁺AG⁻ cells during the culture (Fig EV1A and B). On the other hand, when we used d6 hPGCLCs as a starting cell population, the percentage of BT⁺AG⁺ cells upon passage was higher, at around ~40% for the earlier passages (~c20) and ~20% for later passages (~c30–c120; Fig 2A and B), and accordingly, the percentage of non‐BT⁺AG⁺ cells was lower (~80%), with essentially no BT⁺AG⁻ or BT⁻AG⁺ cells during the culture (Fig 2A and B).

Figure 2. Long‐term expansion of hPGCLCs under a defined condition in vitro .

1. FACS plots for BTAG expression at d6 of hPGCLC induction and at c10 to c120 of BT⁺AG⁺ cell expansion culture starting from d6 hPGCLCs on the m220 feeders in DMEM (1 g/l of glucose) with 15% KSR, 2.5% FBS, 100 ng/ml SCF, 10 μM forskolin, and 20 ng/ml bFGF. The positivity for BTAG expression was defined as shown in Fig 1C.
2. The transitions of the percentages of BT⁺AG⁺, BT⁺, and AG⁺ cells during the BT⁺AG⁺ cell expansion culture until c120. The color coding is as indicated.
3. FACS analyses for the positivity for DRAQ7 incorporation (top), TRA‐1-85 expression (middle), and the ratio for TRA‐1-85⁺/ BT⁺AG⁺, TRA-1‐85⁺/ BT⁻AG⁻cells, TRA-1‐85⁻, and DRAQ7⁺ cells (bottom) among cells at c20 (left) and c60 (right).
4. Relief‐contrast (left) and fluorescence [BT (middle) and AG (right)] images of BT⁺AG⁺ cell expansion culture at c30, 70, and 120. AG fluorescence was dimmer when the AG⁺ cells were cultured on a plate in a two‐dimensional manner. Scale bar, 50 μm.
5. Growth curves of BT⁺AG⁺ cells as indicated by log₁₀(fold increases) during their expansion culture in two experimental replicates until c120. 10,000 hPGCLCs were used as a starting cell population, and 10,000 BT⁺AG⁺ cells were re‐plated at each passage.
6. Immunofluorescence (IF) analyses of the expression of BLIMP1, TFAP2C, SOX17, POU5F1, and NANOG (cyan) in AG⁺ (yellow) cells during BT⁺AG⁺ cell expansion culture at c28 [low (left) and high (middle) magnifications] and at c66 (high magnification). The cells were counterstained with DAPI (white), and merged images are shown on the right. Scale bar, 20 μm.
7. Quantitative PCR (qPCR) analyses of the expression of the indicated genes during hPGCLC induction and BT⁺AG⁺ cell expansion culture. For each gene examined, the ∆Ct from the average Ct values of the two independent housekeeping genes RPLP0 and PPIA (set as 0) was calculated and plotted for two independent experiments. Mean values are connected by a line. *: Not detected or ∆Ct < −10.

Figure EV1. Long‐term expansion of BT ⁺ AG ⁺ cells from d4 hPGCLCs.

1. FACS plots for BTAG expression at d4 of hPGCLC induction and at c10 to c120 of BT⁺AG⁺ cell expansion culture starting from d4 hPGCLCs on the m220 feeders in DMEM (1 g/l of glucose) with 15% KSR, 2.5% FBS, 100 ng/ml SCF, 10 μM forskolin, and 20 ng/ml bFGF. The positivity for BTAG expression was defined as shown in Fig 1C.
2. The transitions of the percentages of BT⁺AG⁺, BT⁺, and AG⁺ cells during the BT⁺AG⁺ cell expansion culture until c120. The color coding is as indicated.
3. FACS analysis for the BTAG fluorescence levels of the TRA-1‐85⁻ cells (m220 feeders). (left) DRAQ7⁻ cells were sorted as live cells and analyzed for TRA-1‐85 expression. (right) The BTAG fluorescence of the TRA-1‐85⁺ human cells and TRA-1‐85⁻ m220 feeders. Note that TRA-1‐85⁻ m220 feeders exhibited a weak autofluorescence, plotted diagonally beneath the BT⁺AG⁺ cells when sorted by the BTAG fluorescence.
4. Karyotype analysis by array comparative genomic hybridization (aCGH) of parental 585B1 BTAG hiPSCs (top) and BT⁺AG⁺ cells in expansion culture at c70 (bottom), indicating that the karyotypes are essentially preserved in BT⁺AG⁺ cells at c70.
5. IF analysis of the expression of GFP (AG), human mitochondrial antigen (hMcd), and SOX2 in the BT⁺AG⁺ cell expansion culture at c29. Note that the AG⁻ cells in culture expressed SOX2 and occurred alongside the AG⁺/SOX2⁻ cells. Scale bar, 50 μm.

We examined the composition of the cell populations upon passage in the d6 hPGCLC expansion culture more closely. In one experiment, at c20, BT⁺AG⁺ cells made up around ~46% of the whole cells, whereas TRA‐1‐85⁺ [a human‐specific antigen (Draper et al, 2002)], non‐BT⁺AG⁺ cells accounted for around ~13% and TRA‐1‐85⁻ cells (m220 feeders) consisted of around ~36% (dead cells: ~5%; Fig 2C). At c60, BT⁺AG⁺ cells accounted for around ~37% of the total cell population, TRA‐1‐85⁺, non‐BT⁺AG⁺ cells made up around ~38% and TRA‐1‐85⁻ cells (m220 feeders) made up around ~20% (Fig 2C). TRA‐1‐85⁻ m220 feeders exhibited a weak autofluorescence, plotted diagonally beneath the BT⁺AG⁺ cells when sorted by the BTAG fluorescence (Fig EV1C). Combined with the analyses of cell‐population transitions with regard to the BTAG expression during the whole culture period (Fig 2A and B, Fig EV1A and B), these findings indicate that under the selected condition, d6 hPGCLCs could be propagated as BT⁺AG⁺ cells in a more stable fashion than d4 hPGCLCs, and although a certain fraction of BT⁺AG⁺ cells differentiated into non‐BT⁺AG⁺ cells (see below), particularly at later passages, a majority of the d6 hPGCLC culture remained as BT⁺AG⁺ cells, with the capacity to form characteristic colonies even at c120 (Fig 2D). Cumulatively, the results showed that d6 hPGCLC‐derived BT⁺AG⁺ cells propagated nearly ~1.0 × 10⁶‐fold by c120 in two experimental replicates (Fig 2E), and a karyotype analysis by comparative genomic hybridization (CGH) at c70 verified that the expanding BT⁺AG⁺ cells were essentially karyotypically normal (Fig EV1D).

In good agreement with the positivity of BTAG expression, immunofluorescence (IF) analysis revealed that d6 hPGCLC‐derived colonies both at c28 and c66 expressed BLIMP1 and TFAP2C in a uniform manner and were also positive for SOX17, POU5F1 (OCT4), and NANOG, key markers for hPGC(LC)s (Fig 2F). We found that a majority of the AG⁻ cells in culture expressed SOX2, and these cells occurred alongside the AG⁺/SOX2⁻ cells (Fig EV1E, see below). We isolated total RNAs of BT⁺AG⁺ cells from c10 to c120 (c10, c30, c50, c70, c90, and c120) and analyzed the expression of key genes by quantitative PCR (qPCR), which showed that the BT⁺AG⁺ cells continuously expressed genes such as PRDM1, TFAP2C, SOX17, POU5F1, and NANOG at a similar level and stably repressed SOX2 up to at least c120 (Fig 2G). These findings indicate that hPGCLCs, particularly d6 hPGCLCs, can be propagated under a defined condition by ~1.0 × 10⁶‐fold or more while maintaining the properties of hPGCs.

Expansion of hPGCLCs with a surface‐marker selection

We next examined whether hPGCLCs can be propagated with a passage procedure that is easily applicable to other hPSC lines. First, based on the observation that BT⁻AG⁺ cells did not appear under our culture condition when using d6 hPGCLCs as starting cell populations (Fig 2A and B), we examined whether hPGCLCs can be propagated with a passage using only the AG reporter. We found that d6 hPGCLCs derived from 585B1 BTAG hiPSCs were propagated with a passage as AG⁺ cells up to c60 or longer (Fig 3A and B), and the resultant cell population at c60 was essentially identical with that passaged with the BTAG positivity (Fig 3C).

Figure 3. Expansion of hPGCLCs from an independent hiPSC line with a surface‐marker selection.

1. FACS plots for AG expression in cultures originating from AG⁺ cells at d6 of hPGCLC induction (from 585B1 BTAG hiPSCs) and passaged with AG positivity at c10 and c50.
2. A growth curve of AG⁺ cells as indicated by log₁₀(fold increases) during their expansion culture until c60. 10,000 AG⁺ cells at d6 of hPGCLC induction were used as a starting cell population, and 10,000 AG⁺ cells were re‐plated at each passage.
3. A FACS plot for BTAG expression of cultures as in (B) at c60.
4. FACS plots for INTEGRINα6 and EpCAM expression in cultures originating from INTEGRINα6^high/EpCAM^high cells (colored in blue) at d6 of hPGCLC induction (from 1383D6 hiPSCs) and passaged with INTEGRINα6^high/EpCAM^high expression at c10, 20, and 30. The INTEGRINα6^low/EpCAM^high cells that emerged during the culture were colored in yellow (see Fig 4).
5. Relief‐contrast images of cultures as in (D) at c10 and 30. Scale bar, 200 μm.
6. (top) The expression pattern of INTEGRINα6 and EpCAM in BT⁺AG⁺ cells (from 585B1 BTAG hiPSCs) at c40 during their expansion culture. (bottom) The expression pattern of BTAG in INTEGRINα6^high/EpCAM^high (blue) or INTEGRINα6^low/EpCAM^high (yellow) cells at c40 (see Fig EV1C).
7. Growth curves (three experimental replicates) of INTEGRINα6^high/EpCAM^high cells (from 1383D6 hiPSCs) as indicated by log₁₀(fold increases) during their expansion culture until c30. 10,000 such cells at d6 of hPGCLC induction were used as a starting cell population, and 10,000 cells were re‐plated at each passage.
8. qPCR analyses of the expression of the indicated genes during hPGCLC induction and INTEGRINα6^high/EpCAM^high cell expansion culture. For each gene examined, the ∆Ct from the average Ct values of the two independent housekeeping genes RPLP0 and PPIA (set as 0) were calculated and plotted for two independent experiments. Mean values are connected by a line. *: Not detected or ∆Ct < −10; **: Detected only in one experimental replicate.

Second, we examined whether hPGCLCs can be propagated with a passage using cell‐surface markers. We induced an independent hiPSC line, 1383D6 (XY), bearing no fluorescence reporters (Yokobayashi et al, 2017), into hPGCLCs, isolated d6 hPGCLCs with the expression of INTEGRINα6 and EpCAM (Sasaki et al, 2015; Yokobayashi et al, 2017; Fig 3D), and cultured them under our selected condition. At c10, we observed a formation of colonies of characteristic morphology (Fig 3D and E), and FACS analysis revealed that the culture was enriched with a distinct population of cells expressing high levels of INTEGRINα6 and EpCAM, with simultaneous occurrence of an INTEGRINα6^low/EpCAM^high cell population (see below for their characterization; Fig 3D). We noted that the BT⁺AG⁺ cells at c40 derived from 585B1 BTAG hiPSCs exhibited a profile of INTEGRINα6 and EpCAM expression similar to that of the INTEGRINα6^high/EpCAM^high cell population induced from 1383D6 hiPSCs, whereas, conversely, the INTEGRINα6^low/EpCAM^high cell population corresponded to the BT⁻AG⁻ cells and the INTEGRINα6⁻/EpCAM⁻ cell population consisted mainly of the m220 feeders (Fig 3F, Fig EV1C).

