SETDB1 is essential for mouse primordial germ cell fate determination by ensuring BMP signaling Free

In most multicellular organisms, germ cells are solely responsible for ensuring successive generations by giving rise to spermatozoa and oocytes, the two sources of developmental totipotency. The union of a sperm and an oocyte via fertilization leads to the generation of a new organism. In mouse embryos, primordial germ cells (PGCs), which are the progenitor cells for spermatogenesis and oogenesis, are fate-determined from undifferentiated epiblast cells at around embryonic day (E) 6.5. Previous gene knockout (KO) studies have shown that bone morphogenetic protein 4 (BMP4) and BMP8B, which signal from the extra-embryonic ectoderm, and WNT3, which signals in epiblasts, are crucial for the fate determination of PGCs from epiblasts (Fig. 1) (Lawson et al., 1999; Ying et al., 2000; Ohinata et al., 2009). BMP4/8B and WNT3 control downstream intracellular signaling pathways during the formation of PGCs, and embryos lacking BMP4/8B or WNT3 fail to form PGCs.

A BMP ligand initiates intracellular signaling by binding to and bringing together type I (BMPR1A, BMPR1B, ACVR1 and ACVRL1) and type II (BMPR2, ACVR2A and ACVR2B) receptor serine/threonine kinases on the cell surface (Yadin et al., 2016). This allows the type II receptor to phosphorylate the type I receptor kinase domain, which then propagates the signal through phosphorylation of the conserved C-terminal residues of SMAD1, SMAD5 or SMAD8 (also known as SMAD9), each of which form heterodimers with SMAD4 and translocate to the nucleus and function as transcriptional regulators (Chen et al., 2004).

During PGC fate determination, BMP4 transmits signals to proximal epiblast cells through SMAD1/5 and SMAD4 at E5.5, which leads to upregulation of Blimp1 (Prdm1) and Prdm14, two of the PR/SET histone methyltransferase domain-containing transcriptional regulators essential for PGC determination, at E6.25 and E6.75, respectively (Fig. 1) (Ohinata et al., 2009). BLIMP1 and PRDM14 further upregulate Tfap2c, another transcriptional regulator crucial for PGC establishment at E7.0. In Prdm14 KO PGCs, levels of histone H3 lysine 9 dimethylation (H3K9me2) and of H3 lysine 27 trimethylation (H3K27me3) are abnormally increased and decreased, respectively, suggesting that PRDM14 is involved in epigenetic regulation in PGCs (Yamaji et al., 2008), but histone methyltransferase enzymatic activity of PRDM14 has not been demonstrated. On the other hand, Blimp1 KO ‘failed’ PGC-like cells have an aberrant transcriptome profile with de-repression of somatic genes, suggesting that BLIMP1 has a role in the repression of somatic genes during PGC fate determination (Ohinata et al., 2005; Kurimoto et al., 2008), but the detailed molecular mechanisms of BLIMP1 functions remain unknown.

As mentioned above, signaling molecules and transcriptional regulators involved in PGC formation have been identified, but the epigenetic regulation of PGC fate determination remains poorly understood. To approach this problem, we recently performed RNA interference (RNAi) screening to identify histone modifier genes involved in PGC fate determination using an in vitro model of PGC-like cell (PGCLC) induction (Hayashi et al., 2011) and identified a histone deacetylase, HDAC3. We demonstrated that selective recruitment of HDAC3 to somatic genes by BLIMP1 and subsequent repression of somatic gene expression are crucial for PGC fate determination (Fig. 1) (Mochizuki et al., 2018).

