In mammals, the pre-gastrula proximal epiblast gives rise to primordial germ cells (PGCs) or somatic precursors in response to BMP4 and WNT signaling. Entry into the germline requires activation of a naïve-like pluripotency gene regulatory network (GRN). Recent work has shown that suppression of OTX2 expression in the epiblast by BMP4 allows cells to develop a PGC fate in a precise temporal window. However, the mechanisms by which OTX2 suppresses PGC fate are unknown. Here, we show that, in mice, OTX2 prevents epiblast cells from activating the pluripotency GRN by direct repression of Oct4 and Nanog. Loss of this control during PGC differentiation in vitro causes widespread activation of the pluripotency GRN and a deregulated response to LIF, BMP4 and WNT signaling. These abnormalities, in specific cell culture conditions, result in massive germline entry at the expense of somatic mesoderm differentiation. Increased generation of PGCs also occurs in mutant embryos. We propose that the OTX2-mediated repressive control of Oct4 and Nanog is the basis of the mechanism that determines epiblast contribution to germline and somatic lineage.
Primordial germ cells (PGCs) represent the founder cells of the germ cell lineage, which ensures the transmission of genetic and epigenetic information across the generations (Wylie, 1999; Surani, 2001; Saitou and Yamaji, 2012; Johnson et al., 2003; Johnson and Alberio, 2015). In mice, both the germline and the somatic lineage, the latter referring to embryonic and extra-embryonic mesoderm (Saitou et al., 2005), originate from the proximal-posterior epiblast of pre-gastrula embryos (Lawson et al., 1999; Saitou, 2009). The generation of PGCs is controlled by a precise mechanism determining the segregation of the germline from the somatic lineage (Ohinata et al., 2009; Surani et al., 2007; Hayashi et al., 2007). The acquisition of PGC identity is associated with loss of somatic mesoderm fate (Günesdogan and Surani, 2016; Saitou et al., 2002). In mouse, germline and somatic lineage originate in response to BMP and WNT signaling (Lawson et al., 1999; Ohinata et al., 2009; Senft et al., 2019). BMP4-mediated activation of the transcription factor (TF) genes Blimp1 (also known as Prdm1), Ap2γ (also known as Tfap2c) and Prdm14 is required for germline entry and suppression of somatic mesoderm identity, which is defined by the expression of T (also known as Brachyury) and Hoxb1 (Magnúsdóttir et al., 2013; Ohinata et al., 2005; Kurimoto et al., 2008; Ancelin et al., 2006; Vincent et al., 2005; Weber et al., 2010; Yamaji et al., 2008; Nakaki et al., 2013). Co-expression of Blimp1, Ap2γ and Prdm14 defines the identity of germ cells in vivo and in vitro. Further work indicated that WNT3 is also required for PGC fate (Aramaki et al., 2013). Germline entry and development of PGCs also requires the activation of a naïve-like pluripotency gene regulatory network (GRN) defined by the TFs Oct4 (also known as Pou5f1), Sox2 and Nanog and representing a unique feature of the PGC unipotent state (Saitou and Yamaji, 2012; Hayashi et al., 2007; Hackett and Surani, 2014; Leitch et al., 2013; Smith, 2017; Okamura et al., 2008; Yamaguchi et al., 2009; Chambers et al., 2007; Zhang and Chambers, 2019; Campolo et al., 2013; Kehler et al., 2004; Yeom et al., 1996). Importantly, Nanog activation in epiblast-like cells (EpiLCs), a cellular state very similar to pre-gastrulation epiblast (Hayashi et al., 2011; Smith, 2017), is sufficient for germ cell induction (Murakami et al., 2016), and Oct4 expression in PGCs requires a transcriptional regulatory switch from the proximal enhancer (PE) to the distal enhancer (DE) (Choi et al., 2016; Wu and Schöler, 2014). The study of germline development has been fuelled by the realization of a stepwise in vitro culture system generating PGC-like cells (PGCLCs) from germline-competent EpiLCs induced, in turn, from naïve embryonic stem cells (ESC) (Hayashi et al., 2011). PGCLCs and PGCs share molecular identity, epigenetic reprogramming and spermatogenic capacity (Hayashi et al., 2011, 2012, 2017). The TF OTX2, a key determinant of brain development (Acampora et al., 1995), is also expressed in pluripotent cells in vivo and in vitro (Acampora et al., 2013). OTX2 is required to: antagonize naïve pluripotency; promote transition of ESCs from naïve to formative pluripotency; specify the heterogeneous identity of ESCs through reciprocal antagonism with NANOG; and define in vitro and in vivo naïve-primed intermediate states of pluripotency (Acampora et al., 2013, 2016, 2017; Buecker et al., 2014; Yang et al., 2014; Neagu et al., 2020). Moreover, in pluripotent cells OTX2 may also bind to the enhancer region of Oct4, Nanog and Sox2 (Acampora et al., 2016). Recent work has revealed that OTX2 is a novel crucial determinant controlling germline entry of formative epiblast cells in mouse (Zhang et al., 2018; Zhang and Chambers, 2019; Laird, 2018). Indeed, by restricting germline entry to a limited number of epiblast-derived cells, OTX2 controls segregation of germ cells from somatic precursors. Here, we hypothesized that, mechanistically, entry into germline requires the release by OTX2 of a repressive state preventing activation of the pluripotency GRN in epiblast-derived cells. Our data indicate that, in EpiLC-derived cells, loss of OTX2 binding to Oct4 or Nanog enhancer region causes widespread activation of the pluripotency GRN, increased germline entry and suppression of somatic fate. Generation of PGCs is increased also in mutant embryos. This study uncovers a novel mechanism by which OTX2 regulates epiblast contribution to germline and somatic mesoderm.
OTX2, OCT4 and NANOG expression during PGC development and PGCLC differentiation
To obtain a detailed view of the temporal appearance of OTX2, OCT4 and NANOG relative to germline and somatic differentiation determinants, embryos were analysed by immunofluorescence (Fig. 1A). At embryonic day (E) 6.5, AP2γ and BLIMP1 were co-expressed in a subset of fragilis+ (also known as IFITM3) cells, which included incipient PGCs (Zhang et al., 2018). From E7 to E7.5 these three markers gradually became colocalized in PGCs (yellow arrows in Fig. 1A). At E6.5 and E7, OTX2 was expressed in some fragilis+-AP2γ− cells (white arrow in Fig. 1A), but became undetectable at E7.5 in fragilis+ cells. During this time window, OCT4 and NANOG were co-expressed in fragilis+ cells, although at E6.5 NANOG was detected at variable levels (Fig. 1A). T was initially expressed throughout the fragilis+ field but by E7.5 disappeared from some fragilis+ cells (Fig. 1A). These data show that AP2γ and BLIMP1 become expressed within fragilis+ cells only when OTX2 has been turned off, suggesting that expression of OTX2 in fragilis+ cells may prevent precocious activation of AP2γ and BLIMP1.