Accordingly, we passaged the INTEGRINα6^high/EpCAM^high cell population and found that such cells had been propagated and passaged successfully up to at least c30 with an expansion of up to ~50‐fold (Fig 3D, E and G). Moreover, a qPCR analysis revealed that the INTEGRINα6^high/EpCAM^high cells expressed key hPGC genes such as PRDM1, TFAP2C, SOX17, POU5F1, and NANOG, and repressed SOX2 in a manner similar to the BT⁺AG⁺ cells in expansion culture (Fig 3H). The somewhat lower expansion efficiency of INTEGRINα6^high/EpCAM^high cells derived from the 1383 D6 line compared with that of BT⁺AG⁺ cells derived from the 585B1 BTAG line (~50‐fold vs. ~100‐ to 200‐fold at c30) may have been related to damage on the starting cell population resulting from more complicated sorting procedures for surface markers or due to a clonal difference in the original hiPSC lines. We conclude that hPGCLCs are propagated successfully with a passage by using cell‐surface markers.

Expanded hPGCLCs retain an early hPGC(LC) transcriptome

To gain more insight into the properties of hPGCLC‐derived cells during the expansion culture, we determined the transcriptomes of the relevant cell types by an RNA‐sequence method (Nakamura et al, 2015) and analyzed their properties in comparison with those of hPGCLC‐derived cells in xrOvaries (Yamashiro et al, 2018). The cell types analyzed were as follows: hiPSCs and iMeLCs (585B1 BTAG, 1383D6), BT⁺AG⁺ cells at c10, c30, c50, c70, c90, and c120 (585B1 BTAG), INTEGRINα6^high/EpCAM^high cells at c10 and c30, INTEGRINα6^low/EpCAM^high cells at c10 (1383D6), BT⁺AG⁺ cells at aggregation culture days (ag) 7, 21, 35, 49, 63, and 77 (585B1 BTAG; Yamashiro et al, 2018), and AG⁺VT⁻/AG^+/−VT⁺/AG⁺VT⁺/AG⁻VT⁺ cells at ag120 [1390G3 AGVT (DDX4 (also known as human Vasa homologue)‐tdTomato)] (Fig EV2A, Dataset EV1, Table EV2).

Figure EV2. Transcriptome analysis and expression dynamics of key genes during hPGCLC induction, BT ⁺ AG ⁺ cell expansion, and differentiation in xrOvaries.

1. Scatter plot comparisons of the transcriptomes between the indicated replicates of BT⁺AG⁺ cells in expansion culture.
2. Expression dynamics determined by RNA‐seq analysis of key indicated genes in the indicated cell types. For each gene shown, log₂(RPM + 1) values from three replicates for hiPSCs, iMeLCs, d6 hPGCLCs, and c10‐c120 BT⁺AG⁺ cells and two replicates for ag7‐ag77 BT⁺AG⁺ cells (except for ag49 BT⁺AG⁺ cells with one replicate) are shown with the mean values connected by lines. Note that the c10 ITGA6^low cells were from 1383D6 hiPSCs, whereas all other cell types were derived from 585B1 BTAG hiPSCs.

Unsupervised hierarchical clustering revealed that the analyzed cells could be classified primarily into two large clusters, one consisting of hiPSCs/iMeLCs and the other consisting of d6 hPGCLC‐derived cells (Fig 4A). Within the large cluster of d6 hPGCLC‐derived cells, the BT⁺AG⁺ cells and INTEGRINα6^high/EpCAM^high cells in expansion culture formed a tight cluster, with those at different culture periods intermingling, and the whole cluster of such cells exhibited similarity to the cluster consisting of hPGCLC‐derived cells in xrOvaries at an early culture period (BT⁺AG⁺ cells at ag7; Yamashiro et al, 2018), as well as to the cluster of d6 hPGCLCs (Fig 4A). On the other hand, the cluster of BT⁺AG⁺/INTEGRINα6^high/EpCAM^high cells in expansion culture exhibited a relatively distant relationship with that of hPGCLC‐derived cells in xrOvaries at late culture periods (BT⁺AG⁺ cells at ag21, ag35, ag49, ag63, and ag77 and AG⁺VT⁻/AG^+/−VT⁺/AG⁺VT⁺/AG⁻VT⁺ cells at ag120), including oogonia/gonocyte‐like cells (cells at ag77 and ag120; Fig 4A; Yamashiro et al, 2018). Interestingly, we found that the INTEGRINα6^low/EpCAM^high cells at c10 were clustered with hiPSCs/iMeLCs and expressed pluripotency‐associated genes and other key genes in a manner similar to hiPSCs/iMeLCs (Figs 4A and EV2B), indicating that they represent a de‐differentiated cell population. In good agreement with these results, principal component analysis (PCA) showed that the BT⁺AG⁺/INTEGRINα6^high/EpCAM^high cells in expansion culture, irrespective of their culture periods, were plotted at positions close to that of BT⁺AG⁺ cells at ag7, which were located between the positions of d6 hPGCLCs and BT⁺AG⁺ cells at ag21 (Fig 4B).

Figure 4. Transcriptome analysis of BT ⁺ AG ⁺ cells during their expansion culture.

1. Unsupervised hierarchical clustering of transcriptomes of BT⁺AG⁺ cells during their expansion culture with relevant cell types [hiPSCs, iMeLCs, d6 hPGCLCs, and d6 hPGCLC‐derived cells cultured in xrOvaries (Yamashiro et al, 2018)]. The original hiPSC lines used and the cellular states are color‐coded as indicated. Numbers following the indications of cellular states denote replicate numbers. c: Culture days in expansion culture; ag: culture days in aggregation culture in xrOvaries.
2. Principal component analysis (PCA) of transcriptomes of the cells as in (A). The cells are plotted in the two‐dimensional plane defined by PC1 and PC2 values. The color coding is as indicated.
3. Heat map representation of the expression of genes that characterize the differentiation process of oogonia/gonocyte‐like cells from hiPSCs (cluster 1 to 5 genes; Yamashiro et al, 2018) in the cells as in (A). Representative genes and gene ontology (GO) functional term enrichments in each cluster are indicated on the right. The color coding is as indicated.
4. Box plot representation of the mean expression levels of the cluster 1 to 5 genes as in (C) in the indicated cell types. The boxes show 25^th and 75^th percentiles, and the bars represent the median.
5. Spearman's rank correlation analysis of the expression profiles of the cluster 1 to 5 genes as in (C) in the indicated cell types. The color coding is as indicated.

Next, we examined the expression of a gene set that characterizes a developmental progression from hPGC(LC)s to oogonia/gonocytes that we defined previously (Yamashiro et al, 2018) in BT⁺AG⁺/INTEGRINα6^high/EpCAM^high cells in expansion culture. As reported previously (Yamashiro et al, 2018) and shown in Fig 4C, the cluster 1 and 2 genes are those that are up‐regulated upon hPGCLC specification, with the cluster 1 genes (e.g., SOX17, PRDM1, NANOS3, KLF4, TCL1A) being expressed continuously thereafter and the cluster 2 genes (e.g., HLA‐DQB1, HLA‐DPA1, MT2A, TRPC5, TRPC6) being repressed progressively after around ag35, the cluster 3 genes are those that are up‐regulated specifically after around ag35 and characterize oogonia/gonocyte development (e.g., DAZL, DDX4, MAEL, TDRD9, PIWIL1), and the cluster 4 and 5 genes are those that are down‐regulated after hPGCLC specification, with the cluster 4 genes (e.g., SOX2, TDGF1, GJA1, OTX2, CDH1) being repressed relatively quickly and the cluster 5 genes (e.g., IFITM2, IFITM3, SLC2A3, SLC2A1) being repressed more gradually. As is evident from Fig 4C and D, the BT⁺AG⁺/INTEGRINα6^high/EpCAM^high cells in expansion culture continued to express the cluster 1, 2, and 5 genes, repressed the cluster 4 genes in a manner similar to d6 hPGCLCs/ BT⁺AG⁺ cells at ag7, and did not up‐regulate the cluster 3 genes except such genes as BRDT, IRX4, and IL12B, exhibiting a property highly similar to that of BT⁺AG⁺ cells at ag7. Consistently, Spearman's rank correlation analysis using this gene set revealed that the transcriptomes of the BT⁺AG⁺/INTEGRINα6^high/EpCAM^high cells in expansion culture at different culture periods were highly correlated and showed the closest correlation with those of BT⁺AG⁺ cells at ag7 and then at ag21 (Fig 4E).

We determined differentially expressed genes [log₂ (RPM + 1) ≧4, log₂ (fold change) ≧1] between BT⁺AG⁺ cells at different culture periods in expansion culture ([Link], [Link]). The number of differentially expressed genes was highest between d6 hPGCLCs and c10 BT⁺AG⁺ cells (377 and 548 genes were up/down‐regulated, respectively, in c10 BT⁺AG⁺ cells), decreased substantially between c10 and c30 BT⁺AG⁺ cells (72 and 56 genes were up/down‐regulated, respectively, in c30 BT⁺AG⁺ cells), and the BT⁺AG⁺ cells after c30 exhibited an almost identical transcriptome, except for a few genes including TCL1A, which was progressively down‐regulated during the expansion culture (Fig 5A, Figs EV2B and EV3, and see below). TCL1A is known to act as a co‐activator of AKT and to promote glycolysis in PSCs (Laine et al, 2000; Nishimura et al, 2017), suggesting the possibility that the AKT and glycolytic pathways are progressively down‐regulated during the expansion culture.

Figure 5. Differentially expressed genes among BT ⁺ AG ⁺ cells during their expansion culture.

1. Scatter plot comparisons of the transcriptomes between the indicated cell types (average values of three replicates) during BT⁺AG⁺ cell expansion from d6 hPGCLCs. The up‐ and down‐regulated genes [log₂ (RPM + 1) ≧ 4, log₂ (fold change) ≧ 1] in the cell types on the y‐axes are colored in red and blue, respectively.
2. GO enrichments and representative genes in differentially expressed genes between d6 hPGCLCs and c10 BT⁺AG⁺ cells. The color coding is as indicated.
3. (left) Scatter plot comparisons of the transcriptomes between c30 and ag7 (top) or ag21 (bottom) BT⁺AG⁺ cells. The color coding is as in (A). (right) GO enrichments and representative genes in differentially expressed genes between c30 and ag7 (top) or ag21 (bottom) BT⁺AG⁺ cells. The color coding is as indicated.

Figure EV3. Transcriptome analysis of BT ⁺ AG ⁺ cells during their expansion culture.

GO enrichments and representative genes in differentially expressed genes between c10 and c30 BT ⁺ AG ⁺ cells. The color coding is as indicated.

The genes up‐regulated in c10 BT⁺AG⁺ cells compared with d6 hPGCLCs were enriched in gene ontology (GO) functional terms such as “negative regulation of growth” and “mineral absorption”, and included metallothionein (MT) 1E, 1F, 1G, 1H, 1X, and 2A, and superoxide dismutase (SOD) 1 and 2, which are known to function for the homeostasis of essential heavy metals such as zinc or to remove reactive oxygen species (ROS; Fig 5B, Table EV3; Haq et al, 2003). We noted that the MT genes were also up‐regulated in BT⁺AG⁺ cells at ag7 in xrOvaries and were down‐regulated thereafter (Fig EV2B). In contrast, the genes down‐regulated in c10 BT⁺AG⁺ cells were enriched in GO terms such as “heart development” (e.g., GATA2, GATA3, HAND1, ID1, ID3, SALL1), “kidney development” (e.g., NPHP3, LZTS2, LHX1, ARID5B, OVOL1), and “extracellular matrix organization” (e.g., COL2A1, COL3A1, COL4A5, LAMA4, LAMB2; Fig 5B, Table EV3), suggesting that genes associated with a somatic program concomitantly activated upon hPGCLC specification are subject to repression during the hPGCLC expansion culture.

