Tissue of Origin, but Not XCI State, Influences Germ Cell Differentiation from Human Pluripotent Stem Cells

Human pluripotent stem cells (hPSCs) are not only a promising tool to investigate differentiation to many cell types, including the germline, but are also a potential source of cells to use for regenerative medicine purposes in the future. However, current in vitro models to generate human primordial germ cell-like cells (hPGCLCs) have revealed high variability regarding differentiation efficiency depending on the hPSC lines used. Here, we investigated whether differences in X chromosome inactivation (XCI) in female hPSCs could contribute to the variability of hPGCLC differentiation efficiency during embryoid body (EB) formation. For this, we first characterized the XCI state in different hPSC lines by investigating the expression of XIST and H3K27me3, followed by differentiation and quantification of hPGCLCs. We observed that the XCI state did not influence the efficiency to differentiate to hPGCLCs; rather, hPSCs derived from cells isolated from urine showed an increased trend towards hPGCLCs differentiation compared to skin-derived hPSCs. In addition, we also characterized the XCI state in the generated hPGCLCs. Interestingly, we observed that independent of the XCI state of the hPSCs used, both hPGCLCs and soma cells in the EBs acquired XIST expression, indicative of an inactive X chromosome. In fact, culture conditions for EB formation seemed to promote XIST expression. Together, our results contribute to understanding how epigenetic properties of hPSCs influence differentiation and to optimize differentiation methods to obtain higher numbers of hPGCLCs, the first step to achieve human in vitro gametogenesis.


Introduction
Human pluripotent stem cells (hPSCs), such as human embryonic stem cells (hESC) and induced pluripotent stem cells (hiPSC), can be maintained in culture for long periods of time while preserving their pluripotency, self-renewal characteristics and the potential to differentiate to virtually all cell types of the body [1,2]. This potential enables tremendous opportunities in regenerative medicine, biomedical applications, and developmental biology research. However, hPSC lines maintained in culture can display (epi)genetic instabilities, significantly impeding differentiation outcomes [3,4]. Female hPSCs are subjected to greater repercussions of epigenetic instability due to the fact that they contain two X chromosomes and rely on a process known as X chromosome inactivation (XCI) to achieve dosage compensation [5,6]. XCI ensures the balance of X chromosome-linked genes between males and females by inactivating more than 1000 genes on the inactivated SOX17, but also of surface makers such as ALPL, PDPN, EPCAM and ITGA6 [22][23][24]. In addition, PGCs undergo significant epigenetic reprogramming, including genome-wide DNA demethylation and remodelling of histone marks/variants, leading to the erasure of genomic imprints [25,26] and, in female hPGCs, the reactivation of the Xi, a process known as X chromosome reactivation (XCR) [26,27]. In this study, we investigated whether different XCI states in hPSCs were associated with the capacity to differentiate to primordial germ cell-like cells (hPGCLCs) in vitro.

Maintenance of hESCs and hiPSCs
The hESC line H9 (WA09) was purchased from WiCell Institute and the hiPSCs used were previously reprogrammed from primary tissues and characterized by the LUMC hiPSC core facility (Table 1) and some have been registered in hPSCreg (https://hpscreg. eu/). The hPSCs were maintained on Vitronectin (Invitrogen, Waltham, MA, USA) or Matrigel (Corning, Corning, NY, USA) coated plates in either mTeSR-Plus or TeSR-E8 media (STEMCELL Technologies, Vancouver, BC, Canada). To coat the plates, Vitronectin was diluted to 5 µg/mL in Dulbecco's phosphate-buffered saline (DPBS) (Invitrogen), whereas Matrigel was diluted (80 times) in DMEM/F12 (Invitrogen) and left on the plate for 1 h (hr) at room temperature (RT). The hPSCs were passaged as small clumps every 5-7 days using ReLeSR passaging reagent (STEMCELL Technologies) and maintained at 37 • C in normoxic conditions (5% CO 2 on air).  [29] 1 These lines are derived from the same donor. 2 These lines are derived from the same donor.
To induce embryoid body (EB) formation, hPSCs were harvested as single cells after treatment with TrypLE Express (Invitrogen) at 37 • C for 5 min. Next, we plated 10,000 cells/well of an ultra-low attachment U-bottom 96-well plate (Greiner, Alphen aan den Rijn, The Netherlands) in 200 µL TeSR-E8 (STEMCELL Technologies) containing 1× RevitaCell Supplement (Invitrogen). The plate was briefly centrifuged at 200× g for 5 min and incubated for 48 h to allow EB formation. To induce spontaneous differentiation in the EBs, the medium was replaced with 200 µL Essential 6 (Invitrogen) with 1% FCS (Sigma-Aldrich) (serum-induced differentiation) or maintained the EBs in TeSR-E8 (3D pluripotency culture) for an additional 6-8 days at 37 • C under normoxic conditions (5% CO 2 on air).