To determine whether these temporal changes in expression were mirrored in vitro, we analysed EpiLC-derived cell aggregates during PGCLC differentiation by using a modification of the established procedure (Hayashi et al., 2011) in which the concentrations of BMP4, BMP8A, SCF and EGF were reduced tenfold (protocol 1) as used previously (Zhang et al., 2018) (Fig. 1B). AP2γ+ cells were first detected at day (d) 0.5; at d1 some AP2γ+ cells co-expressed BLIMP1 and these increased in number up to d2; OCT4 was expressed in all cells up to d1 with expression remaining high in AP2γ+ cells at d2; NANOG was activated at high levels predominantly in AP2γ+-T− cells; and T showed widespread activation beginning at d1 prevalently in AP2γ− cells expressing low levels of NANOG (Fig. 1C). A comparison of OTX2 expression with that of OCT4, NANOG, AP2γ and T showed that, although markedly downregulated between d0 and d1 (Zhang et al., 2018) and virtually undetectable at d2 (Fig. 1C), OTX2 remained detectable at d0.5 and d1 in numerous cells co-expressing OCT4 and low levels of NANOG; from d1 T was also detectable (Fig. 1C). Notably, at d0.5, OTX2 was absent from AP2γ+ cells. This analysis confirms and extends previous findings (Zhang et al., 2018) indicating that, during the germline and somatic differentiation time window, OTX2 is efficiently repressed in epiblast-derived or EpiLC-derived cells as they begin to differentiate towards PGCs or PGCLCs, while it is maintained for longer in somatic precursors. This suggests that differential OTX2 repression may prevent excessive germline differentiation.
OTX2 binding to Nanog or Oct4 enhancer region regulates the generation of PGCLCs and somatic cells
To assess the hypothesis that differential OTX2 repression governs the extent of germline entry, we analysed ESCs in which specific OTX2 binding site(s) (Obs) were mutagenized. NanogΔObs3 ESCs have lost the strongest Obs (Obs3) in the Nanog enhancer (Acampora et al., 2016). In addition, we derived Oct4ΔObs ESC lines mutagenized in the three Obs located in the enhancer region of Oct4 (Fig. S1A-D): Obs1 within the PE, Obs2 at the 5′ of the PE and Obs3 at the 3′ of the DE (Fig. S2A). Compared with wild type, the OTX2-binding activity in Oct4ΔObs ESCs and d2 EpiLCs was virtually abrogated (Fig. S2B). However, wild-type and Oct4ΔObs ESCs showed no differences in: self-renewal at clonal density (Fig. S2C; Table S1); expression of OCT4, NANOG, OTX2 and OCT6 (also known as Pou3f1) (Fig. S2D); ability to convert into the naïve state (Fig. S2E); steady state LIF, FGF, WNT and BMP signaling (Fig. S2F); acute responsiveness to LIF (Fig. S2G) and FGF2 (Fig. S2H); and chimaera-forming capacity (Fig. S2I). Furthermore, analysis of Oct4ΔObs and NanogΔObs3 EpiLCs at d2 showed uniform high level of OCT4, OTX2 and OCT6, absence of NANOG and T (Fig. S2J), and similar steady state levels of phospho (p)-ERK1,2, p-SMAD1,5,8, p-β-catenin, active β-catenin and p-SMAD2 (Fig. S2K). Together, these results indicate that Oct4ΔObs ESCs and both Oct4ΔObs and NanogΔObs3 EpiLCs are indistinguishable from wild-type cells.
To investigate whether Oct4ΔObs and NanogΔObs3 (subsequently referred to as ΔObs) ESCs are affected in their propensity to differentiate into germline cells, we employed the above-mentioned protocol 1 as a PGCLC differentiation method (Fig. 2A). FACS analysis (Fig. S3A-C) for SSEA1 (also known as Fut4) and CD61 (also known as Itgb3), which are co-expressed in PGCLCs, indicated that compared with wild type, ΔObs mutants generated a higher percentage of SSEA1+-CD61+ cells at d6, approaching that of Otx2 null (Otx2KO) cells (Zhang et al., 2018) (Fig. 2B). This finding was confirmed with independently generated ΔObs cell lines (Fig. 2C). To determine whether increased generation of PGCLCs was presaged by altered expression of TFs, cytospin preparations of d2 cell aggregates were analysed. Relative to wild type, ΔObs mutants showed an increased proportion of cells expressing high levels of OCT4, NANOG and BLIMP1, and a decreased proportion of T+ cells (Fig. 2D; Fig. S3D,E; Table S2). Immunohistochemistry analysis of d2 cell aggregates supported cytospin data (Fig. 2E). Real-time quantitative PCR (RT-qPCR) analysis indicated that the expression levels of Oct4, Nanog, Ap2γ and Blimp1 were also increased at d2 in mutants (Fig. 2F). For Nanog and Blimp1 this was apparent at d0.5; for Ap2γ it was evident at d0.25 (Fig. 2F). No difference in Otx2 expression was detected between wild type and ΔObs mutants, whereas the expression levels of T and Hoxb1 were substantially diminished in ΔObs mutants and virtually extinguished in Otx2KO cells (Fig. 2F,G). Previous studies reported that histone H3 tri-methyl lysine 9 (H3K9me3) is enriched at poised enhancers and may distinguish poised and active states for the DE and PE of Oct4 (Wu and Schöler, 2014). Compared with wild type, in Oct4ΔObs d2 cell aggregates, the H3K9me3 level was reduced on the DE (Fig. 2H), suggesting a repressive role for OTX2 on DE activity. Increased generation of SSEA1+-CD61+ cells was observed in mutants also when PGCLC differentiation was initiated from ESCs maintained in serum plus LIF (protocol 2) (Fig. 2I,J; Fig. S4A). Therefore, in contrast to wild type and similar to Otx2KO cells, germline differentiation of ΔObs mutants is increased at the expense of somatic fate.
Cytokine requirements for germline and somatic differentiation of ΔObs mutants
Otx2KO EpiLCs can undergo germline entry in the absence of cytokines (Zhang et al., 2018). To determine whether ΔObs EpiLCs possessed the same capacity, we employed a cytokine-free differentiation protocol (protocol 3) (Fig. 3A). FACS analysis at d6 revealed moderate generation of PGCLCs only in Otx2KO cells (Fig. 3B). Cell counting and immunohistochemistry experiments showed that at d2 ΔObs mutants significantly increased the percentage of total cells expressing high OCT4 and high NANOG, and decreased the proportion of T+ cells (Fig. 3C; Figs S3D,E and S4B; Table S2). The BLIMP1+ cell fraction was expanded only in Otx2KO cells. RT-qPCR analysis showed that repression of Otx2 was equally delayed in all genotypes, and that expression of Nanog, Oct4, Ap2γ , Blimp1 and T at d2 was coherent with cell counting data (Fig. 3D). At earlier time points, the expression of these factors was similar in all genotypes except for Blimp1 and Ap2γ, which were increased and precocious in Otx2KO cells (Zhang et al., 2018) (Fig. 3D). Hoxb1 was inefficiently upregulated in mutants (Fig. 3E). Therefore, in the absence of cytokines, ΔObs mutants, in contrast to Otx2KO cells, are unable to promote germline differentiation. This suggests that cytokine-independent germline differentiation of ΔObs mutants may require OTX2 functions in addition to those depending on OTX2 binding to Oct4 or Nanog enhancer region. LIF is required for proliferation and survival of germ cells (Hayashi et al., 2011; Farini et al., 2005; Koshimizu et al., 1996; Cheng et al., 1994) and to maintain pluripotency and promote self-renewal of ESCs (Matsuda et al., 1999; Niwa et al., 2009). As these functional properties may have an impact on germline differentiation of ΔObs and Otx2KO mutants, we analysed cell aggregates cultured with LIF only (protocol 4) (Fig. 4A). Compared with wild type, ΔObs and Otx2KO mutants generated more PGCLCs at d6 (Fig. 4B), and at d2 showed increased percentage of total cells expressing high OCT4, high NANOG and BLIMP1, and a virtual absence of T+ cells (Fig. 4C; Figs S3D,E and S4C; Table S2). RT-qPCR analysis showed that Otx2 was efficiently repressed in all genotypes (Fig. 4D), while, compared with wild type, Oct4 expression was maintained at high levels in mutant cell aggregates, Blimp1, Ap2γ and Nanog expression was activated earlier, and T and Hoxb1 expression was severely decreased (Fig. 4D,E). Therefore, LIF is sufficient to promote significant germline differentiation in ΔObs mutants and, with a substantially higher efficiency, in Otx2KO cells.