In this study, we focused on another candidate, SETDB1, which catalyzes the tri-methylation of lysine 9 (K9) of histone H3 and was identified by RNAi screening. Methylated K9 of histone H3 is a repressive epigenetic marker, and the expression of genes with this modification is often downregulated. SETDB1 is a member of the SET domain-containing lysine-specific histone methyltransferases (HMTs), which play crucial roles in controlling chromatin structures and transcription in germline cells. For example, the HMTs EHMT2 and SUV39H1 are involved in genome stability during germ cell development and the completion of meiosis (Tachibana et al., 2007; Peters et al., 2001). In addition, SETDB1 is required for silencing endogenous retroviruses in gonadal PGCs (Liu et al., 2014). However, there is little information regarding the functions of HMTs in nascent PGCs. We show here that Dppa2, Otx2 and Utf1 are repressed by SETDB1 in developing PGCs, and that these gene products repress the BMP/SMAD pathway genes Acvrl1 and Smad1. Our results suggest that proper repression of Dppa2, Otx2 and Utf1 by SETDB1 is crucially important for PGC fate determination of epiblast cells.

We have recently established an efficient RNAi screening protocol to identify genes involved in PGC fate determination using PGCLC induction of mouse embryonic stem cells (ESCs) in culture (Mochizuki et al., 2018), and identified SETDB1 histone H3K9 tri-methyltransferase as a promising candidate. Setdb1 is constitutively expressed in epiblasts, PGCs, epiblast-like cells (EpiLCs) and PGCLCs (Fig. 2C; Kurimoto et al., 2008; Kurimoto et al., 2015). Knockdown (KD) of Setdb1 by two different siRNAs resulted in reduction of Blimp1::Venus (BV) reporter expression both at day 2 and at day 4 after induction (Fig. 2A,B,D). We also evaluated the endogenous expression of three important PGC determinant genes (Blimp1, Prdm14 and Tfap2c), all of which were significantly downregulated (Fig. 2E). We also examined gene expression in fluorescence activated cell sorting (FACS)-purified BV-positive (+) and -negative (−) cells, and found that the expression of Blimp1 was downregulated in BV(+) and BV(−) cells (Fig. S1A). Although Prdm14 and Tfap2c were preferentially downregulated in BV(−) cells, their expression in BV(−) cells was low compared with that in BV(+) cells (Fig. S1B), suggesting that influence of Setdb1 KD on the expression of those two genes for PGC specification in this experimental condition is minimal. These results together suggest that SETDB1 positively regulates PGC fate determination. In addition, the expression of early somatic and mesodermal cell-related genes were downregulated in BV(+) and BV(−) cells (Fig. S1), suggesting that mesodermal differentiation is also affected by Setdb1 KD.

Because Setdb1 ablation leads to embryonic lethality shortly after implantation at around E3.5-E5.5 (Dodge et al., 2004), we examined Setdb1 conditional knockout (cKO) embryos obtained from the mating of Setdb1-flox and epiblast-specific Sox2-Cre lines to evaluate the roles of SETDB1 on PGC fate determination in vivo. At the mid-bud stage (around E7.5), a significantly smaller number of BLIMP1-positive nascent PGCs were observed in Setdb1 cKO embryos than in wild-type (WT) embryos (Fig. 3A,B). In addition, the number of alkaline phosphatase-positive PGCs decreased in the cKO embryos at around E8.0 (Fig. S2). We further found that T, encoded by another crucial PGC determinant gene (Aramaki et al., 2013), was coincidently expressed in the BLIMP1-positive cells not only in WT but also in Setdb1 cKO embryos (Fig. 3A), though its expression was apparently decreased in the cKO embryos (Fig. 3C). The localization of phosphorylated (p)-SMAD1/5/8 in the posterior-proximal epiblast of Setdb1 cKO embryos was also significantly reduced, suggesting abnormality in the BMP/SMAD pathway (Fig. 3A,D). Reduction of p-SMAD1/5/8 was also observed in the adjacent mesodermal cells, which is consistent with our transcriptome and RT-qPCR data showing downregulation of mesoderm-related genes, including BMP signal-related genes, by Setdb1-KD in cell aggregates containing PGCLCs and early mesodermal cells as described below (Fig. 4, Fig. S1). The data suggest that SETDB1 is also crucial for development of early mesodermal cells.