To gain insight into a potential mechanism that distinguishes the developmental pathways of BT⁺AG⁺ cells in expansion culture and in xrOvaries, we explored differentially expressed genes [log₂ (RPM + 1) ≧4, fold change ≧3] between c30 BT⁺AG⁺ cells and ag7/ag21 BT⁺AG⁺ cells (Fig 5C, [Link], [Link]). The genes up‐regulated in c30 BT⁺AG⁺ cells compared with ag7 BT⁺AG⁺ cells (221 genes) included CENPN/C/K, CETN3, CDC7/26, CCNA2, and E2F3, and were enriched in GO terms such as “CENP‐A containing nucleosome assembly” and “mitotic nuclear division” (Fig 5C, Table EV5), consistent with the notion that BT⁺AG⁺ cells in expansion culture are mitotically more active than those in xrOvaries. The genes up‐regulated in c30 BT⁺AG⁺ cells compared with ag21 BT⁺AG⁺ cells (162 genes) additionally included the MT gene family members, reflecting their down‐regulation in ag21 BT⁺AG⁺ cells (Fig 5C, Table EV6). In contrast, the genes up‐regulated in ag7/21 BT⁺AG⁺ cells compared with c30 BT⁺AG⁺ cells (101 and 70 genes, respectively) included FOS, FOSB, EGR1, EGR2, and EGR3 (Fig 5C, [Link], [Link]), which are immediate‐early genes in response to a variety of cellular stimuli, suggesting that the xrOvary environment elicits a distinct signaling pathway(s) in BT⁺AG⁺ cells. Collectively, these findings indicate that d6 hPGCLC‐derived BT⁺AG⁺/INTEGRINα6^high/EpCAM^high cells in expansion culture take on a developmental pathway toward oogonia/gonocyte‐like cells, but halt their developmental progression at an early stage, faithfully maintaining the properties of early hPGCs, although the up‐regulation of some of the genes during the culture may represent a culture adaptation.

Expanded hPGCLCs preserve an early hPGC(LC) DNA methylome

We next determined the genome‐wide DNA methylation profiles of BT⁺AG⁺ cells at c10, c70, and c120 in expansion culture by whole‐genome bisulfite sequencing (Miura et al, 2012; Shirane et al, 2016), and examined their properties in comparison with those of other relevant cell types, including hiPSCs, iMeLCs, d6 hPGCLCs, and hPGCLC‐derived cells in xrOvaries (BT⁺AG⁺ cells at ag35 and ag77 and AG⁺VT⁻/AG⁺VT⁺/AG⁻VT⁺ cells at ag120; Yamashiro et al, 2018; Fig EV4A, Table EV7). We reported previously that hiPSCs and iMeLCs exhibited very similar genome‐wide 5‐methylcytosine (5mC) profiles, with an average 5mC level of ~80%, and d6 hPGCLCs underwent a slight, but significant, reduction in the global 5mC levels (~75% in average; Yamashiro et al, 2018; Fig 6A and B, Table EV7). We found that the BT⁺AG⁺ cells in expansion culture exhibited a further reduction in the 5mC levels from d6 hPGCLCs, acquiring an average 5mC level of ~65% at c10 (Figs 6A and B, and EV4A, Table EV7). However, thereafter, they essentially maintained their genome‐wide 5mC levels and profiles at c70 and c120, with a minor increase at c120 (~68%; Fig 6A and B, Fig EV4A, Table EV7). Accordingly, examination of the 5mC levels in representative elements across the genome, including promoters, exons, introns, intergenic regions, representative repeat elements, non‐promoter CpG islands (CGIs), and imprint control regions (ICRs), revealed that all these elements essentially retained their 5mC levels during the expansion culture (Fig 6C–E, Table EV8). These observations are in good agreement with the finding that the BT⁺AG⁺ cells essentially retain the transcriptome of early hPGCs during their expansion culture (Fig 4) and are in sharp contrast to the findings that d6 hPGCLCs differentiate into oogonia/gonocytes and undergo genome‐wide DNA demethylation in xrOvaries (Yamashiro et al, 2018; Fig 6A) and that mPGCLCs undergo genome‐wide DNA demethylation during their expansion culture (Shirane et al, 2016; Ohta et al, 2017).

Figure EV4. Genome‐wide DNA methylation profiles of BT ⁺ AG ⁺ cells during their expansion culture.

1. Scatter plot comparisons of the DNA methylation profiles between the indicated replicates of d6 hPGCLCs/BT⁺AG⁺ cells in expansion culture.
2. DNA methylation profiles around the MT1 gene cluster in cells during hPGCLC induction, expansion, and differentiation in xrOvaries.
3. Expression dynamics determined by RNA‐seq analysis of TCL1A in the indicated cell types as also shown in Fig EV2 (top) and the 5mC levels in the promoter of TCL1A in d6 PGCLCs (three replicates) and BT+AG+ cells at c10, c70, and c120 (two replicates each; bottom).
4. Heat map representation of the indicated imprinted gene expression under the control of the imprint control regions (ICRs) of H19 (chromosomal location: 11p15), IG‐DMR (14q32), and IGF 1R (15q26.3) in the indicated cell types. Numbers following the indications of cellular states are the numbers of replicates. The color coding is as indicated. Note that the expression level of H19 in d6c120 cells was significantly up‐regulated compared with that in d6 hPGCLCs (Welch's t‐test, P = 0.009). Imprinted genes were obtained from the Catalogue of Parent of Origin Effects website (http://igc.otago.ac.nz/home.html; Morison et al, 2005).

Figure 6. Genome‐wide DNA methylation profiles of BT ⁺ AG ⁺ cells during their expansion culture.

• A
  Violin plot representation of the genome‐wide DNA methylation levels [5‐methylcytosine (5mC) levels] determined by whole‐genome bisulfite sequence analysis in BT⁺AG⁺ cells during their expansion culture and the relevant cell types indicated. The mean levels are indicated by red bars.
• B
  Scatter plot comparisons, combined with histogram representations (top and right of scatter plots) of the genome‐wide 5mC levels (genome‐wide 2‐kb windows) between the indicated cell types.
• C
  Heat map representation showing the 5mC levels in the indicated genomic elements on the autosomes in the indicated cell types. HCP/ICP/LCP: high/intermediate/low-CpG promoters. The color coding is as indicated.
• D
  Heat map representation showing the 5mC levels in the indicated repeat elements on the autosomes in the indicated cells. Repeat elements whose repStart ≤ 5 and repEnd > 5 kb for LINE, ERVK, ERVL, and ERV1 and repStart ≤ 5 and repEnd ≥ 310 for SINE were used to exclude truncated elements.
• E
  Heat map representation showing the 5mC levels in the differentially methylated regions of the indicated imprinted genes in the indicated cells.
• F
  IF analysis of the expression of DNMT1 or UHRF1 in hiPSCs and c66 BT⁺AG⁺ cells. hiPSCs were co‐stained with NANOG; c66 BT⁺AG⁺ cells were co‐stained with GFP (AG); both cell types were counterstained with DAPI. Bottom panels are magnifications of the top panels. Scale bar, 20 μm.
• G–I
  Violin plot representations of the genome‐wide 5mC levels in the CpG, CpA, CpT, and CpC sequences in mouse germ cells in vivo (G), in cells during mPGCLC induction and expansion in vitro (H), and in cells during hPGCLC induction, expansion, and differentiation in xrOvaries in vitro (I). The median levels are indicated by red bars. In (G), E: embryonic day; F: female; M: male. In (I), numbers following the indications of cellular states denote replicate numbers. c: Culture days in expansion culture; ag: culture days in aggregation culture in xrOvaries.

Interestingly, we noted that the locus for the MT1 gene cluster, which showed a characteristic up‐regulation in c10 BT⁺AG⁺ cells compared with d6 hPGCLCs (Figs 5A and B, and EV2B), exhibited a uniquely under‐methylated state throughout the hPGCLC induction/differentiation/expansion process (Fig EV4B), and thus, the up‐regulation of the MT genes during the expansion culture was not a consequence of the demethylation of their promoters. In contrast, we found that the promoter of TCL1A, which was among a few genes that showed a progressive repression during the BT⁺AG⁺ cell expansion culture, regained methylation during the culture (Figs 5A and B, and EV4C). Additionally, we noted that the ICRs of IG‐DMR, H19, and IGF1R were slightly demethylated during the BT⁺AG⁺ cell expansion culture (Fig 6E), but this did not appear to affect the expression of the imprinted genes under their control, except for the H19 gene (Fig EV4D).

To gain a mechanistic insight into the regulation of global DNA methylation profiles in BT⁺AG⁺ cells in expansion culture, we first looked at the expression dynamics of genes relevant for DNA methylation/demethylation in these cells. Among the de novo DNA methyltransferases (DNMTs), DNMT3A showed a low expression level [log₂(RPM + 1) ≅ 5, i.e., ~20 copies/cell (Nakamura et al, 2015)] in BT⁺AG⁺ cells in expansion culture as well as in hiPSCs, iMeLCs, and BT⁺AG⁺ cells in xrOvaries, whereas DNMT3B exhibited an acute repression upon hPGCLC specification and showed a low expression level [log₂(RPM + 1) ≅ 5] both in BT⁺AG⁺ cells in expansion culture and in xrOvaries (Fig EV2B). DMNT3L did not show a significant expression in any of the cells examined (Fig EV2B). Among the genes for DNA methylation maintenance, DNMT1 was expressed at a significant level [log₂(RPM + 1) ≅ 8] in all cell types analyzed, whereas, notably, UHRF1, which encodes a key protein for the recruitment of DNMT1 into replication foci (Bostick et al, 2007; Sharif et al, 2007), exhibited a progressive down‐regulation in BT⁺AG⁺ cells in xrOvaries (Fig EV2B; Yamashiro et al, 2018), but continued to be expressed at a level of d6 hPGCLCs/BT⁺AG⁺ cells at ag7 [log₂(RPM + 1) ≅ 5] in BT⁺AG⁺ cells throughout the expansion culture period (Fig EV2B). Among the genes for DNA demethylation, TET1 was expressed at a substantial level, whereas TET2 and TET3 showed only very low‐level expression, in all cell types analyzed (Fig EV2B). These findings suggest that a maintenance of the UHRF1 expression level is a key distinguishing character of BT⁺AG⁺ cells in expansion culture from BT⁺AG⁺ cells in xrOvaries and may underlie the maintenance of global 5mC levels in BT⁺AG⁺ cells in expansion culture.

To explore this point further, we analyzed the expression of DNMT1 and UHRF1 proteins in BT⁺AG⁺ cells in expansion culture by an IF analysis. As shown in Fig 6F, hiPSCs exhibited strong and relatively uniform expression of both DNMT1 and UHRF1 in their nuclei. In some nuclei, DNMT1 exhibited clear punctate localizations, which would correspond to the replication foci (Fig 6F; Leonhardt et al, 2000; O'Keefe et al, 1992). On the other hand, BT⁺AG⁺ cells at c66 robustly expressed DNMT1, but we did not detect its punctate nuclear localizations, presumably due to the slower proliferation kinetics of BT⁺AG⁺ cells compared with hiPSCs. We found that BT⁺AG⁺ cells expressed UHRF1 somewhat weakly and in a heterogeneous fashion; i.e., around half of them expressed UHRF1 in their nuclei, while the others appeared to show little expression of UHRF1 (Fig 6F). We did not detect UHRF1 in the cytoplasm either in hiPSCs or in BT⁺AG⁺ cells at c66 (Fig 6F). These findings led us to conclude that, albeit at a low level and in a somewhat heterogeneous fashion, BT⁺AG⁺ cells in expansion culture do express UHRF1.