To further investigate the contribution of LIF to PGCLC differentiation, we used a protocol in which LIF was the only component omitted from the cytokine cocktail (protocol 5) (Fig. 4F). Compared with wild type, the number of PGCLCs at d6 was increased in ΔObs and Otx2KO mutants (Fig. 4G). At d2, the percentage of total cells expressing high OCT4, high NANOG and BLIMP1 was significantly increased in all mutants, while the number of T+ cells was strongly affected only in Otx2KO cells (Fig. 4H; Figs S3D,E and S4D; Table S2). RT-qPCR analysis showed that repression of Otx2 and activation of Nanog, Blimp1 and Ap2γ were both less efficient than in protocols 1 and 4 (Fig. 4I). Compared with wild type, T and Hoxb1 activation was weakened in mutants (Fig. 4I,J). Taken together, findings from protocols 4 and 5 indicate that LIF alone provides a more effective stimulation of PGCLC differentiation than the remaining cytokines, and is particularly effective in suppressing somatic fate.
Previous work showed that endogenous WNT activity contributes to Otx2 repression in different contexts and can be mimicked by the glycogen synthase kinase 3 inhibitor (GSK3) CHIR99021 (CHIR) (Acampora et al., 2016; Neagu et al., 2020; Zhang et al., 2018). We therefore assessed whether CHIR alone may stimulate the generation of PGCLCs and somatic mesoderm (protocol 6) (Fig. 5A). Only Otx2KO cells produced a substantial proportion of SSEA1+-CD61+ cells at d6 (Fig. 5B). Compared with wild type and Otx2KO mutants, in ΔObs mutants CHIR treatment generated intermediate percentages of cells expressing high OCT4, high NANOG and T (Fig. 5C; Figs S3D,E and S5A; Table S3), and intermediate expression levels of Oct4, Nanog, T and Hoxb1 (Fig. 5D,E). However in wild type and ΔObs mutants, the percentage of BLIMP1+ cells and the expression levels of Blimp1 and Ap2γ were both similar to those described in cytokine-free culture conditions (Fig. 5C,D). RT-qPCR analysis showed that Otx2 repression was comparable with that observed in protocol 5 (Fig. 5D). These results reveal that stimulation of the WNT pathway induces robust PGCLC differentiation only in Otx2KO cells, while in ΔObs and Otx2KO mutants it efficiently suppresses somatic differentiation.
LIF, WNT and BMP4 synergism promotes germline entry in ΔObs mutants
To determine whether WNT stimulation may synergize with LIF signaling during PGCLC differentiation, as reported for ESCs (ten Berge et al., 2011; Karwacki-Neisius et al., 2013; Ogawa et al., 2006), we assessed the effects of simultaneous addition of LIF and CHIR (protocol 7) (Fig. 5F). Compared with wild type, the generation of PGCLCs at d6 was substantially higher in ΔObs and Otx2KO mutants (Fig. 5G). This correlated with increased percentage of total cells expressing high OCT4, high NANOG and BLIMP1, severe decrease of T+ cells (Fig. 5H; Figs S3D,E and S5B; Table S3), and precocious and robust expression of Nanog, Blimp1 and Ap2γ (Fig. 5I). In addition, T and Hoxb1 were not activated and Otx2 was efficiently repressed (Fig. 5I,J). To explore whether low levels of stimulation of other pathways could likewise potentiate the effects of LIF, we assessed differentiation in the presence of LIF and a low dose of BMP4, BMP8a, SCF and EGF (protocol 8) (Fig. 6A), the concentrations of which were ten times lower than those used in protocols 1 and 5. Compared with wild type, ΔObs and Otx2KO mutants showed that PGCLC generation at d6 (Fig. 6B) and the number of cells expressing high OCT4, high NANOG and BLIMP1 at d2 were both markedly increased, while the number of T+ cells was considerably diminished (Fig. 6C; Figs S3D,E and S5C; Table S3). RT-qPCR assays showed expression profiles similar to those obtained with LIF plus CHIR (Fig. 5I,J). Compared with protocol 8, cell aggregates supplemented only with low concentrations of BMP4, BMP8A, SCF and EGF showed that the generation of SSEA1+-CD61+ cells was reduced in Otx2KO cells and virtually absent in ΔObs mutants (Fig. S6A,B). Therefore, data from protocols 7 and 8 indicate that, in ΔObs mutants, LIF addition to CHIR or to a minimal dose of BMP4, BMP8A, SCF and EGF promotes efficient germline entry and severely affects somatic lineage. Based on these results, we asked whether combining LIF with CHIR and low doses of BMP4, BMP8A, SCF and EGF (protocol 9) could further augment PGCLC differentiation (Fig. 6F). Interestingly, the number of PGCLCs (Fig. 6G), the percentage of total cells expressing high OCT4, high NANOG, BLIMP1 and T (Fig. 6H; Figs S3D,E and S5D; Table S3), and the expression profile exhibited by Otx2, Oct4, Nanog, Ap2γ, Blimp1, T and Hoxb1 (Fig. 6I,J) were all similar to those obtained using protocol 1. Furthermore, the level of H3K9me3 detected in Oct4ΔObs cells was decreased on the DE and increased on the PE (Fig. 6K), suggesting that, as shown for protocol 1, OTX2 is required to prevent transition of the DE status from a poised to an active condition. To further assess the similarities between mature PGCLCs generated by protocols 1 and 9, expression of PGCLC-specific mRNAs was examined by RT-qPCR. SSEA1+-CD61+ cells purified at d6 using protocols 1 and 9 showed very similar expression of 11 PGCLC-specific mRNAs in both wild-type and mutant PGCLCs (Fig. S6C). Therefore, in ΔObs mutants, LIF, CHIR and low doses of BMP4, BMP8a, SCF and EGF may cooperatively induce high levels of PGCLCs counterbalanced by a severe decrease in somatic cells. Through this cooperation, ΔObs mutants exhibit close similarity with Otx2KO cells, a finding previously reported only for protocol 1. The significance of these data was reinforced by a further control experiment showing that, in the presence of CHIR and low doses of BMP4, BMP8A, SCF and EGF, the generation of PGCLCs was considerably increased only in Otx2KO cells (Fig. S6D,E). To further investigate the role of LIF, the effect of titrating the LIF concentration on PGCLC differentiation was assessed. In GK15 media, the percentage of SSEA1+-CD61+ cells detected in wild-type cells was low but increased with increasing LIF concentrations, up to about 6% (Fig. S7A). In contrast, the mutant cell aggregates showed an increased yield of SSEA1+-CD61+ cells at 1000 U/ml LIF, compared with 300 U/ml, but this was not further increased by raising the LIF concentration to 3000 U/ml (protocol 4) (Fig. S7B). When a similar LIF titration was performed using the fixed concentrations of CHIR and low doses of BMP4, BMP8A, SCF and EGF (protocol 9; Fig. S7C), a similarly patterned response was observed in both wild-type cells and in all mutant cell lines (Fig. S7D). Although the absolute proportions of SSEA1+-CD61+ cells was increased in each cell line, a saturating response to 1000 U/ml LIF was seen in all the mutants but not in the wild-type cells (Fig. S7B), which increased the percentage of PGCLCs up to 11% in response to 3000 U/ml of LIF.