Fluorescence Activated Cell Sorting (FACS)
Day 6 hPGCLC-EBs were harvested and washed with DPBS (Invitrogen), then digested with an enzyme mix containing 5 mg/mL Collagenase IV (Invitrogen), 2.5 mg/mL Hyaluronidase (Sigma-Aldrich) and 100 U/mL Dnase I (Sigma-Aldrich) for 30 min at 37 • C. A single cell suspension of the digested EBs and of dissociated hiPSC F30 was created by pipetting up and down vigorously and passing through a 40 µm nylon mesh cell strainer (Corning). After washing with FACS buffer (DPBS (Invitrogen) with 0.5% bovine serum albumin (BSA) (Sigma-Aldrich)), cells were treated with conjugated antibodies (Table S1) for 30 min at 4 • C, washed with FACS buffer and resuspend in 200 µL FACS buffer with 1 µL 7AAD live/dead exclusion dye (BD Biosciences, San Jose, CA, USA). The EB cell suspensions were analyzed on an LSR-II flow cytometer (BD Biosciences). The FACS data were collected from FACSDiva Software (BD Biosciences) and analyzed with FlowJo software (BD Biosciences).

Immunofluorescence and Histology
Cells to be used for immunostaining were cultured on coverslips and fixed in 4% paraformaldehyde (PFA) (Sigma-Aldrich) for 15 min at RT. The cells were washed three times with PBS and permeabilized with 0.3% Triton-X100 (Sigma-Aldrich) in PBS for 15 min at RT, followed by three PBS washes. To block, cells were incubated with blocking solution containing 1% BSA in PBST (0.05% Tween 20 (Sigma-Aldrich) in PBS) at RT for 1 h. Next, the cells were incubated with specific primary antibodies (Table S1) diluted in blocking solution and incubated overnight at 4 • C, followed by three PBST washes and incubation with specific secondary antibodies (Table S1) and DAPI (Life Technologies, Carlsbad, CA, USA) for 1 h at RT. Finally, the coverslips were washed three times with PBS and mounted with ProLong Gold (Life Technologies).
For day 6 hPGCLC-EBs and day 6-8 EBs, EBs were fixed in 4% PFA for 24 h at 4 • C, embedded first in a small drop of 4% low-melting agarose (Promega, Madison, WI, USA) in H 2 O and that was subsequently embedded in paraffin. The paraffin sections (5 µm) were generated using a RM2065 microtome (Leica, Wetzlar, Germany) and pasted onto StarFrost slides (Waldemar Knittel, Bielefeld, Germany). For deparaffinization, the slides were treated with xylene and 100%, 90%, 80%, 70% ethanol and water; antigen retrieval was performed by heating up the sections submerged in 0.01 M citric buffer (pH 6.0) for 12 min at 98 • C in a TissueWave 2 microwave (Thermo Fisher Scientific, Waltham, MA, USA). Afterwards, sections were blocked, immunostained and mounted as described for the cells on coverslips.