We next analyzed the dynamics of the methylation levels of non‐CpG sites (CpH: CpA, CpT, and CpC) during the BT⁺AG⁺ cell expansion culture as well as in other relevant contexts including mPGCLC induction and expansion, since such methylation levels serve as an appropriate index for the de novo DNMT activities in given cells (Ramsahoye et al, 2000; Liao et al, 2015). In accord with the previous findings (Seisenberger et al, 2012; Kobayashi et al, 2013; Kubo et al, 2015), our re‐analysis revealed that mouse germ cells at embryonic days (E) 10.5 and 13.5 during the erasure of genome‐wide CpG methylation exhibited low CpH methylation levels (< 1% on average), and subsequently, male, but not female, germ cells acquired high CpH methylation levels, particularly in the case of methyl CpAs (mCpAs; > 2.5% on average), at E16.5, which accompanied the acquisition of genome‐wide androgenetic CpG methylation profiles (Kobayashi et al, 2013; Kubo et al, 2015; Seisenberger et al, 2012; Fig 6G, Table EV7). In the mPGCLC induction/expansion system, mESCs cultured under a “ground state” condition, in which de novo DNMT activities are known to be repressed (Habibi et al, 2013; Yamaji et al, 2013; Shirane et al, 2016), exhibited low CpH methylation levels (on average < 1%), whereas EpiLCs, which up‐regulate de novo DNMT activities (Yamaji et al, 2013; Shirane et al, 2016), substantially elevated mCpAs (~2% on average; Fig 6H, Table EV7). The mCpA levels in mPGCLCs, in which the expression of de novo DNMTs is strongly repressed (Hayashi et al, 2011; Shirane et al, 2016; Ohta et al, 2017), were reduced rapidly to a basal level (< 1% on average; Fig 6H, Table EV7). These findings validate the suitability of mCpH as an index for the de novo DNMT activity. We found that hiPSCs, which are known to have a gene expression property homologous to that of EpiLCs (Nakamura et al, 2016), showed a high mCpA level (~2% on average), which was slightly reduced in iMeLCs (~1.5% on average), and then further reduced to a basal level (< 1% on average) in d6 hPGCLCs, BT⁺AG⁺ cells in expansion culture and BT⁺AG⁺ cells in xrOvaries (Fig 6I, Table EV7). Collectively, these findings support the notion that BT⁺AG⁺ cells in expansion culture show reduced de novo DNMT activity, but retain maintenance DNMT activity, which accounts for the preservation of genome‐wide 5mC profiles of early hPGC(LC)s in BT⁺AG⁺ cells in expansion culture.

Expanded hPGCLCs undergo epigenetic reprogramming to differentiate into oogonia/gonocytes in xrOvaries

We next evaluated whether the BT⁺AG⁺ cells in expansion culture bear a differentiation capacity in the manner of human germ cells. For this purpose, we induced 585B1 BTAG hiPSCs into hPGCLCs, cultured d6 hPGCLCs for 30 days, and generated xrOvaries using BT⁺AG⁺ cells at c30 (5,000 cells) and mouse embryonic ovarian somatic cells at E12.5 (75,000 cells; Yamashiro et al, 2018; Fig 7A). The xrOvaries with c30 BT⁺AG⁺ cells developed in a similar fashion to those with d6 hPGCLCs (Fig 7B), and FACS analyses revealed that in xrOvaries, c30 BT⁺AG⁺ cells exhibited a tendency for a better survival/proliferation as BT⁺AG⁺ cells than d6 hPGCLCs both at ag35 and ag77, albeit the numbers of surviving BT⁺AG⁺ cells were variable among xrOvaries [per xrOvary: ag35: ~149–984; c30ag35: ~189–2,267; ag77: ~0–4,942; c30ag77: ~15–7,322] (Fig 7C). Accordingly, IF analyses at ag77 showed that similarly to d6 hPGCLC‐derived AG⁺ cells, c30 BT⁺AG⁺ cell‐derived AG⁺ cells exhibited a characteristic faint DAPI staining, were positive for SOX17, TFAP2C, and a human mitochondria antigen, and were delineated by FOXL2⁺ mouse granulosa cells (Fig 7D). Importantly, we found that also similarly to d6 hPGCLC‐derived AG⁺ cells, many of such c30 BT⁺AG⁺ cell‐derived AG⁺ cells became positive for DAZL and DDX4, key markers for oogonia/gonocytes (Fig 7D; Gkountela et al, 2015; Guo et al, 2015; Li et al, 2017; Tang et al, 2015; Yamashiro et al, 2018).

Figure 7. Expanded BT ⁺ AG ⁺ cells differentiate into oogonia/gonocyte‐like cells in xrOvaries.

• A
  Schemes for generating xrOvaries with d6 hPGCLCs (top) or d6c30 BT⁺AG⁺ cells (bottom).
• B
  Bright‐field images and FACS plots for BTAG expression in cells of xrOvaries at ag77 with d6 hPGCLCs (top) or d6c30 BT⁺AG⁺ cells (bottom). Scale bar, 200 μm.
• C
  Numbers of BT⁺AG⁺ cells in xrOvaries at ag35 and ag77 derived from d6 hPGCLCs or d6c30 BT⁺AG⁺ cells, as indicated. P‐values were calculated with Welch's t‐test.
• D
  IF analyses of the expression of TFAP2C, FOXL2, SOX17, DAZL, DDX4, and human mitochondria antigen (hMcd; cyan) in AG⁺ (yellow) cells/mouse ovarian somatic cells in xrOvaries at ag77 with d6 hPGCLCs (left) or d6c30 BT⁺AG⁺ cells (right). The cells were counterstained with DAPI (white), and merged images are shown on the right. Scale bar, 20 μm.
• E
  qPCR analyses of the expression of the indicated genes in BT⁺AG⁺ cells isolated from xrOvaries at ag7, ag35, and ag77 generated with d6 hPGCLCs (black) or d6c30 BT⁺AG⁺ cells (red). For each gene examined, the ∆Ct from the average Ct values of the two independent housekeeping genes RPLP0 and PPIA (set as 0) was calculated and plotted for two independent experiments. Mean values are connected by a line. A dot without a mean bar indicates that the gene expression was detected only in one of the two replicates. *: Not detected.
• F
  The 5mC levels of the promoters of the indicated genes in the indicated cell types. The numbers following the indications of cellular states are the numbers of replicates. Note that the mean promoter methylation levels of DPPA3 and PIWIL2 in d6c10 cells were reduced significantly compared with those in d6 hPGCLCs (Welch's t‐test, P < 0.05).
• G
  Box plot representation of the 5mC levels of the promoters of the cluster 3 genes as in (Fig 4C) in the indicated cell types. The boxes show the 25^th and 75^th percentiles, and the bars represent the median. Numbers following the indications of cellular states are the numbers of replicates. Note that the mean promoter methylation levels in d6c10 cells were reduced significantly compared with those in d6 hPGCLCs (Wilcoxon rank‐sum test, P < 0.05).
• H
  PCA of the transcriptomes of BT⁺AG⁺ cells during their expansion culture and differentiation in xrOvaries with the relevant cell types as indicated. Asterisks (*) denote RNA‐seq data for the ag35 and ag77 cells generated in this study. In PCA, blue and red arrowheads indicate the data for d6 hPGCLC‐derived and c30 BT⁺AG⁺ cell‐derived ag35/77 BT⁺AG⁺ cells, respectively. The color coding is as indicated.
• I
  Heat map representation of the expression of cluster 1 to 5 genes in Fig 4C in BT⁺AG⁺ cells isolated from xrOvaries at ag35 and ag77 generated with d6 hPGCLCs or d6c30 BT⁺AG⁺ cells. The color coding is as indicated.
• J
  Violin plot representation of the mean expression levels of the cluster 1 to 5 genes as in (I) in the indicated cell types. The boxes show 25^th and 75^th percentiles, and the bars represent the median. Note that the cluster 3 genes show a significant up‐regulation, and the cluster 4 and 5 genes show a significant down‐regulation in d6c30 BT⁺AG⁺ cell‐derived cells compared with d6 hPGCLC‐derived cells (Wilcoxon rank‐sum test, P < 0.05).
• K
  (left) Violin plot representation of the genome‐wide 5mC levels determined by whole‐genome bisulfite sequence analysis in the cell types indicated. The mean levels are indicated by red bars. (right) Scatter plot comparisons, combined with histogram representations (top and right of scatter plots), of the genome‐wide 5mC levels (genome‐wide 2‐kb windows) between the indicated cell types.
• L, M
  Heat map representation showing the 5mC levels in the indicated genomic elements on the autosomes (L) and in the differentially methylated regions of the indicated imprinted genes (M) in the indicated cells. The color coding is as indicated.

We determined the gene expression of c30 BT⁺AG⁺ cell‐derived BT⁺AG⁺ cells at ag7, ag35, and ag77 (c30ag7, c30ag35, and c30ag77 cells) by qPCR (c30ag7/c30ag35/c30ag77) or RNA‐seq analyses (c30ag35/c30ag77; Dataset EV1, Table EV2). As shown in Fig 7E, such cells continued to express markers for hPGCs, including PRDM1, TFAP2C, SOX17, POU5F1, and NANOG, and like d6 hPGCLC‐derived cells, repressed UHRF1 progressively, while maintaining DNMT1. Interestingly, however, we found that, compared with d6 hPGCLC‐derived cells, they up‐regulated genes for oogonia/gonocytes, including DPPA3, PRAME, PIWIL2, DAZL, and DDX4, more rapidly: They up‐regulated these genes as early as ag7, while by ag77, d6 hPGCLC‐derived cells caught up with c30 BT⁺AG⁺ cell‐derived cells with regard to the expression levels of these genes (Fig 7E). Notably, we found that, compared to d6 hPGCLCs, BT⁺AG⁺ cells in expansion culture reduced 5mC levels of the promoters for DPPA3, PIWIL2, and PRAME by a substantial degree (~52% for DPPA3; ~19% for PRAME; ~16% for PIWIL2) and for DAZL and DDX4 by a genome‐wide average (~10%; Fig 7F). Indeed, the promoters of the cluster 3 genes, which were highly methylated in d6 hPGCLCs, showed substantial demethylation (~10%) during the first 10 days of the expansion culture (Fig 7G, Fig EV5A).

Figure EV5. BT ⁺ AG ⁺ cells in expansion culture differentiate into oogonia/gonocytes in xrOvaries.

1. Box plot representation of the promoter 5mC levels of the cluster 1‐5 genes in Fig 4C in the indicated cell types. The boxes show 25^th and 75^th percentiles, and the bars represent the median. Numbers following the indications of cellular states denote replicate numbers.
2. Unsupervised hierarchical clustering of the transcriptomes of BT⁺AG⁺ cells during their expansion culture and differentiation in xrOvaries with the relevant cell types as indicated. The Euclidean distance and ward agglomeration method were used. Asterisks (*) denote RNA‐seq data for the ag35 and ag77 cells generated in this study. The color coding is as indicated.
3. Heat map representation showing the 5mC levels in the indicated repeat elements on the autosomes in the indicated cell types. Repeat elements whose repStart ≤ 5 and repEnd > 5 kb for LINE, ERVK, ERVL, and ERV1 and repStart ≤ 5 and repEnd ≥ 310 for SINE were used to exclude truncated elements.
4. (top) Upset plot showing the numbers of the DNA demethylation “escapees” (see Materials and Methods) and their intersections among the indicated cell types. hGC: human germ cells at weeks 7‐9 in (Tang et al, 2015). (bottom) Enrichments of the indicated genomic elements in the DNA demethylation “escapees” in the indicated cell types. TSS: transcription start sites; UTR: untranslated regions; TTS: transcription termination sites.
5. Scatter plot comparisons, combined with histogram representations (top and right of scatter plots), of the genome‐wide 5mC levels (genome‐wide 2‐kb windows) between d12 hPGCLCs reported by von Meyenn et al (2016) and the indicated cell types (Tang et al, 2015; Yamashiro et al, 2018; This study). Wk: developmental week. The sexes of the embryos are shown after the indications of the developmental week. F: female; M: male.