Temporal expression changes during early PGCLC differentiation of ΔObs mutants
Previous data showed that the OTX2 repressive control of Oct4 and Nanog expression defines commitment of EpiLC-derived cells to germline or somatic lineage. To investigate this in greater depth, we analysed the combined expression of OTX2, OCT4, NANOG, SOX2, AP2γ and T in wild-type and ΔObs cell aggregates at d0.5 and d1 using protocol 1. OCT4 and SOX2 were uniformly co-expressed in wild-type and ΔObs cells, while NANOG was detected mostly at low level in wild type and at high level in ΔObs cells (Fig. S8). Although, compared with EpiLCs, the Otx2 mRNA was downregulated at d0.5 and d1 (Fig. 2F), OTX2 was detectable in numerous cells co-expressing OCT4 and low NANOG in wild-type aggregates but co-expressing OCT4 and prevalently high NANOG in ΔObs mutants (Fig. S8). All NANOG+ cells co-expressed SOX2 (Fig. S8). In wild-type cells at d0.5, AP2γ was detected in very few OTX2− cells co-expressing high NANOG, whereas in ΔObs mutants, AP2γ was activated in a higher number of cells all co-expressing high NANOG and, frequently, also OTX2 (Fig. S8), suggesting that, in ΔObs mutants, numerous OTX2+ cells are permissive for AP2γ activation. However, by d1, AP2γ co-expression with OTX2 was lost, likely because of further downregulation of OTX2 in AP2γ+ cells. Analysis of T showed that at d1 in wild-type cells, most of the OTX2+ cells co-expressed T and low NANOG, while in ΔObs mutants the few T+ cells co-expressed only low NANOG (Fig. S8). These data suggest that OTX2-dependent widespread activation of the pluripotency GRN determines, in mutants, a precocious and extensive recruitment of EpiLC-derived cells for germline entry rather than for somatic fate.
The Obs2 in the Oct4 enhancer plays a major role in germline and somatic differentiation
To determine which Obs in the Oct4 enhancer is functionally relevant, we restored the wild-type sequence of Obs2, which exhibited the highest OTX2 binding activity, in the Oct4ΔObs mutant and generated the Oct4ΔObs1,3 ESC line (Fig. S1E,F). ChIP analysis showed the expected recovery in OTX2 binding specifically at Obs2 (Fig. S9A,B). FACS analysis showed that the percentage of PGCLCs generated by using protocol 1 was similar in wild type and Oct4ΔObs1,3 mutant (Fig. S9C). Immunohistochemistry and cell counting at d2 (Figs S9D,E and S3D,E; Table S2) confirmed the similarity of Oct4ΔObs1,3 to wild-type cells, suggesting that OTX2 binding to Obs2 modulates the OCT4 requirement for germline and somatic differentiation.
Enforced OTX2 expression inefficiently suppresses germline entry in ΔObs mutants
Previous work has shown that tamoxifen (Tx)-mediated nuclear translocation of OTX2-ERT2 fusion protein in wild-type and Otx2KO cells suppresses PGCLC differentiation (Zhang et al., 2018). We reasoned that, in contrast to wild-type and Otx2KO cells, nuclear translocation of OTX2-ERT2 in ΔObs mutants should not affect functions depending on OTX2 binding to Oct4 or Nanog enhancer region. To test this hypothesis, we generated E14;Otx2ERT2, Oct4ΔObs;Otx2ERT2, NanogΔObs3;Otx2ERT2 and Otx2KO;Otx2ERT2 ESC lines expressing a similar level of OTX2-ERT2 (Fig. S1G,H). In these cell lines, Tx exposure for 1 h induced efficient OTX2-ERT2 nuclear translocation (Fig. S1I). Two independent clones for Oct4ΔObs;Otx2ERT2 and NanogΔObs3;Otx2-ERT2 mutants were compared with E14;Otx2ERT2 and Otx2KO;Otx2ERT2 cell lines using protocol 1. Without Tx, the generation of PGCLCs was similar to that of the same parental cell lines (Fig. 7A). In Tx-treated cell aggregates (from d0 to d2 of the cell aggregation phase), the generation of PGCLCs was suppressed in E14;Otx2ERT2 and Otx2KO;Otx2ERT2 cell lines, severely affected in Oct4ΔObs;Otx2ERT2 cells, but only partially impaired in NanogΔObs3;Otx2ERT2 cells (Fig. 7B). Consistent with these findings, in Tx-treated d2 cell aggregates the generation of AP2γ+-BLIMP1+ cells was virtually abolished in E14;Otx2ERT2 and Otx2KO;Otx2ERT2 cells, decreased in Oct4ΔObs;Otx2ERT2 cells, but less affected in NanogΔObs3;Otx2ERT2 cell aggregates (Fig. 7C). Conversely, the generation of T+ cells was severely impaired only in NanogΔObs3;Otx2ERT2 cells (Fig. 7C). Interestingly, in E14;Otx2ERT2, Otx2KO;Otx2ERT2 and Oct4ΔObs;Otx2ERT2 cell lines, OCT4 expression was low in all cells including those AP2γ+ (Fig. 7C), which did not activate NANOG and SOX2 (yellow and white arrows in Fig. 7C). In contrast, in the NanogΔObs3;Otx2ERT2 mutant, a large fraction of AP2γ+ cells showed high expression of OCT4 (pink arrows in Fig. 7C), NANOG and SOX2 (Fig. 7C). These data suggest that OTX2 binding to the Nanog enhancer is required to efficiently suppress germline differentiation and induce somatic fate. To investigate the possibility that OTX2 may repress germline determinants, we analysed whether BLIMP1+ cells co-expressed nuclear OTX2-ERT2. BLIMP1 and OTX2-ERT2 were colocalized in Tx-treated cells of all genotypes (Fig. 7D), which, in turn, also co-expressed AP2γ (Fig. 7C). In addition, we found that levels of Prdm14 and Blimp1 mRNAs, which are restricted to PGCLCs, showed a similar expression pattern in all cell lines (Fig. 7E). These data suggest that OTX2 does not directly repress Blimp1, Ap2γ and Prdm14 during germline differentiation.
ΔObs embryos generate supernumerary PGCs
To determine whether PGC development was affected also in vivo, we analysed ΔObs homozygous embryos. Immunohistochemistry analysis of ΔObs homozygous embryos at E6.7 and E7.5 revealed that expression of BLIMP1, AP2γ, fragilis, NANOG, OCT4, T and OTX2 was similar to wild type (Fig. 8A). To determine the number of BLIMP1+-fragilis+ PGCs, we analysed wild-type and ΔObs embryos at E7.5 and E8.5. All sections, including BLIMP1+-fragilis+ PGCs were captured for cell counting (Fig. S10). Compared with wild type, ΔObs mutants exhibited about 1.5 times more PGCs at both E7.5 and E8.5 (Fig. 8B; Table S4). Ectopic PGCs were not identified. Therefore, binding of OTX2 to the Oct4 or Nanog enhancer region is required to limit the extent of germline differentiation in vivo.