To investigate gene regulation by SETDB1 during PGCLC induction, we carried out RNA sequencing (RNA-seq) analysis to compare the transcriptome in Setdb1 KD cells with that in control KD cells 2 days after the induction of PGCLCs. RNA-seq analysis showed that 870 and 1355 genes were significantly upregulated or downregulated by Setdb1 KD, respectively (q<0.05, three biological replicates) (Fig. 4A, Table S1). Gene ontology (GO) analysis revealed that mesoderm development-related GO terms were most significantly enriched among the downregulated genes (Table S2) whereas ‘metabolic process’ is the only enriched GO term among the upregulated genes, suggesting that early mesoderm development is impaired by Setdb1-KD.

We then speculated that repressive transcriptional regulators for the downregulated mesoderm-related genes may exist among the genes upregulated by Setdb1 KD (Fig. S3). To assess this possibility, we first identified 513 genes encoding putative transcription factors enriched in the vicinity (TSS±5 kbp) of the downregulated genes (Table S1) by in silico ChIP analysis (ChIP-Atlas - Enrichment Analysis; http://chip-atlas.org/enrichment_analysis) of public ChIP-seq datasets (more than 6000 experiments) in various mouse cell types (Table S3). We next extracted genes common to the genes upregulated by Setdb1 KD and the selected putative transcription factor genes for the downregulated genes, and identified 28 candidate genes (Fig. 4B, Fig. S3). Of these 28 candidates, we found that only the expression levels of Dppa2, Olig2, Otx2 and Utf1 were drastically (|log2 level|>2.0) decreased in day 2 PGCLCs compared with EpiLCs, corresponding to transient early mesodermal differentiation (Fig. 4C, Fig. S4A). Re-analysis of transcriptome data from previous studies [Nakamura et al., 2011 (GSE31581); Buecker et al., 2014 (GSE56096); Jia et al., 2012 (GSE39513)] showed that upregulated genes in Dppa2- or Utf1-defcient ESCs and Otx2-deficient EpiLCs (i.e. target genes of DPPA2, UTF1 or OTX2) among genes downregulated by Setdb1 KD in PGCLC-containing cell aggregates included mesoderm-related genes (Fig. S4C, Table S4), suggesting that SETDB1 indirectly activates and DPPA2, OTX2 or UTF1 repress common mesoderm-related genes in PGCLCs. Therefore, we focused on Dppa2, Otx2 and Utf1 as promising candidate SETDB1-targeted genes in PGC specification. We confirmed the binding of SETDB1 protein at the flanking regions of Dppa2, Otx2 and Utf1 in cell aggregates containing PGCLCs (Fig. 4E) and the upregulation of Dppa2, Otx2 and Utf1 by Setdb1 KD (Fig. 4D). We also found decreased levels of H3K9 trimethylation (me3) at the flanking regions of Dppa2, Otx2 and Utf1 by Setdb1 KD in cell aggregates containing PGCLCs (Fig. 4F). These results suggest that SETDB1 directly represses Dppa2, Otx2 and Utf1 via H3K9me3 (Fig. 4G). In addition, the expression of UTF1 was upregulated in BLIMP1-positve PGCs in Setdb1 cKO embryos (Fig. 5, Fig. S5), suggesting that Utf1 is under the control of SETDB1 in PGCs in vivo.