Unsupervised hierarchical clustering of the transcriptomes of relevant cell types revealed that, unlike the BT⁺AG⁺ cells in expansion culture, both c30ag35 and c30ag77 cells were classified into a cluster of the oogonia/gonocyte‐like cells, with c30ag35 and c30ag77 cells clustering with early‐ and late‐stage oogonia/gonocyte‐like cells, respectively (Fig EV5B), and PCA and the expression profiles of a gene set that characterizes a developmental progression from hPGC(LC)s to oogonia/gonocytes provided a consistent outcome (Fig 7H). Notably, consistent with the qPCR analysis (Fig 7E), compared with d6 hPGCLC‐derived ag35 and ag77 cells (ag35 and ag77 cells), c30ag35 and c30ag77 cells exhibited a more advanced property, e.g., they down‐regulated the cluster 2 genes and up‐regulated the cluster 3 genes in a more significant fashion (Fig 7I and J). These findings are consistent with the notion that, compared with d6 hPGCLCs, BT⁺AG⁺ cells in expansion culture bear developmentally more advanced properties, responding to signals/environments provided by xrOvaries more rapidly to differentiate into oogonia/gonocyte‐like cells.

We performed a whole‐genome bisulfite sequence analysis of the c30ag77 cells (Table EV7), which revealed that they underwent a substantial genome‐wide DNA demethylation, reaching the genome‐wide 5mC level of ~15% (Fig 7K), a value equivalent to that of human oogonia/gonocytes at the 7−10 week of development (Guo et al, 2015; Tang et al, 2015) and of ag120 cells (Yamashiro et al, 2018). Consequently, the c30ag77 cells erased 5mCs from across the genome, including promoters, exons, introns, intergenic regions, representative repeat elements, non‐promoter CGIs, and ICRs (Fig 7L and M, EV5C), acquiring an epigenetic “naïve” state of the human germ cell lineage. We analyzed the “escapees” that evade the DNA demethylation (5mC > 20%) in ag77 cells, c30ag77 cells, and ag120 AG⁺VT⁺ cells in comparison with those in human germ cells in vivo (Tang et al, 2015; Table EV9), and found that the vast majority of the escapees were overlapped among these cells, with their numbers correlating with the remaining 5mC levels in these cell types (Fig EV5D), indicating that d6 hPGCLC/c30 BT⁺AG⁺ cell‐derived cells in xrOvaries erase their genome‐wide DNA methylation in a manner similar to that of human germ cells in vivo. Accordingly, the escapees were enriched around the repeat elements such as SVA, ERVK, and ERV1, but were depleted around other key genomic elements (Fig EV5D). Taken together, these findings demonstrate that the BT⁺AG⁺ cells in expansion culture fully bear the capacity to differentiate as human germ cells, and therefore, we conclude that the culture system we have developed amplifies hPGCLCs as bona fide hPGCs in vitro.

Discussion

We have shown that hPGCLCs can be propagated by a magnitude of at least ~1 × 10⁶ fold (~20 doublings) during a period of ~4 months with the maintenance of early hPGC properties in the presence of SCF, bFGF, and forskolin, a chemical that stimulates the production of intracellular cAMPs. The BT⁺AG⁺ cells in expansion culture up‐regulate metallothionein gene families, down‐regulate genes associated with a somatic program, and show a close similarity in transcriptome to the BT⁺AG⁺ cells at ag7 in xrOvaries. Although the transcriptome of hPGCs around their specification has not been determined due to technical/ethical difficulties, the occurrence of a similar state in two different culture systems, i.e., expansion culture and xrOvaries, would support the notion that it represents a state that hPGCs acquire soon after their specification. Alternatively, given that the metallothionein gene families are activated in response to various stimuli, including cytokines (Haq et al, 2003), this state may represent an adaptation in expansion culture, which wanes during the differentiation of hPGCLCs into oogonia in xrOvaries.

Our findings in regard to hPGCLC expansion are highly divergent from those in regard to mPGCLC expansion in several respects, although the effects of the different culture conditions between hPGCLCs and mPGCLCs on such divergence warrant investigation. First, hPGCLCs exhibit a much greater level of expansion (at least ~1 × 10⁶ fold) and much longer duration of culture period (at least ~4 months) than mPGCLCs, which propagate up to ~50‐fold during a period of ~1 week and are refractory, if not impossible, to passage (Ohta et al, 2017). This could be a reflection of fundamental differences in intrinsic properties between human and mouse embryonic germ cells: Human germ cells have been reported to increase their numbers up to ~7,000,000 until ~14−19 weeks of development in females (germ cells are either oogonia or oocytes in the first meiotic prophase; Baker, 1963; Kurilo, 1981; Mamsen et al, 2011) and up to ~2,000,000 until ~19 weeks of development in males (germ cells are either gonocytes or fetal spermatogonia; Fukuda et al, 1975; Mamsen et al, 2011), whereas mouse germ cells propagate in number up to ~25,000 until E13.5 both in females (germ cells are either oogonia or oocytes in the first meiotic prophase) and in males (germ cells are pro‐spermatogonia; Kagiwada et al, 2013; Tam & Snow, 1981). Thus, there could exist a mechanism in human cells that allows their long‐term propagation and/or in mouse cells that restrict the numbers of their mitotic cell cycles.

Second, the genetic program for hPGCLC specification can be dissociable from that for epigenetic reprogramming, including genome‐wide DNA demethylation. We have shown that mPGCLCs activate a program for epigenetic reprogramming, including genome‐wide DNA demethylation, upon their specification, and compared with EpiLCs (global 5mC level: ~79%), d6 mPGCLCs exhibit a large‐scale reorganization of chromatin signatures and undergo a genome‐wide DNA demethylation (global 5mC level: ~39%; Shirane et al, 2016). Upon expansion culture, while maintaining their chromatin signatures in a relatively stable fashion, mPGCLCs continue the genome‐wide DNA demethylation most likely in a replication‐coupled, passive manner, acquiring a “naïve” epigenome of the germ line (global 5mC level: ~6%) within a week of culture (Ohta et al, 2017). These findings indicate that the genetic program for mPGCLC specification is tightly linked with and sufficient to induce epigenetic reprogramming. The acquisition of a “naïve” epigenome in cultured mPGCLCs might be a mechanism that restricts their further propagation. In contrast, hPGCLCs continue to propagate without altering their genome‐wide DNA methylation levels and profiles for up to at least ~4 months (global 5mC level: ~68%; the histone modification states during human germ cell development have not been reported), although some unique elements gain or lose methylation during the hPGCLC expansion. Thus, an additional mechanism(s) should be required for the induction of epigenetic reprogramming in hPGC(LC)s. We have provided evidence that both BT⁺AG⁺ cells in expansion culture and BT⁺AG⁺ cells in xrOvaries repress de novo DNMT activity, whereas the former, unlike the latter and mPGCLCs, failed to fully repress UHRF1, a key machinery for maintenance DNMT activity, which would explain why the former maintains its global DNA methylation levels and profiles. The examination of the effects of inducible knockout/knockdown of UHRF1 during the hPGCLC expansion culture would be an interesting experimental option, and the identification of signals/mechanisms that induce epigenetic reprogramming in BT⁺AG⁺ cells in expansion culture warrants future investigation.

von Meyenn et al (2016) reported an hPGCLC induction system from “naïve” hESCs. In their study, naïve hESCs cultured under a distinct condition (Takashima et al, 2014) were induced into human EpiLCs for 4 days, which were then induced into hPGCLCs for 12 days. The authors showed that hPGCLCs induced under this condition underwent genome‐wide DNA demethylation, albeit only partially, acquiring the genome‐wide 5mC levels of ~55% at d12 (von Meyenn et al, 2016). Although interesting, these data would need to be evaluated cautiously, since the sorted KIT⁺ hPGCLC population exhibited somewhat aberrant gene expression and might not represent hPGCLCs [e.g., compared with hPGCs, the levels of BLIMP1, NANOG, and NANOS3 in hPGCLCs were ~1/4, ~1/4, and ~1/8, respectively, and the hPGCLCs expressed DNMT3B strongly (von Meyenn et al, 2016)]. Accordingly, the genome‐wide DNA methylation profiles of hPGCLCs by von Meyenn et al exhibited poor correlation with those of human germ cells in vivo and of hPGCLCs and their progenies reported here (Fig EV5E). The culture condition for capturing naïve pluripotency in humans is still not optimal (Takashima et al, 2014; Theunissen et al, 2014), and “naïve” hESCs exhibited aberrant genome‐wide DNA methylation profiles, including imprint erasure (Takashima et al, 2014; Theunissen et al, 2014), and this might influence subsequent differentiation processes, including the methylation profiles of differentiated cells. An appropriate induction system of hPGCLCs from “naïve” hPSCs warrants future exploration.

Under the condition we established herein, a fraction of BT⁺AG⁺ cells lose their BTAG expression and high‐INTEGRINα6 state and de‐differentiate into cells similar to hiPSCs/iMeLCs. It has been widely accepted that PGCs are the source for teratomas both in mice and in humans (Stevens, 1967; Oosterhuis & Looijenga, 2019), and accordingly, an instability of the PGC state in culture has been well‐documented in mice, in which mPGCs differentiate into cells similar to mESCs in culture (mouse embryonic germ cells: mEGCs; Durcova‐Hills et al, 2008; Leitch et al, 2010; Matsui et al, 1992). Collectively, these findings lead to the concept that PGCs bear a “potential” pluripotency (Hackett & Surani, 2014). Our finding that d6 hPGCLCs/ BT⁺AG⁺ cells in expansion culture also bear a potential to de‐differentiate into a pluripotent state serves as further evidence of the potential pluripotency of hPGC(LC)s. bFGF, which we have used to enhance the propagation of BT⁺AG⁺ cells, might also play a role in their de‐differentiation, since it is known as a key signal for the maintenance of pluripotency in hPSCs (Thomson et al, 1998). It should be noted that during the preparation of our manuscript, Gell et al reported that hPGCLCs can survive for ten days under the condition for mPGCLC expansion (Ohta et al, 2017; Gell et al, 2020).

In summary, the establishment of a system for hPGCLC expansion should serve as a key not only to accelerate a mechanistic understanding of hPGCLC specification, survival, and propagation, but also to explore a strategy for the expansion of human germ cells at different developmental stages, such as gonocytes/oogonia, or of PGCLCs from other relevant species such as cynomolgus monkeys (Sakai et al, 2019). Such studies will provide a foundation for further promoting the human germ cell biology and the understanding of human germ cell development from an evolutionary viewpoint.