Differentiation of the epiblast into germline and somatic cells both in vivo and in vitro requires the action of BMP4, but the interplay with other signals, particularly WNT3, is also notable (Lawson et al., 1999; Saitou and Yamaji, 2012; Günesdogan and Surani, 2016; Aramaki et al., 2013). Previous studies have reported that LIF is required to sustain proliferation and survival of germ cells (Farini et al., 2005; Koshimizu et al., 1996; Cheng et al., 1994). The initial transition of epiblast-derived cells into the germline-competent state requires dismantling of the formative pluripotency GRN, characterized by high Otx2 expression and PE-driven expression of Oct4, alongside activation of the PGC pluripotency GRN, characterized by extinction of Otx2, re-expression of Nanog, high Sox2 expression and DE-driven expression of Oct4 (Zhang and Chambers, 2019; Smith, 2017; Choi et al., 2016). Conversely, somatic differentiation requires transition of formative pluripotency into a primed state characterized by cell heterogeneity in expression of pluripotency TFs, including OTX2, NANOG and SOX2, as well as somatic gene expression.
Recent work has revealed that, in a germline permissive environment, downregulation of Otx2 determines the number of epiblast-derived cells entering the germline differentiation program (Zhang et al., 2018; Zhang and Chambers, 2019; Laird, 2018). Nevertheless, the mechanisms by which OTX2 regulates the contribution of epiblast cells to germline and somatic lineage are unknown. As OTX2 binds to regulatory regions of Oct4 and Nanog in EpiLCs (Acampora et al., 2016), we hypothesized that germline entry may require release of OTX2 repression on the PGCLC pluripotency GRN. Our data indicate that, through this binding, OTX2 influences germline and somatic fate choice of EpiLC-derived cells in response to LIF and high doses of BMP4, BMP8A, EGF and SCF. Indeed, ΔObs and Otx2KO cell aggregates exhibit comparable phenotypes in terms of widespread activation of the pluripotency GRN, increased generation of PGCLCs and suppression of somatic fate. However, ΔObs and Otx2KO mutants do not share all aspects of their phenotype in all protocols. Importantly, without cytokines, ΔObs mutants, unlike Otx2KO cells (Zhang et al., 2018), do not generate PGCLCs. This suggests that OTX2 blocks cytokine-independent germline differentiation through functions other than those operating via binding to Oct4 or Nanog. Such additional OTX2 functions could involve repression of PGCLC TFs. However, our evidence suggests that OTX2 does not directly repress Blimp1, Prdm14 and Ap2γ.
To determine the signaling pathway modulations that may provide equally efficient PGCLC differentiation in ΔObs and Otx2KO mutants, we analysed the effects of LIF, WNT or BMP4 alone (protocols 4-6). None of these protocols raises the efficiency of PGCLC generation by ΔObs cells to the level achieved in Otx2KO mutants. However, the combined action of LIF, CHIR and a low dose of BMP4, BMP8A, EGF and SCF suppresses the OTX2 functions that hamper ΔObs mutants from highly efficient, Otx2KO-like, germline entry to the same extent as in culture conditions with LIF and high doses of BMP4, BMP8A, EGF and SCF. These data suggest that the generation of germ cells may be quantitatively correlated to specific combinations of cytokines whose inducing efficiency may also be influenced by the genetic state of the responding cells. In this context it would be interesting to assess whether, in embryos, the proximal-posterior epiblast includes restricted subdomains where identity and fate of epiblast-derived cells are precisely regulated by different concentrations of LIF, BMP and WNT signaling molecules.
Together with dose-dependent experiments (Fig. S7), our data indicate that LIF has an important role in PGCLC differentiation. We suggest that LIF signaling may improve the generation of PGCLCs by enhancing the proliferation/self-renewal of those cells that have released OTX2-dependent suppression of the pluripotency GRN. This interpretation may explain the beneficial effect of LIF on the generation of PGCLCs in all protocols where it is included, and, in addition, supports the similarity with ESCs where LIF is required to maintain pluripotency and self-renewal (Matsuda et al., 1999; Niwa et al., 2009), even through synergism with WNT and BMP4 signaling (ten Berge et al., 2011; Karwacki-Neisius et al., 2013; Ogawa et al., 2006; Ying et al., 2003, 2008). Notably, as Otx2KO ESCs can self-renew in the absence of LIF (Acampora et al., 2013), this might explain the propensity of Otx2KO ESCs to enter the germline in the absence of LIF. We propose that, in wild-type cells, LIF, BMP4 and WNT cooperate to repress Otx2 early during the cell aggregation phase (Fig. 9). This generates two types of EpiLC-derived cells: those lacking OTX2 (OTX2−) (right in Fig. 9A) and those transiently retaining OTX2 (OTX2+) (left in Fig. 9A). OTX2− cells cannot repress the PGCLC pluripotency GRN; they enter the unipotent germline state, are expanded by LIF or LIF in synergism with BMP4 and/or WNT, and differentiate into PGCLCs (right in Fig. 9A). Conversely, OTX2-mediated repression of the PGCLC pluripotency GRN in OTX2+ cells promotes entry into primed pluripotency and activation of somatic determinants leading to differentiation into somatic mesoderm (left in Fig. 9A). Our data do not exclude the possibility that BMP4 and WNT may also facilitate activation of somatic mesoderm genes (left in Fig. 9A). In ΔObs mutants (Fig. 9B), Otx2 repression by LIF, BMP4 and WNT, and generation of OTX2− (right in Fig. 9B) and OTX2+ (left in Fig. 9B) EpiLC-derived subtypes is comparable with wild-type cells. In ΔObs mutants, the OTX2− EpiLC-derived cell fraction corresponds to the wild-type OTX2− cell fraction, is naturally competent for germline differentiation and is unaffected by loss of OTX2 binding to Oct4 or Nanog enhancer region (right in Fig. 9A,B). However, in the OTX2+ cell fraction, progression towards the somatic fate is impaired by the inability of OTX2 to repress the pluripotency GRN (left in Fig. 9B). Consequently, cells that activate the pluripotency GRN may be further expanded by cytokines and commit to germline entry. These early events result in an increased recruitment of EpiLC-derived cells for germline entry before somatic differentiation is initiated with a consequent switch from somatic to germline fate (Fig. 9B). According to this model, nuclear translocation of OTX2-ERT2 can revert the germline to somatic fate switch in all cell lines except NanogΔObs3;Otx2ERT2 cells (Zhang et al., 2018), indicating that OTX2 binding to the Nanog enhancer region efficiently suppresses the PGCLC pluripotency GRN. Interestingly, compared with Oct4ΔObs;Otx2ERT2 cells, NanogΔObs3;Otx2ERT2 mutants produced a higher proportion of presumptive PGCLCs in the presence of Tx, indicating that OTX2 suppression of Nanog transcription may play a greater role in preventing PGCLC differentiation than suppression of Oct4 transcription. A possible explanation may reside in the expression baseline exhibited by Nanog and Oct4 in EpiLCs. Indeed for Nanog, transition from the EpiLC state to the cell-aggregation phase is accompanied by re-activation from a silent state; for Oct4 this transition does not involve activation of expression, which is similar in EpiLCs and in early EpiLC-derived cell aggregates, but a change in its expression control from the PE to DE. Moreover, only in Tx-treated NanogΔObs3;Otx2ERT2 cells is activation of Nanog accompanied by induction of Sox2 in presumptive PGLCs, suggesting that NANOG is required to induce Sox2 and that both of them activate DE-driven expression of Oct4 when OTX2 is suppressed. Notably, NANOG and SOX2 have been reported to bind to the DE of Oct4 (Wu and Schöler, 2014).