To assess how the de-repression of Dppa2, Otx2 and Utf1 affects PGC determination, we next explored target genes of these transcription factors among the mesoderm development-related genes downregulated by Setdb1 KD. Specifically, we focused on the BMP4 signaling-related genes Acvrl1, Bmpr2, Acvr2b and Smad1 (Table S2), which should be involved in PGC fate determination. We first confirmed their downregulation in PGCLC-containing cell aggregates by Setdb1 KD (Fig. 4D). In addition, Smad1 tended to be downregulated in the mesodermal region, which contains PGCs, in Setdb1 cKO embryos (Fig. S5). We then tested whether Dppa2, Otx2 and Utf1 directly repress the expression of the BMP4 signaling-related genes. We examined the enrichment of DPPA2, OTX2 and UTF1 at the flanking regions of two BMP4 signaling-related genes (Acvrl1 and Smad1) in EpiLCs transfected with expression vectors encoding FLAG-DPPA2, HA-OTX2 and MYC-UTF1. As expected, FLAG-DPPA2, HA-OTX2 and MYC-UTF1 were enriched at these genes (Fig. 6A-C). We then evaluated the effects of overexpression (OE) of Dppa2, Otx2 and Utf1 on the expression of the PGC determinant genes. We transduced EpiLCs with a Dppa2-Otx2-Utf1 lentiviral vector and cultured the cells under PGCLC induction conditions, then examined expression of the BMP4 signaling-related genes and the PGC determinant genes at day 2 and/or 4 after transduction. We confirmed overexpression of Dppa2, Otx2 and Utf1 (Fig. 6D), and found that the expression of two BMP4 signaling-related genes (Acvrl1 and Smad1) and three PGC determinant genes (Blimp1, Prdm14 and Tfap2c) were significantly downregulated, or showed a tendency to be downregulated (although not statistically significant), suggesting repression of BMP signaling and of PGCLC formation by the ectopic expression of Dppa2, Otx2 and Utf1 (Fig. 6E-H). Finally, we confirmed that the levels of p-SMAD1/5/8 in the posterior-proximal region of epiblasts (where forming PGCs locate) were lower in Setdb1 cKO embryos compared with stage-matched WT embryos (Fig. 3A,D). Taken together, these results suggest that the repression of Dppa2, Otx2 and Utf1 by SETDB1 is functionally important for PGC fate determination by ensuring BMP signaling.

In this research, we aimed to elucidate the molecular mechanism underlying PGC fate determination by the histone methylation enzyme SETDB1 and demonstrated that SETDB1 is responsible for the H3K9 methylation-mediated repression of Dppa2, Otx2 and Utf1. These gene products suppress the expression of BMP4 signaling-related genes, including Smad1; consequently, SETDB1 ensures BMP4 signaling required for PGC formation (Fig. 1).

A previous study reported that SETDB1 was required for peri-implantation development. In contrast, the present study showed that SETDB1 plays a role in mesodermal differentiation because mesoderm-related genes are enriched among the genes downregulated by Setdb1 KD (Table S2). We also found that some genes downregulated by Setdb1 KD were upregulated by BMP4 in PGCLCs, some of which are mesoderm related (Fig. S4B, Table S5), further suggesting that SETDB1 controls downstream genes including mesoderm-related genes via BMP signaling. BMP4 phosphorylates SMAD1/5, which is required for both mesodermal differentiation and PGC formation, and further interaction of p-SMAD1/5 with SMAD4 specifically plays a role in PGC formation (Chu et al., 2004).

In addition, genes upregulated by the KO of Dppa2, Otx2 or Utf1 in pluripotent cells overlap with genes downregulated by Setdb1 KD in PGCLCs, and some of these genes are mesoderm related (Fig. S4C, Table S4). For example, mesoderm-related Tbx3, Hoxb1 and Hoxa1 are upregulated in Dppa2 KO ESCs (Nakamura et al., 2011), in Otx2 KO EpiLCs (Buecker et al., 2014) and in Utf1 KO ESCs (Jia et al., 2012), respectively (Table S4). These findings further support the idea that SETDB1 activates mesoderm-related gene expression by suppressing Dppa2, Otx2 and Utf1. The BMP4 signaling-related genes, which are targets of DPPA2, OTX2 and UTF1 in PGCLCs (Fig. 6A-C), are not listed among the genes significantly upregulated by Dppa2, Otx2 or Utf1 KO in pluripotent cells (Table S4), perhaps owing to different cellular contexts between PGCLCs and pluripotent cells, such as the involvement of additional redundant transcription factors for the repression of BMP4 signaling-related genes in pluripotent cells.