Materials and Methods

Reagents and Tools table

Reagent/resource	Reference or source	Identifier or catalog number
Experimental models
585B1‐BTAG	Sasaki et al (2015)	BTAG 585B1‐868
1390G3‐AGVT	Yamashiro et al (2018)	1390G3 AGVT_5‐2‐6
1383D6	Yokobayashi et al (2017)	1383D6
m220	Ohta et al (2017)	N/A
ICR	Shimizu Laboratory Suppliers	Slc:ICR
Antibodies
Mouse anti‐BLIMP1, monoclonal	R&D Systems	Cat # MAB36081
Goat anti‐SOX17, polyclonal	R&D Systems	Cat # AF1924
Mouse anti‐AP2γ/TFAP2C, monoclonal	Santa Cruz	Cat # sc‐12762
Mouse anti‐OCT4/POU5F1, monoclonal	Santa Cruz	Cat # sc‐5279
Goat anti‐NANOG, polyclonal	R&D Systems	Cat # AF1997
Goat anti‐SOX2, polyclonal	Santa Cruz	Cat # sc‐17320
Goat anti‐FOXL2, polyclonal	Novus Biologicals	Cat # NB100‐1277
Mouse anti‐DAZL, monoclonal	Santa Cruz	Cat # sc‐390929
Goat anti‐DDX4, polyclonal	R&D Systems	Cat # AF2030
Mouse anti‐human mitochondria, monoclonal	Millipore	Cat # MAB1273
Rat anti‐GFP, monoclonal	Nacalai Tesque	Cat # 04404‐84
Mouse anti‐UHRF1, monoclonal	Millipore	Cat # MABE308
Rabbit anti‐DNMT1, polyclonal	Abcam	Cat # ab19905
Rat anti‐CD49f/ITGA6 conjugated with Brilliant Violet 421, monoclonal	BioLegend	Cat # 313624
Mouse anti‐EpCAM conjugated with APC, monoclonal	BioLegend	Cat # 324208
Mouse anti‐TRA‐1‐85 conjugated with Brilliant Violet 421, monoclonal	BD Horizon	Cat # 563302
Anti‐SSEA‐1 antibody, monoclonal	Miltenyi Biotec	Cat # 130‐094‐530
Anti‐CD31 antibody, monoclonal	Miltenyi Biotec	Cat # 130‐097‐418
Donkey anti‐rat IgG conjugated with Alexa Fluor 488	Invitrogen	Cat # A21208
Donkey anti‐mouse IgG conjugated with Alexa Fluor 568	Invitrogen	Cat # A10037
Donkey anti‐mouse IgG conjugated with Alexa Fluor 647	Invitrogen	Cat # A31571
Donkey anti‐goat IgG conjugated with Alexa Fluor 647	Invitrogen	Cat # A21447
Donkey anti‐rabbit IgG conjugated with Alexa Fluor 647	Invitrogen	Cat # A31573
Oligonucleotides and sequence‐based reagents
qPCR primers	This study	Table EV10
Chemicals, enzymes, and other reagents
Stem Fit AK03/AK03N	Ajinomoto	N/A
iMATRIX	Nippi	892014
TrypLE select	Gibco	12563‐011
EDTA	Nacalai Tesque	06894‐14
Trypsin–EDTA	Gibco	15400‐054
DNase I	Sigma	DN25
Knockout serum replacement	Gibco	10828‐028
NEAA	Gibco	11140‐050
Penicillin and streptomycin	Gibco	15140‐122
l‐glutamine	Gibco	25030‐081
Sodium pyruvate	Gibco	11360‐070
2‐mercaptoethanol	Gibco	21985‐023
FBS	Gibco / Sigma	10437‐028 / F7524
BSA for FACS buffer	Gibco	15260‐037
BSA for immunofluorescence	Sigma	A3059
Normal donkey serum	Jackson ImmunoResearch	017‐000‐121
Triton‐X 100	Nacalai Tesque	35501‐02
ibidi mounting media	ibidi	50001
Mitomycin C	Sigma / Kyowa Kirin	M0503 / MITOMYCIN Injection
Fibronectin	Merck Millipore	FC010
Ascorbic acid	Sigma	A4403
DNase‐free water	Gibco	15230‐162
Unmethylated λ phage DNA	Promega	D1521
Phusion Hot Start II DNA polymerase	Thermo Fisher Scientific	F‐549S
AxyPrep MAG PCR Clean‐Up Kit	Corning	MAG‐PCR‐CL‐250
Activin A	PeproTech	AF‐120‐14
BMP4	R&D Systems	314‐BP
SCF	R&D Systems	255‐SC
LIF	Merck Millipore	LIF1010
EGF	R&D Systems / PeproTech	236‐EG / AF‐100‐15
bFGF	Wako	064‐04541
Forskolin	Sigma	F3917
Rolipram	Abcam	ab120029
Cyclosporin A	Sigma	30024
CHIR99021	Tocris	4423
Y27632	Wako / Tocris	253‐00513 / 1254
DMEM (4.5 g/l glucose)	Gibco	10313‐021
αMEM	Gibco	32571‐036
Other media used in the culture condition considerations	N/A	Table EV1
Software
FV10‐ASW	Olympus	N/A
XnConvert	https://www.xnview.com/	N/A
R version 3.6.1	R Core Team (2019)	N/A
DAVID 6.8	Huang et al (2009)	N/A
Feature Extraction 12.1.0.3	Agilent Technologies	N/A
Cytogenomics 5.0.2.5	Agilent Technologies	N/A
cutadapt v1.9.1	Martin (2011)	N/A
Tophat v2.1.0	Kim et al (2013)	N/A
Bowtie2 v2.2.7	Langmead and Salzberg (2012)	N/A
HTSeq v0.9.1	Anders et al (2015)	N/A
Trim_galore program	http://www.bioinformatics.babraham.ac.uk/projects/trim_galore	N/A
Bismark v0.17.0	Krueger and Andrews (2011)	N/A
Samtools v1.3	Li et al (2009)	N/A
IGVTools v2.3.52	Robinson et al (2011)	N/A
Methpipe v3.4.3	Song et al (2013)	N/A
bedtools v2.29.2	Quinlan and Hall (2010)	N/A
Homer v4.9.1	http://homer.ucsd.edu/homer	N/A
Other
RNeasy Micro Kit	Qiagen	74004
NucleoSpin RNA XS	MACHEREY‐NAGEL	U0902A
NucleoSpin Tissue	MACHEREY‐NAGEL	U0952Q
EZ DNA Methylation‐Gold Kit	Zymo Research	D5005
Qubit RNA HS Assay Kit	Invitrogen	Q32855
Power SYBR Green PCR Master Mix	Applied Biosystems	4367659
SureTag Complete DNA Labeling Kit	Agilent Technologies	5190‐4240
SurePrint G3 Human CGH Microarray 8 × 60K	Agilent Technologies	G4405A
V‐bottom 96‐well plate	NOF / Greiner	51011612 / 651970
U‐bottom 96‐well plate	Thermo Fisher Nunc	174925
Falcon cell strainer	Corning	352235
Film‐bottom dish	Matsunami Glass	FD10300
CKX41 inverted microscope	Olympus	N/A
DS‐Fi2	Nikon	N/A
M205C microscope	Leica	N/A
DP72	Olympus	N/A
FV1000‐IX81 confocal Microscope system	Olympus	N/A
CFX384 Touch Real‐Time PCR detection system	Bio‐Rad Laboratories	N/A
SureScan Microarray scanner G2600D	Agilent Technologies	N/A
NextSeq 500/550	Illumina	N/A
HiSeq 2500	Illumina	N/A

Methods and Protocols

Animals and hiPSCs

All animal experiments were performed under the ethical guidelines of Kyoto University (Approval No. MedKyo19001). All experiments using hiPSCs were approved by the Institutional Review Board of Kyoto University and were performed according to the guidelines of the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan.

hiPSC culture

All the cells used in this study were maintained in a humidified incubator with 5% CO₂ at 37°C. hiPSC lines 585B1 BTAG (46 XY; Sasaki et al, 2015) and 1383D6 (46 XY; Yokobayashi et al, 2017) were cultured in StemFit AK03 or AK03N (Ajinomoto) on a plate coated with laminin‐511 E8 fragments (iMATRIX‐511; Nippi, 892014). For a passage, cells were dissociated into single cells by treatment using a 1:1 mixture of TrypLE select (Gibco, 12563‐011) and PBS (−) containing 0.5 mM EDTA (Nacalai Tesque, 06894‐14). 10 μM of ROCK inhibitor (Y‐27632; Wako, 253‐00513 or Tocris, 1254) was added upon plating, and on the next day, the medium was changed to a fresh one without Y27632. Images of hiPSC colonies were taken with a CKX41 inverted microscope (OLYMPUS) equipped with a DS‐Fi2 camera (Nikon).

hPGCLC induction

Human primordial germ cell‐like cells were induced from hiPSCs via iMeLCs as described previously (Sasaki et al, 2015; Yokobayashi et al, 2017). Briefly, iMeLCs were induced from hiPSCs (1.0–2.0 × 10⁵ cells/well) in the GK15 medium [15% knockout serum replacement (KSR; Gibco, 10828‐028), 1% MEM non‐essential amino acid solution (NEAA; Gibco, 11140‐050), 1% penicillin and streptomycin, 2 mM l‐glutamine, 2 mM sodium pyruvate (Gibco, 11360‐070), and 0.1 mM 2‐mercaptoethanol in GMEM (Gibco, 11710‐035)] supplemented with 3 μM CHIR99021 (Tocris, 4423), 50 ng/ml activin A (PeproTech, AF‐120‐14), and 10 μM Y‐27632 on a fibronectin (Millipore, FC010)‐coated 12‐well plate (the induction time was 44–48 h for 585B1 BTAG hiPSCs and 60 h for 1383D6 hiPSCs). hPGCLCs were induced from iMeLCs (3.0–6.0 × 10³ cells/well) in the GK15 medium supplemented with 200 ng/ml BMP4 (R&D Systems, 314‐BP), 100 ng/ml SCF (R&D Systems, 255‐SC), 10 ng/ml LIF (Merck Millipore, LIF1010), 50 ng/ml EGF (R&D Systems 236‐EG or PeproTech AF‐100‐15), and 10 μM Y‐27632 in a V‐bottom 96‐well plate (NOF, 51011612 or Greiner, 651970). After 4 or 6 days from hPGCLC induction, hPGCLCs were isolated by FACS sorting [see the Fluorescence‐activated cell sorting (FACS) section]. Images of iPSCs and iMeLCs were taken with a CKX41 inverted microscope (OLYMPUS) equipped with a DS‐Fi2 camera (Nikon). Images of iMeLC aggregates under the hPGCLC induction condition were taken with a M205C microscope (Leica) equipped with a DP72 camera (Olymbus).

xrOvary culture

The xrOvaries were generated and cultured as described previously (Yamashiro et al, 2018). Briefly, 5.0 × 10³ d6 hPGCLCs or d6c30 hPGCLCs were aggregated with 7.5 × 10⁴ mouse embryonic ovarian somatic cells of the ICR strain at E12.5 (Shimizu Laboratory Supplies). After a floating culture in GK15 containing 10 μM Y‐27632 in a U‐bottom 96‐well plate (Thermo Fisher Nunc, 174925) for 2 days to form the aggregates, the aggregates were transferred and cultured in the xrOvary culture medium [αMEM (Gibco, 32571‐036) with 10% FBS, 1% penicillin and streptomycin, 150 μM l‐ascorbic acid (Sigma, A4403), and 55 μM 2‐mercaptoethanol] on a Transwell‐COL membrane insert (Corning, 3496) under a liquid–gas interface condition. Half of the xrOvary culture medium was replaced with the new one every 3 days. The isolation of mouse embryonic ovarian somatic cells was performed as described previously (Hayashi & Saitou, 2013; Yamashiro et al, 2018). For magnetic‐activated cell sorting (MACS), dissociated somatic cells were incubated with anti‐SSEA‐1 antibody (Miltenyi Biotec, 130‐094‐530) and anti‐CD31 antibody (Miltenyi Biotec, 130‐097‐418) for 20–40 min on ice.