Importantly, this study shows that the repressive control exerted by OTX2 on Oct4 or Nanog also occurs in ΔObs homozygous embryos, which exhibit increased numbers of PGCs. Notably, PGC generation is not detected at ectopic sites in mutant embryos. This suggests that only proximal-posterior epiblast-derived cells, which lie close to the BMP4 source in the extra-embryonic ectoderm, respond to loss of OTX2 binding to the Oct4 or Nanog enhancer region by expanding generation of PGCs. However, our data do not currently distinguish between mechanisms involving a somatic mesoderm to germline fate switch of epiblast-derived cells or an alternative, cell-autonomous preferential proliferation of PGC precursors. In sum, previous work identified Otx2 as a crucial determinant controlling allocation of epiblast cells to germline or somatic lineage (Zhang et al., 2018). Here we show that repression of Oct4 and Nanog is a major mechanism by which OTX2 balances the contribution of epiblast cells to germline and somatic lineage, and, in addition, that OTX2 plays further roles preventing germline entry in the absence of cytokines.
MATERIALS AND METHODS
Generation of ESC and mouse mutants
Mutant ESC lines were generated by homologous recombination into E14Tg2a cells. The NanogΔObs3 ESC lines corresponding to clones 1 and 7, have previously been described (Acampora et al., 2016). In this study, clone 1 was used for all experiments and clone 7 for independent confirmatory assays (Fig. 2). As for the Oct4ΔObs ESC line, the mutant version of Obs1, Obs2 or Obs3 (Fig. S1A) was first introduced in the 5′ flanking sequence of the Oct4 gene by PCR-mediated mutagenesis; this region was then inserted into the targeting vector. Correct targeting into ESCs was assessed by Southern blot with probe a (Fig. S1B,C). Two Oct4ΔObs-neo/+ ESC clones were subjected to alternate rounds of neomycin cassette removal and re-targeting to obtain two independent homozygous ESC lines corresponding to clones 3 and 24, as confirmed by Southern blot and PCR analysis (Fig. S1C,D). Clone 24 represented an independent clone for confirmatory experiments (Fig. 2). The Oct4ΔObs clone 3 was employed to perform all the experiments reported in this study. The Oct4ΔObs/+ ESC clone used to generate the Oct4ΔObs clone 3, was also injected into C57BL/6 blastocysts to establish a mouse colony, which was kept in the B6D2 genetic background. Oct4ΔObs homozygous mice were healthy and fertile. The homozygous Oct4ΔObs1,3 ESC line was also obtained by two alternate rounds of targeting and neomycin cassette removal in the homozygous Oct4ΔObs cell line, by using a targeting vector carrying the mutant version of Obs1 and Obs3 sequences, but with the Obs2 sequence reverted to that of the wild-type gene (Fig. S1E). Genotyping of the homozygous Oct4ΔObs1,3 cells was confirmed by PCR with allele-specific primers (Fig. S1F) and Southern blot (not shown). Primers for genotyping are listed in Table S5. The pPyCAGOtx2-ERT2 plasmid (Zhang et al., 2018) (Fig. S1G) was electroporated into E14 NanogΔObs3, Oct4ΔObs and Otx2KO ESCs to generate cell lines showing cytoplasmic and ubiquitous distribution of the OTX2-ERT2 fusion protein, the nuclear translocation of which was dependent on 4-hydroxytamoxifen (Tx) (Sigma) administration. Twelve randomly integrated clones for each mutant cell line were assayed by western blot to determine transgene expression. ESC lines exhibiting comparable levels of OTX2-ERT2 were identified and those showing a level approximately twice that exhibited by the endogenous OTX2 were selected (Fig. S1H). These clones were then assayed by immunostaining with an antibody directed against ER. In the absence of Tx, the OTX2-ERT2 protein was prevalently localized to the cytoplasm, whereas 1 h of Tx (200 nM) exposure was sufficient to confine OTX2-ERT2 to the nucleus (Fig. S1I).
All ESC lines of this study tested negative for mycoplasma contamination. Experiments involving the use of animals were carried out in accordance under authorization 1196/2015-PR from the Italian Ministry of Health.
Cell culture experiments
ESCs were either cultured in Glasgow Minimum Essential Medium (GMEM) (Sigma) supplemented with 12% FBS (HyClone) and 1000 U/ml of leukaemia inhibitory factor (LIF) (Millipore) (referred to as FBS+LIF), or in serum-free N2B27 medium (Gibco) supplemented with 500 U/ml LIF, 1 µM of the mitogen-activated protein kinase kinase 1/2 (MEK1/2) inhibitor PD325901 (PD) and 3 µM of the glycogen synthase kinase 3 (GSK3) inhibitor CHIR99021 (CHIR) small molecules (both from Calbiochem) (referred to as LIF+2i) (Ying et al., 2008). Cell aggregates were generated from EpiLCs induced from ESCs adapted to LIF+2i medium for a minimum of seven passages or maintained in FBS+LIF. ESCs (2×105) were cultured in a six-well plate coated with human fibronectin (Sigma) (16.7 µg/ml) in serum-free N2B27 medium containing 1% KnockOut Serum Replacement (KSR, Gibco ThermoFisher), 12 ng/ml of basic fibroblast growth factor (FGF2) (PeproTech) and 20 ng/ml activin A (R&D Systems) for 48 h (Buecker et al., 2014). EpiLCs were then dissociated with TrypLE reagents (Gibco ThermoFisher) and resuspended at a concentration of 8×104 cells/ml in GMEM medium containing 15% KSR (GK15) or together with various combinations of cytokines and small molecules. Next, 25 µl of cell suspension (about 2000 cells) was plated on the lids of tissue culture dishes and incubated, as hanging drops, over PBS-containing bottom plates. After 2 days, cell aggregates were either collected for immunohistochemistry on sections and cytospin experiments, or transferred to ultralow adhesion 24-well plates (Corning) where they were incubated in the same culture media for an additional 4 days before FACS analysis. Medium was replaced every other day. Culture conditions for cell aggregates were based on the method reported by Hayashi et al. (2011).