RNA-Fluorescence In Situ Hybridization (FISH)
For RNA-FISH, cultured cells were fixed on coverslips and EBs were fixed and paraffin embedded/sectioned as described for immunofluorescence. RNA-FISH was performed with RNAscope Multiplex fluorescent reagent kit v2 (323100, Advanced Cell Diagnostics, Newark, CA, USA), following manufacturer instructions. Briefly, paraffin sections were deparaffinized, pre-treated with provided H 2 O 2 solution and incubated in retrieval solution for 15 min, subsequently treated with Protease Plus for 15 min at 40 • C in a HybEZ II oven (321720, Advanced Cell Diagnostics). Cells fixed on coverslips were treated with diluted Protease III (1:5) at RT for 10 min after H 2 O 2 incubation. Next, the probe mix of RNAscope Probe-Hs-XIST-C2 (311231-C2) and RNAscope Probe-Hs-HPRT1-C1 (453191-C1) (1 volume of XIST-probe in 50 volume of HPRT1-probe) was added and incubated for 2 h at 40 • C. After hybridization, the mRNA signal was amplified sequentially for each channel (C1 and C2). Fluorophores used to detect signals were Opal 520 (1:800, FP1487001KT, Akoya Biosciences, Marlborough, MA, USA) and Opal 570 (1:1500, FP1488001KT, Akoya Biosciences). Nuclei were counterstained with DAPI and the samples were mounted using ProLong Gold (Life Technologies).

Imaging and Quantification
Imaging of immunofluorescence and/or RNA-FISH was performed on either SP8 confocal laser scanning microscope (Leica) or a LSM 900 Airyscan 2 confocal laser scanning microscope (Zeiss, Oberkochen, Germany). Grayscale images were edited for brightness and contrast on Image J 1.53c (FIJI) [33]. For quantification of the XCI status, cells were categorized based on XIST/HPRT expression in confocal images. For cultured hPSCs (n > 200 cells per cell line), cells were manually counted and cells without HPRT dots were omitted. For sections of hPGCLC-EBs and EBs (n varied between 56-174 cells per cell line), cells were manually counted and, in hPGCLC-EBs, SOX17+ cells were identified as hPGCLCs, and SOX17− cells as soma. Cells of the hPGCLC-EBs and EBs without HPRT and XIST dots were omitted. For the quantification of H3K27me3, cells (n > 200 cells per cell line) were manually counted. The cell counts of the multiple fields were summed, and relative category frequency (percentage) was visualized as a stacked bar plot using R (version 4.0.2; https://cran.r-project.org/bin/windows/base/) and ggplot2 package (version 3.3.3; https://cran.r-project.org/web/packages/ggplot2/index.html).

Statistics
Statistical analysis of the percentage of cells showing non-overlapping HPRT and XIST single dots (XaXi) between Class II and Class III hPSCs, the percentage of H3K27me3+ cells between Class II and Class III hPSCs, the percentage of PGCLCs between Class II F31 and Class III F20 hPSCs, the percentage of H3K27me3+ PGCLCs between Class II F31 and Class III F20 hPSCs, the percentage of cells showing non-overlapping HPRT and XIST single dots (XaXi) between Class II and Class III EBs. The FACS results was performed using Student's t-test (unpaired) and data are shown as mean ± standard deviation (SD), as indicated (*, p-value < 0.05; **, p-value < 0.01; ***, p-value < 0.001; ns, not significant).