For the statistical analysis of the number of BT⁺AG⁺ cells in xrOvaries, we employed Welch's t‐test by using the t.test function with the “var.equal = F” option in R software version 3.6.1.

Fluorescence‐activated cell sorting (FACS)

Human primordial germ cell‐like cells were isolated from the aggregates of iMeLCs as described previously (Sasaki et al, 2015; Yokobayashi et al, 2017) with some modifications. Typically, the aggregates were washed with PBS (−) once and incubated in 1:1 mixture of 0.5% trypsin–EDTA (Gibco, 15400‐054) and PBS (−) at 37°C for 15 min. The aggregates were dispersed by pipetting, and the trypsin treatment was neutralized with the STOP medium [DMEM (Gibco, 10313‐021) with 10% FBS, 1% penicillin and streptomycin, 2 mM l‐glutamine, 10 μM Y‐27632, and 0.1 mg/ml DNase I(Sigma, DN25)]. After pelleting, the cells were resuspended in the FACS buffer [PBS (−) with 0.1% BSA and 10 μM Y‐27632] and strained with a Falcon Cell Strainer (Corning, 352235).

To isolate hPGCLCs with surface markers (Sasaki et al, 2015; Yokobayashi et al, 2017), the dissociated iMeLC aggregates were incubated with BV421‐conjugated anti‐CD49f (INTEGRINα6, ITGA6) antibody (BioLegend, 313624) and APC‐conjugated anti‐CD326 (EpCAM) antibody (BioLegend, 324208) on ice for 15 min in the dark in FACS buffer and washed with PBS (−) once. After pelleting, the cells were resuspended in FACS buffer and strained with a Falcon Cell Strainer.

BT⁺AG⁺ or INTEGRINα6^{high or low}/EpCAM^high cells were sorted with FACSAria III (BD Bioscience). For the expansion culture, the cells were sorted into FACS buffer. To collect the cells for the following analysis (RT–qPCR, RNA‐seq, and WHOLE‐GENOME BISULFITE SEQUENCE), the cells were sorted into CELLOTION (ZENOAQ, CB051). For the xrOvary culture, BT⁺AG⁺ cells were sorted into GK15 medium supplemented with 10 μM of Y‐27632.

For the passage of expanded hPGCLCs, cells were washed with PBS (−) once and treated with a 1:4 mixture of 0.5% trypsin–EDTA and PBS (−) at 37°C for 5 min. After the rigorous pipetting, the trypsin treatment was neutralized with the STOP medium or PBS (−) containing 10% FBS, 10 μM Y‐27632, and 0.1 mg/ml DNase I. Then, the cells were pelleted and resuspended in FACS buffer. After removing cellular clumps with Falcon Cell Strainer, the cells were subjected to FACS sorting. To passage the expanded hPGCLCs on the basis of the expression of CD49f and CD326, the dissociated cells were immune‐labeled as described. BT⁺AG⁺ or INTEGRINα6^figh/EpCAM^high cells were isolated with FACSAria III. For the passage, BT⁺AG⁺ or INTEGRINα6^figh/EpCAM^high cells were sorted into FACS buffer.

To analyze the expression of TRA‐1‐85 in hPGCLC‐derived cells in the expansion culture, the cells suspended in FACS buffer were incubated with anti‐TRA‐1‐85 antibody (BD Horizon, 563302) on ice for 30 min in the dark and washed with PBS (−) once. After pelleting, the cells were resuspended in FACS buffer and strained with a Falcon Cell Strainer. 3 mM DRAQ7 (Abcam, ab109202) was added to the cell suspension, and the suspension was incubated for 10 min at room temperature and subjected to a FACS analysis.

hPGCLC expansion culture

The m220‐5 cell line (Dolci et al, 1991; Majumdar et al, 1994; Ohta et al, 2017; Miyauchi et al, 2018) was maintained in DMEM (Gibco, 10313‐021) containing 10% FBS (Gibco 10437‐028 or Sigma F7524), 1% penicillin and streptomycin (Gibco, 15140‐122), and 2 mM l‐glutamine (Gibco, 25030‐081). Mitomycin C (MMC, Sigma M0503, or KYOWA KIRIN MITOMYCIN Injection: 2 or 4 μg/ml, 2 h)‐treated m220‐5 cells were used as feeder cells for the hPGCLC expansion culture. Before MMC treatment, m220‐5 cells were passaged with 10 μM forskolin (Sigma, F3917) and 10 μM rolipram (Abcam, ab120029) 3 to 4 times for the expansion and adaption to the supplemented chemicals.

To evaluate the effects of cytokines and/or chemicals on hPGCLC expansion, hPGCLCs were cultured on the m220‐5 cells treated with MMC in GK15 containing 2.5% FBS and 100 ng/ml SCF supplemented with cytokines and/or chemicals [10 ng/ml LIF; 50 ng/ml EGF; 20 ng/ml bFGF; 10 μM forskolin; 10 μM rolipram; 5 μM cyclosporin A (Sigma, 30024)], and passaged every 10 days [see the Fluorescence‐activated cell sorting (FACS) section]. We plated approximately 5.0 × 10³ cells in the 24‐well plate on the day of the passage (0.5 ml of the medium containing 10 μM of Y27632) and added 0.5 ml of the medium without Y‐27632 on the next day. On the third day, the whole medium was changed (1 ml/well). On the 5^th, 7^th, and 9^th day after the passage, half of the medium was exchanged with the new one. To evaluate the effects of the basal medium and the glucose concentration, hPGCLCs were cultured on the m220‐5 cells treated with MMC in various media supplemented with 2.5% FBS, 10 μM forskolin, 100 ng/ml SCF, and 20 ng/ml bFGF, and passaged every 10 days [see the Fluorescence‐activated cell sorting (FACS) section]. The media used for these considerations are listed in Table EV1. We plated 4.5–5.0 × 10³ cells in the 24‐well plate on the day of the passage (0.5 ml of the medium containing 10 μM of Y27632) and added 0.5 ml of the medium without Y‐27632 on the next day. The medium changes were performed as described above.

The contents of the defined hPGCLC expansion culture medium were as follows: DMEM (1.0 g/l glucose; Gibco, 11054‐020) with 15% KSR, 2.5% FBS, 1% NEAA, 2 mM GlutaMAX (Gibco, 35050‐061), 1% penicillin and streptomycin, and 0.1 mM 2‐mercaptoethanol (Gibco, 21985‐023) supplemented with 10 μM forskolin, 100 ng/ml SCF, and 20 ng/ml bFGF. The passage and medium change procedures are as described above.

Detailed step‐by‐step protocols for hiPSC culture, hPGCLC induction, xrOvary culture, and associated FACS analysis are described in Yamashiro et al (2020).

A step‐by‐step protocol for hPGCLC expansion culture is as follows:

1. Coat a flat‐bottom cell culture plate (i.e., Falcon, 353043) with 0.1% gelatin solution at room temperature for at least 1 h.

2. Plate MMC‐treated (4 μg/ml, 2 h) m220‐5 feeders on a gelatin‐coated cell culture plate. Typically, 5.0 × 10⁵ cells are plated in one well of a 12‐well plate with 1 ml of DMEM containing 10% FBS, 1% penicillin and streptomycin, and 2 mM l‐glutamine.

3. Harvest d6 hPGCLCs or expanded hPGCLCs (BT⁺AG⁺ cells or INTEGRINα6^high/EpCAM^high cells) by FACS sorting (see the above sections) in FACS buffer and pellet them down by centrifugation at 200 g for 7 min.

4. After discarding the supernatant, suspend the cells in hPGCLC expansion culture medium with 10 μM Y27632 so that the cell concentration becomes 1.0 × 10⁴ cells/ml.

5. Plate d6 hPGCLCs or expanded hPGCLCs on MMC‐treated m220‐5 feeders. Typically, 1 ml of cell suspension equivalent to 1.0 × 10⁴ cells is plated in one well of a 12‐well plate.

6. On the next day, add hPGCLC expansion culture medium. Typically, 1 ml of the medium is added to one well of a 12‐well plate.

7. On the 3^rd day, replace all the medium with new medium (2 ml).

8. On the 5^th/7^th/9^th day, replace half the medium with new medium (1 ml).

9. On the 10^th day, passage the expanded hPGCLCs to a new culture plate.

Immunofluorescence (IF) analysis

The expanded PGCLCs were cultured in the film‐bottom dish (Matsunami Glass, FD10300) for several days and washed with PBS (−) once. The cells were fixed in 4% paraformaldehyde for 15 min at room temperature and washed with PBS (−) 3 times. For blocking and permeabilization, the fixed cells were incubated in blocking buffer [PBS (−) with 10% normal donkey serum (Jackson ImmunoResearch, 017‐000‐121), 3% BSA (Sigma, A3059), and 0.1% Triton‐X 100 (Nacalai Tesque, 35501‐02)]. Then, the cells were incubated in 0.5× blocking buffer [a 1:1 mixture of blocking buffer and PBS (−)] containing primary antibodies. The conditions used for the incubation with primary antibodies were as follows: For mouse anti‐BLIMP1, goat anti‐SOX17, mouse anti‐AP2γ/TFAP2C, mouse anti‐OCT4/POU5F1, goat anti‐SOX2, and mouse anti‐human mitochondria, the incubation was performed at 4°C overnight; for mouse anti‐UHRF1 and rabbit anti‐DNMT1, the incubation was performed at room temperature for 30 min; and for goat anti‐NANOG, the incubation was performed at 4°C overnight or room temperature for 30 min. After the primary antibody reaction, the cells were washed with PBS (−) 3 times and incubated in 0.5× blocking buffer containing 1 μg/ml of DAPI and secondary antibodies for 1 h at room temperature. The cells were washed with PBS (−) 3 times and mounted in ibidi Mounting Media (ibidi, 50001).

The IF analysis of xrOvaries was performed as described previously (Yamashiro et al, 2018). To mount the stained sections, ibidi Mounting Media without DAPI, instead of VECTASHIELD mounting medium (Vector Laboratories, H‐1000) with DAPI, was used.

The following primary antibodies were used at the indicated dilution: mouse anti‐BLIMP1 (1/200; R&D Systems, MAB36081); goat anti‐SOX17 (1/100; R&D Systems, AF1924); mouse anti‐AP2γ/TFAP2C (1/100; Santa Cruz, sc‐12762); mouse anti‐OCT4/POU5F1 (1/100; Santa Cruz, sc‐5279); goat anti‐NANOG (1/100; R&D Systems, AF1997); goat anti‐SOX2 (1/100; Santa Cruz, sc‐17320); mouse anti‐UHRF1 (1:200; Millipore, MABE308); rabbit anti‐DNMT1 (1:200; Abcam, ab19905); mouse anti‐human mitochondria (1:500; Millipore, MAB1273); goat anti‐FOXL2 (1/500; Novus Biologicals, NB100‐1277); mouse anti‐DAZL (1/100; Santa Cruz, sc‐390929); goat anti‐DDX4 (1/400; R&D Systems, AF2030); and rat anti‐GFP (1/250; Nacalai Tesque, 04404‐84).

The following secondary antibodies from Invitrogen were used at a 1/800 dilution: Alexa Fluor 488 donkey anti‐rat IgG (A21208); Alexa Fluor 568 donkey anti‐mouse IgG (A10037); Alexa Fluor 647 donkey anti‐mouse IgG (A31571); Alexa Fluor 647 donkey anti‐goat IgG (A21447); and Alexa Fluor 647 donkey anti‐rabbit IgG (A31573).

Images were taken with the FV1000‐IX81 confocal microscope system (Olympus) and processed with the FV10‐ASW software. Processed images were trimmed with XnConvert software (https://www.xnview.com/en/xnconvert/).

