The developmental dismantling of pluripotency is reversed by ectopic Oct4 expression

During early embryonic development, cells have the capacity to form derivatives of all three embryonic germ layers, a property known as pluripotency. In the environment of the early embryo, pluripotent founder cells respond to patterning signals to differentiate appropriately. However, following grafting to permissive sites, such as the adult kidney capsule in the mouse, all known pluripotent cell types seed teratomas – tumours containing a chaotic mixture of differentiated derivatives of the three embryonic germ layers. A major subset of these, termed teratocarcinomas, also contain self-renewing embryonal carcinoma (EC) cells and are consequently malignant (Andrews, 2002; Solter, 2006). In newborn humans, the rare spontaneous incidence of teratomas and teratocarcinomas along the midline from the forebrain to the base of the spine suggests that developmental suppression of pluripotency normally prevents occurrence of these tumours (Oosterhuis and Looijenga, 2005). Previous analyses have shown that teratocarcinoma-forming potential is lost in mouse during the broad developmental period from ∼E7.5 to ∼E8.5 (Damjanov et al., 1971).

Pluripotency during pre-implantation development and in embryonic stem (ES) cells is governed by a gene regulatory network centred on the core transcription factors Oct4, Sox2 and Nanog (Avilion et al., 2003; Chambers et al., 2007; Masui et al., 2007; Mitsui et al., 2003; Nichols et al., 1998; Niwa et al., 2000; Silva et al., 2009) (reviewed by Chambers and Tomlinson, 2009). Oct4, Sox2 and Nanog are also expressed in the pluripotent post-implantation epiblast and in derivative epiblast stem cell (EpiSC) lines (Brons et al., 2007; Tesar et al., 2007). Classical teratogenesis experiments showed that pluripotency persists in the epiblast at E7.5 but is lost a full day later (Beddington, 1983; Damjanov et al., 1971). However, the role of Oct4, Sox2 and Nanog in the pluripotency of post-implantation embryos or EpiSCs has never been defined.

Here, we determine the precise time at which somatic cells cease to be pluripotent during embryonic development. We correlate this loss of pluripotency with changes occurring in the expression of Oct4 and Nanog. Closure of the regulatory elements of the Nanog and Oct4 genes contributes to this loss of pluripotency. An initial period during which pluripotency can be rapidly reinstated by enforcing Oct4 expression is followed by stable epigenetic repression of pluripotency factor genes.

Mice were maintained on a 12-hour light/12-hour dark cycle. For timed matings, noon on the day of the vaginal plug was designated E0.5. Except where stated, wild-type strains were MF1 or 129/Ola. Gastrulation-stage embryos were staged as described (Downs and Davies, 1993): ES, early streak; MS, mid-streak; LS, late streak; OB, no allantoic bud; EB, early allantoic bud; LB, late allantoic bud; EHF, early headfold; LHF, late headfold.

Embryos were cultured in static 4-well dishes in an incubator at 5% CO₂ in air in 50% rat serum as described (Copp, 1990). Blastocyst injection and embryo transfer was performed using standard procedures.

Ubiquitously induced Oct4 expression was achieved by crossing wild-type females with homozygous males carrying both the doxycycline-inducible reverse tet transactivator (rtTA) targeted at Rosa26 and a doxycycline-responsive Oct4 transgene (TgOct4) targeted at Col1a1 [B6;129-Gt(ROSA)26Sor^{tm1(rtTA*M2)Jae} Col1a1^{tm2(tetO-Pou5f1)Jae/J} (Hochedlinger et al., 2005)].

Wild-type (C57BL/6, 129 and CBA) embryo subregions were dissected in M2 (Sigma) as described (Cambray and Wilson, 2007). To dissect distal/anterior (Nanog:GFP negative) and proximal/posterior (Nanog:GFP positive) regions of the E7.5 embryo, embryos were first ordered relative to developmental stage and imaged to assess GFP-positive and -negative regions. A transverse cut was made to separate the most proximal epiblast, containing primordial germ cells, together with the extra-embryonic region, from the distal epiblast. A second cut was made to separate the distal/anterior from posterior/proximal regions. Finally, the embryos were viewed with fluorescence optics to check the accuracy of microdissection; any remaining Nanog-GFP-positive cells were trimmed from distal-anterior regions. To dissect prospective forebrain (region 1) and adjacent region (region 2), extra-embryonic membranes were removed and the embryo laid flat dorsoventrally, with the aid of a transverse cut near the node if necessary (see Fig. 5). Forebrain and adjacent region were then separated with glass needles.