In protocol 1, cell aggregates were cultured in GK15 supplemented with LIF (1000 U/ml), bone morphogenetic protein 4 (BMP4) (50 ng/ml), BMP8A (50 ng/ml), stem cell factor (SCF) (10 ng/ml) (all from R&D Systems) and epidermal growth factor (EGF) (10 ng/ml) (PeproTech) (Zhang et al., 2018; Hayashi et al., 2011) (Figs 1B and 2A). Protocol 2 differed from protocol 1 in that ESCs were maintained in FBS+LIF (Fig. 2I). In protocol 3, cell aggregates were cultured in only GK15 medium (Fig. 3A). In protocol 4, cell aggregates were cultured in GK15 with LIF (1000 U/ml) only (Fig. 4A). In protocol 5, cell aggregates were cultured in GK15 supplemented with BMP4 (50 ng/ml), BMP8A (50 ng/ml), SCF (10 ng/ml) and EGF (10 ng/ml), and without LIF (Fig. 4F). In protocol 6, cell aggregates were cultured in GK15 supplemented with CHIR (3 µM) only (Fig. 5A). In protocol 7, cell aggregates were cultured in GK15 supplemented with LIF (1000 U/ml) and CHIR (3 µM) (Fig. 5F). In protocol 8, cell aggregates were cultured in GK15 supplemented with LIF (1000 U/ml) and low doses of BMP4 (5 ng/ml), BMP8A (5 ng/ml), SCF (1 ng/ml) and EGF (1 ng/ml) (Fig. 6A). In protocol 9, cell aggregates were cultured in GK15 supplemented with LIF (1000 U/ml) and CHIR (3 µM), and low doses of BMP4 (5 ng/ml), BMP8A (5 ng/ml), SCF (1 ng/ml) and EGF (1 ng/ml) (Fig. 6F).
Two additional culture conditions without LIF were employed as control experiments: in the first condition, cell aggregates were cultured in GK15 supplemented with low doses of BMP4 (5 ng/ml), BMP8A (5 ng/ml), SCF (1 ng/ml) and EGF (1 ng/ml) (Fig. S6A); in the second condition, cell aggregates were cultured in GK15 supplemented with CHIR (3 µM) and low doses of BMP4 (5 ng/ml), BMP8A (5 ng/ml), SCF (1 ng/ml) and EGF (1 ng/ml) (Fig. S6D).
For experiments performed to assess the effect of different concentrations of LIF, wild-type and mutant cell aggregates were cultured according to the standard and modified versions of protocols 4 and 9. For protocol 4, cell aggregates were cultured for 6 days in GK15 medium supplemented with 300, 1000 (standard protocol 4) and 3000 U/ml of LIF only (Fig. S7A,B); for protocol 9, cell aggregates were cultured in the presence of fixed concentrations of CHIR (3 µM) and low doses of BMP4 (5 ng/ml), BMP8A (5 ng/ml), SCF (1 ng/ml) and EGF (1 ng/ml) in combination with 300, 1000 (standard protocol 9) and 3000 U/ml of LIF (Fig. S7C,D).
For experiments involving E14;Otx2ERT2, Oct4ΔObs;Otx2ERT2, NanogΔObs3;Otx2ERT2 and Otx2KO;Otx2ERT2 cell lines, Tx (200 nM) was administered from d0 to d2 of the cell aggregation phase both for FACS experiments on d6 and immunohistochemistry and RT-qPCR analyses on d2. For these cell lines, EpiLC-derived cell aggregates were cultured following protocol 1 conditions.
Alkaline phosphatase assay, and FGF2 and LIF response experiments
For alkaline phosphatase (ALP) staining, 8×102 ESCs cultured in FBS+LIF medium were seeded as triplicates in six-well plates and grown for 5 days. Upon fixation in 10% neutral buffered formalin (Sigma), ALP activity was revealed by 30 min incubation in a freshly made solution containing 0.1 M Tris-HCl (pH 8.3), Naphthol AS MX-PO4 (5 mg, pre-dissolved in 200 µl of N,N-Dimethylformamide) as phosphatase substrate and Fast Red Violet (30 mg) as a dye (both from Sigma) in a 50 ml final volume. After a few washes in water, colonies were scored as those exhibiting homogeneous staining and counted. For LIF response experiments, ESCs were kept for up to 1 h in GMEM containing 15% KSR in the absence of LIF, which was then added back to the medium at a concentration of 103 Units/ml. p-STAT3 was monitored after 20 min by western blots (Acampora et al., 2017). For the FGF2 response, ESCs were washed and kept in GMEM containing 5% KSR with or without FGF2 at a concentration of 5 ng/ml for 15 and 45 min (Acampora et al., 2016). The FGF2 response was determined by assessing the level of p-ERK1,2 in western blots.
ChIP experiments were performed as previously described (Lee et al., 2006) on wild-type and Otx2KO ESCs and EpiLCs with a rabbit polyclonal OTX2 antibody (Acampora et al., 2016), and on d2 cell aggregates with a rabbit histone H3 tri-methyl lysine 9 (H3K9me3) antibody and normal rabbit IgG. On the day before chromatin preparation, protein A dynabeads (Invitrogen) were washed three times with PBS containing 0.5% BSA and pre-adsorbed overnight with the different antibodies and normal rabbit IgG. For aggregates only, cells were washed with PBS and trypsinized before fixation. Formaldehyde (1%) crosslinked cells were then quenched with 125 mM glycine and washed twice with ice-cold PBS. Upon collection and centrifugation at 4°C, cells were lysed in LB1 buffer (50 mM Hepes-KOH, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40 and 0.25% Triton X-100), then resuspended in LB2 buffer [10 mM Tris-HCl (pH 8.0), 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA] with rocking at 4°C for 10 min each time. Isolated nuclei were resuspended into LB3 buffer [10 mM Tris-HCl (pH 8), 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate and 0.5% N-lauroylsarcosine Na-salt] and kept on ice. All lysis buffers were prepared with freshly added protease inhibitor cocktail plus PMSF. Sonication was performed with a Misonix 2000 sonicator to obtain fragment size with a peak of ∼500 bp or 300 bp, respectively, for OTX2 or H3K9me3 ChIP assays. SDS 0.1% (only for OTX2 ChIP assays) and Triton X-100 (1%) were added to sonicated chromatin before centrifugation to remove debris. Supernatant was then incubated with the antibody- or IgG-coated beads on a rotator at 4°C for at least 18 h (Acampora et al., 2016). Beads were collected at any step through the magnetic stand and sequentially washed with low-salt solution [0.1% SDS, 1% Triton, 2 mM EDTA (pH 8), 20 mM Tris-HCl (pH 8) and 150 mM NaCl], high-salt solution [0.1% SDS, 1% Triton, 2 mM EDTA, 20 mM Tris-HCl (pH 8) and 500 mM NaCl], LiCl solution [0.25 M LiCl, 1% NP40, 1% Na-Deoxycholate and 10 mM Tris-HCl (pH 8)] and TE buffer [Tris-HCl 10 mM (pH 8) and 1 mM EDTA] containing 50 mM NaCl. Beads were then incubated for 15 min at 65°C with elution buffer [50 mM Tris-HCl (pH 8), 10 mM EDTA and 1% SDS]. The immunoprecipitated (IP) DNA was separated from the beads and reverse crosslinked along with input (non-IP DNA) by overnight incubation at 65°C. After RNA and protein digestion, DNA samples were purified by phenol-chloroform extraction and ethanol precipitation. DNA was resuspended in 10 mM Tris-HCl (pH 8) and used in standard qPCR reactions. Fold enrichment of OTX2 ChIP samples was calculated relative to ChIP in Otx2KO cells, used as a negative control, according to the 2−ΔΔCt formula. ChIP values for H3K9me3 and control IgG were calculated as percentage of input, according to the formula: % Input=2−ΔCt(normalized ChIP). Data are mean±s.d. of three independent experiments, each in technical triplicates. ChIP primers are listed in Table S5.