Characterisization of XCI State in Several hiPSC Lines
We determined the XCI state of eight different female hPSCs (hESC H9 and hiPSC lines listed in Table 1) by visualizing the expression of XIST (coating the Xi) and HPRT, an X-linked gene subjected to XCI (expressed on Xa) [18]. In female hPSC lines F99, F31, F30, F71, F197 and F198, most cells showed single XIST clouds and separate single HPRT dots indicative of XaXi, whereas in female hPSC lines F20 and H9, most cells showed absence of XIST and a single dot of HPRT, indicative of XaXe ( Figure 1A and Figure S1A). Male hiPSC lines M54 and M72 that contain one active X chromosome showed, as expected, an absence of XIST and one dot of HPRT ( Figure S1B). Therefore, we deemed hPSC lines F99, F31, F30, F71, F197 and F198 to be Class II and hPSC lines H9 and F20 to be Class III. To further confirm the XCI states of the hPSC lines, we additionally examined the localization of H3K27me3 and EZH2 by immunofluorescence and observed the typical perinuclear colocalization of H3K27me3 and EZH2, indicative of the Xi in the six Class II hPSC lines, but not in the two Class III hPSC lines ( Figure 1B and Figure S1C) or in the male lines ( Figure S1B). Quantification for XIST and HPRT ( Figure 1C) and H3K27me3 ( Figure 1D) revealed significant statistical differences between Class II and Class III hPSCs as well as heterogeneity of cell types in culture in agreement with other studies [21,34].
Next, we investigated the localization of H3K9me3, a histone mark that has been associated with gene-poor regions in the Xi [35] and largely maintained on the Xe [16]. In agreement, we observed a similar nuclear pattern of expression between Class II and III lines and the male lines ( Figure S1B,D).

XCI State of Class II and Class III hPSCs Is Maintained upon Spontaneous Differentiation
Class II (XaXi) and Class III (XaXe) hPSCs maintain the XCI state upon differentiation, while Class I cells are subjected to random XCI, although this has been challenged [34]. To investigate this, we differentiated several hPSC lines in FCS-containing medium and examined their XCI states. To confirm differentiation, we first validated loss of pluripotencyrelated markers POU5F1 and SOX2 ( Figure S2A).
After differentiation, Class II lines F99, F31 and F30 maintained one XIST cloud per cell, whereas no XIST cloud could be detected in the Class III lines F20 and H9 ( Figure 2A). However, regarding both H3K27me3 ( Figure 2B) and H3K9me3 ( Figure S2B), the signal was more homogenous across the entire nucleus in the differentiated Class II and Class III cells, except for Class III H9, which showed very low levels of both H3K27me3 and H3K9me3. In differentiated male lines M54 and M72, with no POU5F1 and SOX2 expression ( Figure S2C) and no XIST ( Figure S2D), we observed comparable high levels of H3K27me3 throughout the nucleus ( Figure S2D), whereas the human female HEK293T line (derived from fetal kidney, here used as positive control) showed both a pronounced XIST cloud and accumulation of H3K27me3 in most cells, indicative of XaXi state ( Figure S2D). In agreement with the broad nuclear expression of H3K27me3 in differentiated lines, EZH2 was also now expressed across the nucleus in both differentiated Class II and Class III cells, except for Class III H9 ( Figure 2C).

Influence of XCI State in Differentiation Efficiency to hPGCLCs
Considering the importance of epigenetic resetting in hPGCs, we investigated whether the XCI state of Class II and Class III hPSCs would affect differentiation efficiency to hPGCLCs. To address this question, we differentiated the Class II and Class III hPSCs to hPGCLCs following a previously established protocol using EBs [32] ( Figure 3A). The differentiation efficiencies to hPGCLCs were quantified based on the percentages of established hPGC markers, ITGA6 and EPCAM [24], and ALPL and PDPN [36,37], measured by flow cytometry ( Figure 3B). Note that most ITGA6+/EPCAM+ hPGCLCs were also ALPL+/PDPN+ ( Figure 3B). However, we obtained a substantial variation in differentiation efficiencies to hPGCLCs varying from 2% to 30% using both sets of surface markers, not only between the Class II lines, but also in the male ( Figure 3C). Statistical analysis comparing PGCLC differentiation efficiencies showed neither significant differences between Class II and Class III lines nor between skin-derived Class II and Class III lines (Figure 3C). Interestingly, the hPSCs lines that performed better in PGCLC differentiation were all derived from cells isolated from urine (presumably kidney epithelial cells) (Table 1). However, using ITGA6+/EPCAM+ as a readout to compare skin-derived with urine-derived