SETDB1 may also be involved in WNT3A signaling which is crucial for PGC specification. We found that Wnt3a, but not the closely related gene Wnt3, was downregulated by Setdb1 KD (Fig. 4A, Table S2). Wnt3a was upregulated by KO of Dppa2 in ESCs (Table S4; Nakamura et al., 2011; GSE31581), which is repressed by SETDB1 (Fig. 4A), suggesting that SETDB1 stimulates Wnt3a expression by repressing Dppa2 expression. Our present results suggest that OXT2 represses BMP signaling genes (Fig. 6B,E), and that OXT2 may also have additional repressive functions on WNT3A signaling during epiblast differentiation to early mesoderm/PGCs. Previous studies suggest that OTX2 is required for direct activation of Dkk1, an antagonist of WNT signaling, to suppress the activity of WNT3A signaling (Perea-Gomez et al., 2001; Ip, et al., 2014). Therefore, SETDB1 likely promotes WNT3A signaling through the suppression of Dppa2 as well as Otx2 for mesodermal differentiation as well as for PGC formation.

In this research, we showed that SETDB1 was recruited to Dppa2, Otx2 and Utf1 and repressed their expression through H3K9 methylation during PGC determination (Fig. 1). SETDB1 localizes to retroviral sequences and represses their transcription via H3K9me3 in male/female gonadal PGCs (Liu et al., 2014), and is also required for the H3K9me3-mediated transcriptional repression of cell cycle genes as well as retroviral sequences in female meiotic germ cells (Eymery et al., 2016). Therefore, SETDB1 represses different sets of genes in the processes of PGC specification and of germ cell differentiation, and thus the mechanisms underlying the selective recruitment of SETDB1 to those different genes are attractive subjects for future studies.

Murine EpiLC-PGCLC induction was performed essentially as described previously (Hayashi et al., 2011). Briefly, Blimp1::Venus/Stella::CFP (BVSC) ESCs were adapted to 2i (PD0325901, CHIR99021)+LIF feeder-free culture conditions. The EpiLCs were induced by plating ESCs in a well coated with human plasma fibronectin (Millipore) in N2B27 medium containing Activin A (Peprotech), basic fibroblast growth factor (Gibco) and KnockOut Serum Replacement (KSR) (Gibco). The PGCLCs were induced under floating aggregation culture conditions by plating EpiLCs in a well of a low-cell-binding U-bottom 96-well plate (Nunc) in GMEM-based serum-free medium in the presence of the cytokines BMP4 (R&D), LIF (ESGRO; Chemicon), SCF (R&D) and epidermal growth factor (R&D). Human embryonic kidney (HEK) 293T cells were cultured with DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS).

Setdb1 KD during PGCLC induction was achieved using siRNAs that targeted Setdb1 (QIAGEN; siRNA#1-SI02693803, 5′-CCGGATCAGACCCATGAGAAA-3′; siRNA#2-SI02741375, 5′-CCGAGACTTCATAGAGGAATA-3′) transfected into EpiLCs. siRNAs were transfected into cells using Lipofectamine 3000 (Invitrogen) according to the manufacturer's reverse transfection protocol. Briefly, Lipofectamine 3000 (1.2 µl) and siRNA (28.8 pmol) were diluted with 60 µl of OptiMEM and incubated for 10 min. An aliquot of 6000 BVSC EpiLCs in 300 µl of the PGCLC induction medium was added to each Lipofectamine/siRNA mixture and mixed briefly. The EpiLCs/siRNA mixtures were seeded into three wells (technical triplicates) of low-cell-binding U-bottom 96-well plates and the cells were cultured in a conventional 5% CO₂ incubator at 37°C. Two independent pre-designed siRNAs for each gene were purchased from QIAGEN. BVSC EpiLCs were transfected with individual siRNAs in a 96-well format as described above and transfected cells were cultured for up to 4 days with technical triplicates. Blimp1 siRNA (QIAGEN; Blimp1 siRNA-SI01387505) and AllStars negative control siRNA (QIAGEN; Control siRNA-SI03650318) were used as a positive control (for reduction of Blimp1::Venus reporter fluorescence) and negative control (for no change of Blimp1::Venus reporter fluorescence), respectively, in each experiment. The changes in aggregate fluorescence were semi-automatically quantified by the ‘Spot Detector’ application of a Cellomics ArrayScan cell image analyzer (Thermo Fisher) using a 5× objective. Briefly, after capturing the outline of an aggregate by autofocusing with Channel 1 (bright-field), an automatic image of Channel 2 (Venus) was acquired. In Channel 2, spots with an intensity above a set threshold were extracted and marked. The threshold value was set by using the aggregates transfected with Blimp1 positive control siRNA, and proper detection of positive signals with Allstars negative control siRNA was confirmed in each experiment. Quantitative analysis using the analysis mode generated ObjectAreaChannel1 (approximate total area of aggregate) and SpotTotalAreaChannel2 (approximate total area of Blimp1::Venus-positive regions). Finally, the percentage of Blimp1::Venus-positive regions in each aggregate was calculated. Wells with significantly darker or brighter fluorescence than the control wells were again identified by direct visual inspection using a Leica AF6000 system at day 2 and at day 4 after PGCLC induction. Thereafter, KD cells were lysed for total RNA extraction.