RNA extraction, reverse transcription, and cDNA amplification

For the RNA extraction, cells collected in CELLOTION were pelleted, lysed, and stored at −80°C until use. RNA extraction was performed with an RNeasy Micro Kit (Qiagen, 74004) or NucleoSpin RNA XS (MACHEREY‐NAGEL, U0902A) according to the manufacturer's instruction. When using NucleoSpin RNA XS, we added an additional centrifugation just before the elution step. The concentration of RNA was measured with a Qubit RNA HS Assay Kit (Invitrogen, Q32855). One nanagram of total RNA from each sample was used for the reverse transcription, and the cDNA amplification was performed as described previously (Nakamura et al, 2015). When the RNA concentration was too low to be measured (in the present study, RNA extraction was performed with 1,606 to 10,968 cells), we centrifuged the RNA solution to reduce its volume and used the whole concentrated RNA solution for the reverse transcription and the cDNA amplification.

Quantitative PCR (qPCR) and RNA‐sequence (RNA‐seq) analysis

The amplified cDNAs as described were used for qPCR analysis. We performed qPCR with Power SYBR Green PCR Master Mix (Applied Biosystems, 4367659) on a CFX384 Touch Real‐Time PCR detection system (Bio‐Rad Laboratories). Primer sequences used were as listed in Table EV10.

For RNA‐seq analysis, sequence libraries were constructed with amplified cDNAs as described previously (Nakamura et al, 2015; Ishikura et al, 2016) with some modifications. When purifying cDNA libraries, we used AxyPrep PCR purification MAG PCR Clean‐Up Kit (Axygen, MAG‐PCR‐CL‐250), instead of Agencourt AMPure XP (Beckman Coulter, A63881). Sequences were performed on an Illumina NextSeq 500/550 platform with a NextSeq 500/550 High Output Kit v2.5 (75 cycles; Illumina).

Whole‐genome bisulfite sequence

The whole‐genome bisulfite sequence libraries were generated with a post‐bisulfite adaptor tagging method (http://www.crest-ihec.jp/english/epigenome/index.html; Miura et al, 2012; Shirane et al, 2016). 10,000–15,000 cells were lysed in the lysis buffer [0.1% SDS and 1 μg/μl Proteinase K in DNase‐free water (Gibco, 15230‐162)] containing unmethylated λ phage DNA (Promega, D1521) at 37°C for 60 min and then incubated at 98°C for 15 min to inactivate proteinase K. Assuming that a single cell bears 6 pg of genomic DNA, we added the unmethylated λ phage DNA equivalent to 0.5% of the input DNA. For the bisulfite treatment, the cell lysate was treated with an EZ DNA Methylation‐Gold Kit (Zymogen, D5005) according to the manufacturer's instruction. To construct sequence libraries, the bisulfite‐treated DNA was subjected to the serial synthesis of the single‐end adaptor‐tagged complementary DNA fragments (http://www.crest-ihec.jp/english/epigenome/index.html). When performing the PBAT method, we used Phusion Hot Start II DNA Polymerase (Thermo Fisher Scientific, F549S) and an AxyPrep MAG PCR Clean‐Up Kit (Corning, MAG‐PCR‐CL‐250) instead of Phusion Hot Start High‐Fidelity DNA Polymerase (Thermo Fisher Scientific, F540S) and Agencourt AMPure XP, respectively. Sequencing was performed on an Illumina HiSeq 2500 platform with the TruSeq SR Cluster Kit v3‐cBot‐HS and TruSeq SBS Kit v3‐HS (Illumina).

Array comparative genomic hybridization (aCGH)

For extracting genomic DNA, cells were collected in CELLOTION and pelleted, and genomic DNA extraction was performed with NucleoSpin Tissue (MACHEREY‐NAGEL, U0952Q) according to the manufacturer's instruction. An aCGH experiment and image processing of the microarray were performed with a SureTag Complete DNA Labeling Kit (Agilent Technologies, 5190‐4240) by Takara Bio. Human Reference DNA Male (Agilent Technologies) was used for the sex‐matched reference genomic DNA. Briefly, 200 ng genomic DNA was digested with Alu1 and Rsa1, and then labeled with Cyanine 3 (reference genomic DNA) or Cyanine 5. Labeled genomic DNA was hybridized on a SurePrint G3 Human CGH Microarray 8 × 60K (Agilent Technologies, G4450A) at 60°C for 24 h at 20 rpm. After washing, the slide was scanned with a SureScan Microarray Scanner G2600D (Agilent Technologies). The acquired images were processed with Feature Extraction 12.1.0.3 (Agilent Technologies; see aCGH data analysis).

RNA‐seq data processing and analysis

The human genome sequence and transcript annotation GFF3 file of annotation release 107 (GRCh38.p2) were obtained from the NCBI ftp site (ftp://ftp.ncbi.nlm.nih.gov/). Reference gene annotations at the 3’‐ends were extended up to 10 kb to fully cover the transcription termination sites (TSSs).

Processing of reads into expression levels was accomplished as described previously, with modifications (Nakamura et al, 2015). Briefly, all reads were processed with cutadapt v1.9.1 (Martin, 2011) to remove low‐quality (quality value < 30) or adapter sequences. Reads (≥ 30 mer) were mapped on the human genome, GRCh38.p2, using Tophat (v2.1.0)/Bowtie2 (v2.2.7) with the “‐no‐coverage‐search” option (Langmead & Salzberg, 2012; Kim et al, 2013), followed by HTSeq (v0.9.1; Anders et al, 2015) to estimate forward read counts per gene. Read counts were converted to reads per million mapped reads (RPM) using the sum of all forward read counts for all genes. Genes whose log₂ (RPM + 1) ≧ 4 in at least one sample were used for unsupervised hierarchical clustering (UHC) and principal component analysis (PCA). Unsupervised hierarchical clustering and PCA were carried out using R software version 3.6.1 (R Core Team, 2019). PCA was performed with prcomp function. Unsupervised hierarchical clustering was performed with Dist function with method = “euclidean” setting and with hclust function with method = “ward.D2” setting. Gene ontology (GO) analysis was carried out using DAVID 6.8 (Huang et al, 2009).

Whole‐genome bisulfite sequence data processing and analysis

Genome release GRCh38.p2 was used as a human genome sequence and transcript reference. Promoter regions were defined as regions between 900‐bp upstream and 400‐bp downstream of transcription start sites (TSSs). High‐, intermediate‐, and low‐CpG promoters (HCP, ICP, and LCP, respectively) were calculated as described in Borgel et al (2010). The list of CpG islands (CGI) was obtained from Illingworth et al (2010). CGIs in the hg19 genome were converted to GRCh38 using the LiftOver program (https://genome.ucsc.edu/cgi-bin/hgLiftOver). Imprinted differentially methylated regions (DMR) were from Court et al (2014). Regions that show differential methylation in spermatozoa and oocytes (≥ 50% in spermatozoa or ≥ 50% in oocytes, plus a more than 50% difference in the number of spermatozoa and oocytes) are shown in the figure. RepeatMasker data for human GRCh38 were used for genomic repetitive element information and were obtained from UCSC table browser (https://genome.ucsc.edu/cgi-bin/hgTables). Intergenic regions were randomly selected 2‐kb bins between genes excluding repetitive elements. Two‐hundred randomly selected regions are shown in the heat map. The read processing, mapping, and estimation of the methylated C levels were performed as described previously (Shirane et al, 2016). Briefly, all reads were processed with the Trim_galore program (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) to remove four bases from the 5’‐ends, one base from the 3’‐ends, adaptor sequences, and low‐quality reads (quality score < 20). The processed reads were mapped onto the human genome (GRCh38.p2) with the bismark program from Bismark (v0.17.0) with the “–pbat” option (Krueger & Andrews, 2011). Mapped data (BAM format) were converted to methylation levels using the bismark_methylation_extractor program from Bismark. Samtools (v1.3; Li et al, 2009) and IGVTools (v2.3.52; Robinson et al, 2011) were used to manipulate BAM files and create wig files.

Only CpGs in which the read depth is between 4 and 200 are used for % methylated C (mC) calculations. Only regions in which CpG (read depth ≥ 4) counts were four or more within genome‐wide 2‐kb bins, promoters, exons, introns, intergenic regions, or imprinted DMRs were used for calculating averages or other analyses. For the analysis on CpA, CpC, and CpT methylations, percent methylations were calculated as % mC per kb as follows: The numbers of methylation calls over genome‐wide 1‐kb bin were summed and divided by the sum of the numbers of unmethylation calls in the same region. The methylation levels per kb for CpA, CpC, and CpT were calculated separately throughout the whole genome.

The “escapees” from DNA demethylation in c30ag77 cells (this study), ag77_2, and ag120 AG⁺VT⁺ cells (Yamashiro et al, 2018), and pooled human germ cells from week 7 to 9 (Tang et al, 2015) were defined by using hypermr program in MethPipe v3.4.3 package with default setting (Song et al, 2013). The escapees in common among these cells or specific to some cell types were calculated using merge and intersect commands of bedtools v2.29.2 (Quinlan & Hall, 2010). The annotations of the escapees were performed using annotatePeaks program in Homer v4.9.1. The data were visualized with R software using “upset” function in “UpSetR” library. The enrichment scores were calculated from the numbers of escapees overlapped with indicated genomic elements over the numbers of shuffled bins overlapped with the same category (average of five trials).

Data processing for hPGCLC‐derived cells in xrOvaries (Yamashiro et al, 2018), spermatozoa, and oocytes (Okae et al, 2014) was performed as described above. Data for mPGCLCs (Shirane et al, 2016) were mapped onto mouse genome mm10/GRCm38 and processed in the same manner as described above.

aCGH data analysis

aCGH raw data were analyzed with CytoGenomics 5.0.2.5 (Agilent Technologies) with the analysis method “Default Analysis Method ‐ CGH v2”, which detects aberrant intervals with the algorithm ADM2. Additionally, “Fuzzy Zero” correction was applied to mitigate the erroneous calls of aberrant intervals. The list of copy number variations in Human Agilent Reference DNA Male against hg19 is available from CytoGenomics 5.0.2.5.

Statistical analysis

All statistical tests were performed using R software version 3.6.1 (R Core Team, 2019). Welch's t‐test was performed with t.test function with “var.equal=F” option, and the Wilcoxon rank‐sum test was performed with wilcox.exact function.

Author contributions

Y.M. performed all the experiments and analyzed the data. Y.Y., T.N., and T.Y. contributed to the analyses of RNA‐seq/whole‐genome bisulfite sequence data. H.O. and C.Y. assisted in the induction and expansion of hPGCLCs. Y.M. and M.S. designed the experiments and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

Table EV4

Table EV5

Table EV6

Table EV7

Table EV8

Table EV9

Table EV10

Dataset EV1

Review Process File

The EMBO Journal (2020) 39: e104929

Data availability

The accession numbers of the data generated in this study: the RNA‐seq data: GSE147498 (the GEO database); the whole‐genome bisulfite sequence data: GSE147499 (the GEO database); and the aCGH data: GSE148415 (the GEO database).

The accession numbers for the RNA‐seq data: hiPSCs/iMeLCs: GSE99350 (the GEO database); d6 hPGCLCs and ag7/21/35/49/63/77/120 cells: GSE117101; the whole‐genome bisulfite sequence data: hiPSCs/iMeLCs/d6 hPGCLCs and ag35/77 cells: DRA006618 (the DDBJ database); and ag120 cells: DRA007077 (the DDBJ database).

The accession numbers for the whole‐genome bisulfite sequence data of d4c3/d4c7 mPGCLCs: DRA005166; mESCs/EpiLCs/d2/d4/d6 mPGCLCs: DRA003471; and E10.5/E13.5/E16.5 germ cells: DRA000607.