Influence of XCI State in Differentiation Efficiency to hPGCLCs
Considering the importance of epigenetic resetting in hPGCs, we investigated whether the XCI state of Class II and Class III hPSCs would affect differentiation efficiency to hPG-CLCs. To address this question, we differentiated the Class II and Class III hPSCs to hPGCLCs following a previously established protocol using EBs [32] ( Figure 3A). The differentiation efficiencies to hPGCLCs were quantified based on the percentages of established hPGC markers, ITGA6 and EPCAM [24], and ALPL and PDPN [36,37], measured by flow cytometry ( Figure 3B). Note that most ITGA6+/EPCAM+ hPGCLCs were also ALPL+/PDPN+ ( Figure 3B). However, we obtained a substantial variation in differentiation efficiencies to hPGCLCs varying from 2% to 30% using both sets of surface markers, not only between the Class II lines, but also in the male ( Figure 3C). Statistical analysis comparing PGCLC differentiation efficiencies showed neither significant differences between Class II and Class III lines nor between skin-derived Class II and Class III lines ( Figure 3C). Interestingly, the hPSCs lines that performed better in PGCLC differentiation were all derived from cells isolated from urine (presumably kidney epithelial cells) (Table 1). However, using ITGA6+/EPCAM+ as a readout to compare skin-derived with urine-derived Class II lines did not yield statistical significance ( Figure 3C). In contrast, comparing the mean PGCLC differentiation efficiencies between the five urine-derived hiPSCs and the five skin-derived hiPSCs regardless of their sexes and XCI classes revealed significantly higher efficiency in the urine lines ( Figure 3D). Although we do not have a Class III hPSCs derived from cells isolated from urine and hence their differentiation efficiency to hPGCLCs remains to be determined, we cannot exclude the possibility that differentiation efficiency to hPGCLCs may depend more on the tissue of origin (epigenetic memory) rather than XCI state. Class II lines did not yield statistical significance ( Figure 3C). In contrast, comparing the mean PGCLC differentiation efficiencies between the five urine-derived hiPSCs and the five skin-derived hiPSCs regardless of their sexes and XCI classes revealed significantly higher efficiency in the urine lines ( Figure 3D). Although we do not have a Class III hPSCs derived from cells isolated from urine and hence their differentiation efficiency to hPGCLCs remains to be determined, we cannot exclude the possibility that differentiation efficiency to hPGCLCs may depend more on the tissue of origin (epigenetic memory) rather than XCI state.  Since these four surface markers were also expressed at relatively high levels in hPSCs ( Figure S3A), we further confirmed their specificity by performing immunostaining on the hPGCLCs in EBs, together with established PGC markers SOX17, TFAP2C and POU5F1 [24,32,36]. In EBs, PDPN and ALPL specifically colocalized in TFAP2C+ hPGCLCs, POU5F1+ hPGCLCs and SOX17+ hPGCLCs ( Figure 3E and Figure S3B,C).

Influence of XCI State of hPSCs on the XCI State of Differentiated hPGCLCs
To evaluate the XCI state of hPGCLCs, we performed RNA-FISH on paraffin sections of hPGCLC-EBs targeting XIST and HPRT, with subsequent immunostaining for SOX17 to mark the hPGCLCs ( Figure 3F,G and Figure S4A). The Class II lines (F31, F99, F197 and F198) all maintained XIST clouds in the majority of SOX17+ hPGCLCs and SOX17− soma ( Figure 3F,G and Figure S4A), suggesting that hPGCLCs have not undergone XCR at this stage, or, alternatively, that XCR may be XIST independent in humans. Unexpectedly, most cells in hPGCLC-EBs, including SOX17+ hPGCLCs and SOX17− soma, from Class III F20 showed XIST foci ( Figure 3F,G), suggesting proper XCI.
Next, we investigated accumulation of H3K27me3 in hPGCLC-EBs ( Figure 3H,I and Figure S4B). In agreement with the presence of XIST in Class II hPGCLCs, we observed H3K27me3 foci in four of the Class II hPSCs (F197, F198, F99 and F31) and diffused nuclear staining Class II hPSC F30 ( Figure S4B). By contrast, Class III lines (F20 and H9) displayed low levels of H3K27me3 across the nucleus ( Figure S4B). Analysis of H3K9me3 in hPGCLC-EBs showed a similar pattern of expression in both SOX17+ hPGCLCs and SOX17− soma ( Figure S4C). Together, our results suggested that the XCI state is maintained in hPGCLCs differentiated from either Class II or Class III, with no evidence of XCR.