PGCLC-containing cell aggregates were dissociated using Trypsin-EDTA (10 min, 37°C) and washed with DMEM supplemented with 10% FBS. After passing through a 40 µm pore nylon mesh (BD Falcon), BV-positive and -negative cells were purified using a Bio-Rad S3e fluorescence-activated cell sorter.

Total RNAs were extracted from cell cultures and purified using either an RNeasy Micro kit (QIAGEN) or an RNeasy Plus Mini kit (QIAGEN). For embryos after immunostaining, pieces including PGCs and a mesodermal region were cut using a tungsten needle under a microscope. Total RNA was extracted by using RNeasy FFPE Kit (QIAGEN) according to the manufacturer's instructions. Total RNAs were reverse transcribed using Superscript III (Invitrogen) and the cDNAs were used for quantitative PCR with Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) for cultured cells or PowerUp SYBR Green Master Mix (Thermo Fisher Scientific) for embryos. PCR signals were detected using CFX Connect (Bio-Rad). The sequences of the PCR primers are listed in Table S6.

RNA-seq libraries of Setdb1 KD and control KD cells during PGCLC induction (three biological replicates of each) were prepared from 100 ng of total RNA using a TruSeq RNA Library Preparation kit (Illumina). Libraries were sequenced on an Illumina HiSeq 2500 instrument. Reads were generated by 100 bp single end sequencing. For gene expression analysis, reads were mapped to the mouse genome (UCSC mm9 genome assembly and NCBI RefSeq database) using TopHat software. Cufflinks software was used to estimate gene expression levels based on reads per kilobase of exon per million mapped reads (RPKM) normalization. Differentially expressed genes (DEGs) were extracted from the Cuffdiff results with statistical significance (q<0.05). GO analyses were performed using the PANTHER (Protein ANalysis THrough Evolutionary Relationships) classification system (http://www.pantherdb.org).

For epiblasts, EpiLCs, PGCs and PGCLCs (Kurimoto et al., 2008; Kurimoto et al., 2015), microarray data (GSE11128; GSE46855) were analyzed with GEO2R software in Gene Expression Omnibus (GEO). Log2 normalized expression values relative to that of Arbp (Rplp0) are shown. For Dppa2 KO ESCs (Nakamura et al., 2011), microarray data (GSE31581) were analyzed with GEO2R software in GEO. Upregulated genes in Dppa2 KO were extracted with statistical significance (P<0.05, log2 fold change>1.5). Upregulated genes in Otx2 KO (q<0.01) and upregulated genes in Utf1 KO (P<0.00001) were already listed in Buecker et al., 2014 and Jia et al., 2012, respectively.

Mice were kept and bred at the Animal Unit of the Institute of Development, Aging and Cancer (Tohoku University, Japan), an environmentally controlled and specific pathogen-free facility. Epiblast-specific Setdb1 KO embryos were generated using the Setdb1^tm1.1Yshk (Tan et al., 2012) and Sox2-Cre (Tg(Sox2-cre)1Amc) (Hayashi et al., 2002) lines. Briefly, Setdb1^flox/flox females were mated with a Setdb1^Δ/+ Sox2^cre/+ male. For timed matings, the day of appearance of the vaginal plug was considered to be E0.5. The females were sacrificed at E6.5-E7.5 and the embryos were collected in DMEM (Gibco) supplemented with 10% FBS.