Immunocytochemistry and immunohistochemistry
For immunocytochemistry experiments on ESCs and EpiLCs, adherent cells grown on four-well chamber slides (Sarstedt) were PBS washed and fixed with 4% paraformaldehyde (PFA). Blocking (1 h) and incubation with primary antibodies (overnight) were performed in PBS containing 0.3% Triton X-100 and 2% skimmed milk powder (BioRad). Cells were washed three times with PBS before and after incubation with Alexa Fluor secondary antibodies (90 min). Cells were then counterstained with DAPI (Acampora et al., 2016). Immunohistochemistry experiments on d2 PGCLCs were performed on an average of 100-150 cell aggregates per experiment. After washing in PBS, cell aggregates were fixed overnight in 4% PFA, then dehydrated and processed for paraffin wax embedding and microtome sectioning. Slides were then xylene deparaffinized and rehydrated through a descending series of alcohol to water before being boiled in citrate buffer (pH 6.0) for antigen retrieval. Sections were then incubated with a blocking solution containing 0.5% milk, 10% FBS and 1% BSA, and hybridized with the primary and Alexa Fluor secondary antibodies as described before (Acampora et al., 2013). For immunostaining of cytospin preparations, d2 cell aggregates were collected, washed twice with PBS, dissociated with trypsin and resuspended in 5% FBS in PBS to obtain a single cell suspension. After centrifugation, cells were resuspended at 1×106 cells/ml in 1% BSA in PBS and 2×105 cells were loaded onto cytofunnels and centrifuged for 7 min at 1200 g Cells were fixed in 4% PFA and processed for immunostaining as described for ESCs and EpiLCs.
For in vivo PGC counting, embryos obtained from natural matings of age-matched (3 months old) randomly selected breeding pairs of wild-type and ΔObs mutants were collected at E7.5 and E8.5, fixed overnight in 4% PFA in PBS, dehydrated, paraffin wax embedded and sectioned. Embryos matched for morphology or number of somites, were processed for immunostaining with fragilis and BLIMP1 antibodies as for cell aggregates. All sections, including PGCs, were selected and captured for cell counting. All immunostaining images were captured with a Nikon eclipse NI microscope. For immunohistochemistry experiments performed with three compatible antibodies, fluorescence was excited at 405 nm for DAPI and at 488 nm, 555 nm and 647 nm for secondary antibodies. Eventually, red colour for fluorescence excited at 555 nm was converted to green pseudocolour to allow merging of all combinations. All antibodies employed for immunostaining experiments are listed in Table S6.
Cell counting and statistical analysis
Cytospin assays were immunostained and three non-overlapping fields per experiment were captured, images were printed in A4 format and cell number manually determined. Standard deviation (s.d.) was calculated from the analysis of three independent experiments for each differentiation protocol and for each cell line (Tables S2 and S3). Similarly, also for PGC counting in embryos, immunostained sections were selected, captured, printed in A4 format and fragilis+-BLIMP1+ cells counted. Embryos were selected by morphology at E7.5, and by morphology and number of somites at E8.5. Standard deviation was determined by the analysis of six and seven embryos/genotype, respectively, at E7.5 and E8.5 (Table S4). P values were determined by using the one-tailed or two-tailed unpaired Student's t-test.
FACS analysis was performed on d6 cell aggregates. An average of 50-60 aggregates were collected and dissociated into single cells with trypsin and neutralized in PBS containing 10% FBS. A maximum of 5×105 cells were collected by centrifugation and the pellet resuspended in 100 μl PBS plus 10% FBS supplemented with Alexa Fluor 647 anti-mouse/human CD15 (SSEA1) (Biolegend) and PE anti-mouse/rat CD61 (Biolegend) (Zhang et al., 2018), diluted 1/200 and 1/500, respectively, and incubated for 30 min at 4°C. Cells were washed twice in 1 ml PBS plus 10% FBS before analysis on a BD FACSAria III (Becton Dickinson). Data were acquired using BD FACSDiva software 8.0.1 and analysed using FlowJo software version 10.7.1. Gating strategy is shown in Fig. S3A. The dissociated cell population was first gated on the basis of the FSC (size) and SSC (complexities) scatter plot; singlets were then selected based on the linear correlation between the FSC area (FSC-A) and the FSC height (FSC-H); dead cells were excluded by 7-AAD dye (Fig. S3A-C). All FACS assays were performed in at least three independent experiments.
c57BL/6 blastocysts were injected with Oct4ΔObs ESCs (n=10 cells/embryo) for a total of 50 embryos and transferred into B6D2F1 foster mothers. Chimaerism was assessed by extent of coat colour mosaicism.
RNA extraction, cDNA synthesis and RT-qPCR experiments
Total RNA was extracted from cell lysates using TRIzol Reagent (Invitrogen) and reverse transcribed using Superscript II (Invitrogen) according to the manufacturer's instructions. First-strand cDNA was used for RT-qPCR analysis using Power SYBR Green Master Mix (Applied Biosystems). RT-qPCR experiments were performed on RNA extracted from about 80 cell aggregates per time point (from EpiLC, until day 2) for protocols 1 and 3-9, using primers specific for Otx2, Oct4, Nanog, Ap2γ, Blimp1, T and Hoxb1. RT-qPCR experiments were also performed on RNA extracted from SSEA1+-CD61+ cells, FACS-sorted with the same gating strategy previously described at day 6 for PGCLC differentiation in protocols 1 and 9. For SSEA1+-CD61+ sorted cells, primers were specific for Blimp1, Prdm14, Ap2γ, Nanos3, Ddx4 (also known as Mvh), Prmt5, fragilis, Alpl (also known as Tnap), Gcna1, Dazl (also known as Dazla) and Kit. Gene expression was normalized to TATA-box binding protein (Tbp) transcripts according to the 2(Ct TBP−Ct target) formula. Experiments were performed as biological triplicates and technical duplicates. Oligonucleotide sequences for RT-qPCR analysis are listed in Table S5.
Western blot assay
Total cell lysates were loaded and run on a 10.5% SDS-PAGE gel. Proteins were transferred onto Protran nitrocellulose membranes (GE Healthcare), blocked in 5% skimmed milk (BioRad) for 1 h at room temperature and incubated with primary antibodies (4°C, overnight). PBS plus Tween 20 washes were carried out before and after secondary antibody incubation (1 h at room temperature). Protein expression was revealed by ECL reactions (GE Healthcare). Western Blot antibodies are listed in Table S6.
We thank members of the I.C. lab and V. Wilson for comments on the manuscript, L. Pisapia for support with FACS data analysis, and the staff of the IGB animal house and IGB FACS facility. We also thank D. Graniero for typing and formatting the manuscript.
Conceptualization: L.G.D.G., D.A.; Methodology: L.G.D.G., D.A., D.O.; Formal analysis: V.N.; Investigation: L.G.D.G., D.A., D.O., P.B., E.B.; Writing - original draft: D.A., A.S.; Writing - review & editing: I.C., A.S.; Supervision: A.S.; Funding acquisition: A.S.
This work was supported by the Fondo Europeo di Sviluppo Regionale (SATIN-POR CAMPANIA FESR 2014/2020) and the PRIN project from the Ministero dell'Istruzione, dell'Universita' e della Ricerca (20157JF8P5_004 to A.S.), and by the Medical Research Council (MR/T003162/1 to I.C.). E.B. is supported by a Marie Skłodowska-Curie fellowship (H2020-MSCA-IF-2018/843879).
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199166
The authors declare no competing or financial interests.