Re-Expression of XIST from Class III hPSCs in EBs
Several studies have reported that the loss of XIST in eroded hPSCs is irreversible and cannot be recovered upon differentiation [18,19,21,34]. We obtained similar results by differentiating Class III hPSCs in monolayer in the presence of FCS (Figure 2). However, when we investigated the XIST expression in hPGCLC-EBs, we observed XIST dots from Class III hPSCs (F20) (Figure 3F,G).
To understand whether 3D differentiation in EBs would result in upregulation of XIST in Class III hPSCs, we differentiated both Class II and Class III hPSCs in EBs in the presence of FCS and examined the XCI state ( Figure 4). Surprisingly, the majority of cells in the EBs generated with both Class II and Class III showed XIST dots ( Figure 4A,B). However, we noticed a difference in the size of the XIST clouds, larger in Class II EB cells compared to Class III EB cells ( Figure 4A). Nonetheless, H3K27me3 staining only showed enrichment in F99 and F31, but not F30 and the Class III hPSCs (Figure 4C), consistent with what we observed in hPGCLC-EBs ( Figure 3G-I and Figure S4B). Our results indicated that XIST could be upregulated in Class III hPSCs after 3D differentiation in EBs. We further asked whether culturing Class III hPSCs in 3D aggregates in pluripotency medium (TeSR E8) could result in re-expression of XIST. This was indeed the case in Class III F20 grown in 3D aggregates in pluripotency medium ( Figure S4D). Together, our results suggested that culturing XaXe hiPSCs (Class III) in 3D aggregates may contribute to restore aspects of erosion. in the EBs generated with both Class II and Class III showed XIST dots ( Figure 4A,B). However, we noticed a difference in the size of the XIST clouds, larger in Class II EB cells compared to Class III EB cells ( Figure 4A). Nonetheless, H3K27me3 staining only showed enrichment in F99 and F31, but not F30 and the Class III hPSCs ( Figure 4C), consistent with what we observed in hPGCLC-EBs ( Figures 3G-I and S4B). Our results indicated that XIST could be upregulated in Class III hPSCs after 3D differentiation in EBs. We further asked whether culturing Class III hPSCs in 3D aggregates in pluripotency medium (TeSR E8) could result in re-expression of XIST. This was indeed the case in Class III F20 grown in 3D aggregates in pluripotency medium ( Figure S4D). Together, our results suggested that culturing XaXe hiPSCs (Class III) in 3D aggregates may contribute to restore aspects of erosion. The differences observed between 2D and 3D culture could be explained by the activation of a hypoxic response during EB formation [38]. Hypoxic conditions are closer to the physiological conditions of the embryo and have been shown to maintain XIST expression and H3K27me3 accumulation in culture, in contrast to normoxic conditions [39,40]. To investigate whether hypoxia was the cause for Class III cells to re-express XIST when in 3D reaggregates, we next differentiated Class III hPSCs F20 in monolayer in the presence of FCS under hypoxic conditions (5% O2) and compared the XCI state with that of undifferentiated F20 hPSCs cultured under similar hypoxic conditions ( Figure S4E). In contrast to differentiation to EBs, differentiation of Class III hPSCs in monolayer under hypoxic conditions (5% O2) is not sufficient to upregulate the expression of XIST. The differences observed between 2D and 3D culture could be explained by the activation of a hypoxic response during EB formation [38]. Hypoxic conditions are closer to the physiological conditions of the embryo and have been shown to maintain XIST expression and H3K27me3 accumulation in culture, in contrast to normoxic conditions [39,40]. To investigate whether hypoxia was the cause for Class III cells to re-express XIST when in 3D reaggregates, we next differentiated Class III hPSCs F20 in monolayer in the presence of FCS under hypoxic conditions (5% O 2 ) and compared the XCI state with that of undif-ferentiated F20 hPSCs cultured under similar hypoxic conditions ( Figure S4E). In contrast to differentiation to EBs, differentiation of Class III hPSCs in monolayer under hypoxic conditions (5% O 2 ) is not sufficient to upregulate the expression of XIST.