For immunohistochemical analysis, isolated embryos were fixed in 4% paraformaldehyde in PBS for 5 min (for BLIMP1, T and p-SMAD1/5/8) or 1 h (for BLIMP1 and UTF1) at 4°C, washed three times with PBS-0.1% Triton X-100 (PBT), and incubated in blocking solution (1% bovine serum albumin, 10% fetal bovine serum, in PBT) overnight. Embryos were then incubated with primary antibodies in the blocking solution overnight, washed four times with PBT, incubated with secondary antibodies and 4,6-diamidino-2-phenylindole (DAPI) overnight in the blocking solution, then washed four times with PBT. The primary antibodies used were: anti-p-SMAD1/5/8 (rabbit monoclonal; Cell Signaling Technology 13820), anti-BLIMP1 (rat monoclonal; Santa Cruz sc-47732), anti-T/BRACHYURY (goat polyclonal; R&D AF2085) and anti-UTF1 (rabbit polyclonal; Abcam ab24273). The secondary antibodies used were: Alexa Fluor 488-conjugated anti-rabbit IgG (Invitrogen, A-21206 or A-11008) and Alexa Fluor 568-conjugated anti-rat IgG (Abcam, ab175475) or Alexa Fluor 647-conjugated anti-goat IgG (Invitrogen, A-21469). After washing with PBT, the ectoplacental cone was used to determine the genotype by PCR. The primer sequences were: Setdb1 Forward, 5′-CAGCTTGGAGGAATTGGTTC-3′; Setdb1 Reverse, 5′-TCCCAAACCTCATAGGGTAAAA-3′; Sox2Cre Forward, 5′-AACATTCTCCCACCGTCAGT-3′; Sox2Cre Reverse, 5′-CATTTGGGCCAGCTAAACAT-3′. Embryos were finally mounted in Vectashield Mounting Medium (Vector Laboratories) and immunofluorescence images were taken using an SP8 confocal microscope (Leica).

To count the number of PGCs in the embryos, the entire BLIMP1-positive region was scanned using the SP8 confocal microscope and the number of BLIMP1-positive cells located in epiblasts (not in the surrounding visceral endoderm) was manually counted in each focal plane at 5 µm intervals. The developmental stages of the tested embryos were determined by morphological criteria (Downs and Davies, 1993).

The LAS X program (Leica) was used to quantify the fluorescence signals of T, pSMAD1/5/8 and UTF1 in PGCs. The average pixel value in the nucleus of each cell was estimated and then plotted. The fluorescence signal intensities of ten randomly selected extra-embryonic ectoderm cells (T and pSMAD1/5/8) and ten randomly selected visceral endoderm cells (UTF1) in which Setdb1 was not deleted by epiblast-specific Sox2-Cre, together with BLIMP1-positive cells in each section, of WT/Hetero and Setdb1 cKO were measured using a confocal microscope. Confocal sections from five (T and pSMAD1/5/8) or three (UTF1) embryos of each genotype were observed. The average value of the extra-embryonic ectoderm or visceral endoderm cells in each section was set as 1, and the fluorescence values of the BLIMP1-positive cells were normalized to the values of the extra-embryonic ectoderm cells or visceral endoderm cells in the same section.

A plasmid encoding FLAG-DPPA2-P2A-HA-OTX2-P2A-MYC-UTF1 expressed under the control of an EF1a promoter was generated by inserting the open reading frames of Dppa2, Otx2 and Utf1 amplified using gene-specific primers with FLAG-, HA- or MYC-tag sequences at their 5′ ends into the multiple cloning site of pCS2-EF-MCS (Miyoshi et al., 1998) (RIKEN BRC; http://cfm.brc.riken.jp/lentiviral-vectors/plasmid-list/). The sequences of the PCR primers are listed in Table S6.