Discussion
Dosage compensation of X-linked gene expression (XCI) is an essential process to counteract the male to female chromosomal imbalance. Later on during development, XCR occurs in the female germ line as part of the general epigenetic resetting that takes place prior to meiotic entry [41]. Newly developed protocols to generate hPGCLCs from hPSCs in vitro hold great promise to investigate the timing and the molecular mechanisms governing these unique epigenetic events [42].
We characterized the XCI state in six Class II hPSCs (XaXi) and two as Class III (XaXe) hPSCs. Interestingly, hPGCLCs generated from Class II hPSCs maintained the XCI state, suggesting that XCR did not occur in newly specified hPGCLCs. This is consistent with the notion that epigenetic remodeling in hPGCs occurs over a prolonged period of time, particularly because female gametogenesis is strongly asynchronous [27,43,44]. In mice, mPGCs lose H3K27me3 enrichment on the Xi during migration as early as E8.0, but full XCR does not complete until E14.5, long after their arrival in the gonads [45,46]. In 4-week of development (WD) human embryos, hPGCs already show low global 5-methylcytosine (5mC), while 5-hydroxymethylcytosine (5hmC) is high [25,26,47]. In addition, 4WD embryos show distinct H3K27me3 foci in (gonadal) somatic cells, but not in hPGCs, where it is globally enriched [27,48]. However, studies on allelic expression of X-linked transcripts have shown that a proportion (about 30%) of the early (POU5F1+) hPGCs present throughout development may still retain monoallelic expression of X-linked genes [27]. As human fetal samples are difficult to obtain due to ethical constraints and limited availability, complementary studies using novel in vitro gametogenesis models that would allow the development of hPGCLCs beyond that very early stage will be required.
In vitro generated hPGCLCs are likely equivalent to premigratory hPGCs just after specification [24,32,36]. mPGCLCs were shown to upregulate H3K27me3 and displayed low levels of H3K9me2 and 5mC [49,50]. When mPGCLC were cultured in conditions that promote their expansion, they showed further reduction in 5mC levels to a level equivalent to embryonic day (E)13.5 germ cells [51]. In agreement, hPGCLCs also show 5mC depletion [47,52] compared to hPSCs, but in contrast to mPGCLCs, no further reduction occurred upon expansion in culture [53]. A study comparing levels of DNA methylation between mPGCLCs and hPGCLCs reported that DNA demethylation occurs much slower in hPGCLCs and is incomplete [54]. Further decrease in 5mC in hPGCLC required prolonged coculture with mouse somatic ovary cells, which reduces 5mC to levels comparable to those in 7-10WD human germ cells [52]. In addition, while newly specified hPGCLC showed monoallelic X-linked expression, after 122 days in coculture with mouse gonads, hPGCLCs showed partial biallelic expression of some X-linked genes, albeit with low efficiency (in up to 20% of cells) [52]. The hPSC line used by Yamashiro and colleagues was Class III, based on its lack of XIST expression, and the XCR dynamics in the derived hPGCLCs may not be representative for the situation in vivo. It would be interesting to examine how Class I or II hPSCs-derived hPGCLCs would behave after prolonged cultured with mouse gonadal cells. In summary, mPGCLCs appear inherently capable of epigenetic remodeling, but hPGCLCs may require extrinsic factors originating from the migratory or gonadal niche. Identifying these factors and incorporating them in novel in vitro culture systems will be important to elucidate human XCR.