The FLAG-DPPA2-P2A-HA-OTX2-P2A-MYC-UTF1 expression vector was transfected into EpiLCs using Viafect (Promega) according to the manufacturer's instructions. ChIP assays were performed as previously described with minor modifications (Mochizuki et al., 2012). Briefly, about 5,000,000 EpiLCs cells were cross-linked by directly adding 16% formaldehyde (Polysciences) to the cell suspension to a final concentration of 1%, followed by incubation at room temperature with gentle inverting for 10 min. Cell aggregates containing PGCLCs at day 2 after induction, comprising about 5,000,000 cells, were dissociated and treated with protein-protein cross-linker EGS [ethylene glycol bis(succinimidyl succinate) solution (Thermo Fisher)] at a final concentration of 1.5 mM at room temperature with gentle inverting for 20 min, followed by cross-linking with formaldehyde as described above. The chromatin was fragmented by sonication (Diagenode Bioruptor; 30 s ON and 30 s OFF; total processing time of 12 min; output level set at medium). Antibodies used in this assay were obtained from commercial sources with the following catalog numbers: anti-SETDB1, ProteinTech 11231-1-AP; anti-H3K9me3, Abcam ab8898; anti-FLAG, Sigma F1804; anti-HA, Abcam ab9110; and anti-MYC, Abcam ab9132. For each immunoprecipitation reaction, 10-25 µl of Magna ChIP Beads Protein A (Millipore) or Dynabeads Protein G and/or A (Invitrogen) were incubated with 5 µg of the indicated antibody in 500 µl RIPA-150 mM NaCl for between 4 h and overnight at 4°C with rotation, then washed twice with 500 µl of ice-cold RIPA-150 mM NaCl. An aliquot of the fragmented chromatin (500 µl) was incubated with antibody-bound Dynabeads overnight at 4°C with rotation. The enrichment of specific regions in each immunoprecipitated DNA sample was analyzed by quantitative PCR using Power SYBR Green PCR Master Mix. PCR signals were detected using CFX Connect. The sequences of the PCR primers are listed in Table S6. The data were obtained from two independent experiments.

The pCS2-EF-FLAG-DPPA2-P2A-HA-OTX2-P2A-MYC-UTF1 vector and the pCS2-EF-mCherry reporter vector were prepared for Dppa2/Otx2/Utf1-OE and as a negative control, respectively. Lentivirus particles were produced by transfection into HEK293T cells with pCMV-VSV-G-RSV-Rev and pCAG-HIVgp (Miyoshi et al., 1998) (RIKEN BRC; http://cfm.brc.riken.jp/lentiviral-vectors/plasmid-list/). The Dppa2-Otx2-Utf1 lentiviral vector or the mCherry lentiviral vector was transduced into EpiLCs at multiplicities of infection of 10. Dppa2-Otx2-Utf1 OE and control mCherry OE EpiLCs were cultured for 2 and 4 days under PGCLC induction conditions. Thereafter, OE cells were lysed for total RNA extraction followed by RT-qPCR.

Data were analyzed using the Student's t-test. P-values less than 0.05 were considered statistically significant.

All animal experiments were performed under the ethical guidelines of Tohoku University, and animal protocols were reviewed and approved by the Tohoku University Animal Studies Committee. All lentiviral experiments were performed under the ethical guidelines of Tohoku University, and lentiviral protocols were reviewed and approved by the Tohoku University Center for Gene Research.

We thank Dr Yoichi Shinkai for Setdb1-flox mice, Dr Katsuyoshi Takaoka for Sox2-cre mice, Dr Mitinori Saitou for Blimp1::Venus/Stella::CFP (BVSC) ESCs and Dr Katsuhiko Hayashi for technical advice for PGCLC induction. We also thank Drs Jafar Sharif, Tatsuya Takemoto, Shinya Oki, Rieko Ajima, Aiko Kawasumi, Masaki Hosogane, Eri Hosogane-Kobayashi, Satomi Tanaka, Daiji Okamura, Ari Itoh-Nakadai, Yasuhiro Suzuki, Kenji Iemura, Hiroshi Kitamura and all the members of Cell Resource Center for Biomedical Research for helpful discussions, and the Biomedical Research Core of Tohoku University Graduate School of Medicine and the Center of Research Instruments of Institute of Development, Aging and Cancer (IDAC), Tohoku University for use of instruments and technical support.