It has been reported that the erosion of the Xi is a source of variation in female hiPSC lines, and Class III iPSC (XaXe) lines have been associated with poorer in vitro differentiation and teratoma formation [17,18]. In this study, we did not observe a significant difference between the ability of Class II (XaXi) and Class III (XaXe) hiPSCs to undergo PGCLC differentiation or a correlation between the level heterogeneity and differentiation efficiency. However, we observed that the hiPSC lines generated from urine-derived cells showed an increasing trend to generate hPGCLCs compared to skin fibroblast-derived lines. hiPSCs are known to retain levels of epigenetic memory from the cell source of origin, which subsequently can result in bias towards differentiation into particular cell lineages [55][56][57][58][59]. We hypothesize that the origin of urine-derived cells (intermediate mesoderm, which also gives rise to the gonadal tissue) may facilitate differentiation into hPGCLCs. Variation in hPGCLC-differentiation efficiency between hiPSC lines has been reported by others [22,60], and higher efficiency was found to positively correlate with the levels of mesoderm markers EOMES, MIXL1 and T at the iMeLC stage. It remains to be determined whether hPSCs of mesodermal origin show higher expression of early mesenchymal markers at the iMeLC stage.
In line with previous reports, we found that Class II and Class III iPSC lines do not change their XCI state upon differentiation in monolayer. However, we did observe the re-expression of XIST in cells from Class III lines, when differentiated in (three types of) EBs/aggregates. Differentiation of pluripotent stem cells is a commonly used approach to study XCI dynamics in vitro, but the method of differentiation is generally thought not to affect the XCI outcome. Our results suggest that in fact, the method used should be carefully considered, as 3D culture conditions may affect XIST expression. Why 3D culture resulted in the upregulation of XIST in Class III hPSCs remains to be investigated. Intriguingly, the upregulation of XIST in Class III hPSCs EBs was not accompanied by the expected accumulation of H3K27me3. We speculate that this may be attributed to the low quantity of XIST transcripts in EBs from Class III cells, in contrast to the larger XIST clouds observed in EBs from Class II cells. H3K27me3 accumulation is dependent on XIST coating as it is catalyzed by EZH2 recruited by XIST; therefore, the limited XIST coating could explain why H3K27me3 accumulation was not observed in Class III EBs. In this respect, it would be interesting to investigate XIST and H3K27me3 after a longer period of 3D culture.
In 3D culture, diffusion of nutrients and gasses such as oxygen is more restricted compared to 2D monolayer, where each cell is in direct contact with the culture medium. Low oxygen levels were shown to be essential in deriving and maintaining hESCs in a pre-XCI state [39], but also to preserve the XCI fidelity of Class II hPSCs [40]. Moreover, H3K27me3 enrichment is oxygen-dependent, mediated by the oxygen-sensitive activity of histone demethylase KDM6A [61]. We were unable to observe upregulation of XIST in monolayer differentiation in hypoxia. However, culture in 3D aggregates may prove to be an important mechanism to restore XCI erosion. Differences in cell-extracellular matrix interaction and cell-cell interactions can greatly affect cell behavior [62][63][64]. Therefore, other properties specific to 3D aggregates that may influence XCI need to be considered.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/cells10092400/s1, Figure S1: Characterization of XCI hiPSC lines. Figure S2: Differentiation controls and characterization of FCS-induced monolayer differentiated cells. Figure S3: Expression of canonical hPGC markers in PGCLC-EBs. Figure S4: Characterization of XCI state in hPGCLCs and in Class III (F20) cells under Hypoxic and pluripotent 3D conditions. Table S1: List of antibodies used in this study.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of Leiden University Medical Center (protocol code P13.080 and date of approval 16 July 2015).
Informed Consent Statement: Informed consent was obtained from all subjects providing material for the generation of hiPSCs used in the study.
Data Availability Statement: Not applicable.