Kidney capsule grafts were performed as described (Tam, 1990), except that transfer to NOD/SCID mice was performed in a laminar flow cabinet. After transplanting E8.5 TgOct4/+;rtTA/+ forebrains, half of the host animals were administered 1 mg/ml doxycycline (Sigma) through the drinking water for 4 weeks. Tumours were recovered at 4-6 weeks and fixed in 4% paraformaldehyde (PFA) for 1-7 days depending on tumour size. Tissue was processed and stained as described (Bancroft and Gamble). Tissues representative of different germ layers used for scoring all tumours are shown in supplementary material Fig. S2.

To derive E7.5-E8.0 (OB-2s) EpiSC lines, embryos were dissected, dissociated (without removal of endoderm or mesoderm) and explanted into EpiSC medium (Tesar et al., 2007) supplemented with 20 ng/ml activin-A (PeproTech) and 10 ng/ml bFGF (R & D) and plated on irradiated mouse embryo fibroblasts (MEFs). After 2-3 days, primary explants were passaged by incubation with 1× accutase (Sigma, Catalogue number A 6964) (5 minutes) and then triturated into 10- to 100-cell clumps, neutralized with EpiSC medium and replated. Subsequent passages were performed every 5-6 days. Cultures were designated as cell lines when they had reached passage 3, exhibiting EpiSC morphology and robust proliferation.

TgOct4-overexpressing explants and EpiSCs were derived from TgOct4/TgOct4;rtTA/rtTA homozygous male crossed with either 129/Ola wild-type or Nanog:GFP [129-Nanog^{tm1(GFP-IRES-Puro)}] females, derived from TNG targeted ES cells (Chambers et al., 2007). Medium was initially supplemented with 1 mg/ml doxycycline, and reduced to 0.5 mg/ml after three passages. Nanog^–/– EpiSCs were derived in vitro from ES cells as described (Guo et al., 2009) and passaged in accutase as above. Oct4GiP EpiSCs were derived from ES cells carrying an Oct4-promoter driven eGFPiresPuroR-polyA transgene (Ying et al., 2002). When added, chemical inhibitors were used at these concentrations: PD0325901 (Stemgent), 1 μM; SB 431542 (Sigma), 10 μM; JAK inhibitor I (Calbiochem 420099), 0.6 mM.

Tumours were isolated and chopped roughly before dissociation in 0.5% trypsin/2.5% pancreatin in PBS (37°C, 15 minutes). After trituration, cells were plated onto irradiated MEFs in EpiSC media, passaged the next day and then every 2-3 days.

Total RNA (30 ng-1 μg) isolated using the RNeasy microkit or minikit (Qiagen), was used for cDNA synthesis using SuperScript III (Invitrogen). Quantitative PCR was performed using Light Cycler 480 SYBR Green I Master Mix (Roche). PCR primer sequences are available on request. Values for each gene were normalized to expression of TATA-box binding protein (TBP) and expressed as mean±s.e.m. of at least three replicates, relative to ES cell levels.

Whole embryos or cultured cells were fixed using 4% PFA in PBS for 2 hours (embryos) or 10 minutes (cells) at 4°C followed by permeabilization in 0.5% Triton X100 (Sigma) in PBS for 15 minutes. Samples were incubated in 1 M glycine in PBS/0.1% TritonX100 (PBST) for 20 minutes and blocked overnight at 4°C using 3% serum (Sigma), 1% BSA (Sigma) in PBST. Primary antibodies were diluted in blocking buffer to the working concentrations indicated below and applied for 1-2 hours at room temperature (cells) or 48 hours at 4°C (embryos). After three (cells) or four to six (embryos) PBST washes (15 minutes at room temperature), Alexa-conjugated secondary antibodies (Molecular Probes; 2 μg/ml in blocking buffer) were applied for 1 hour (cells) or 3 hours (embryos) in the dark at room temperature. Cells/embryos were then washed at least three times in PBST followed by incubation in 4′,6-diamidino-2-phenylindole (DAPI) to stain nuclei. Primary antibody concentrations were: Nanog, 1.4 μg/ml (Chambers, 2004) (embryos) or 2.5 μg/ml (ab14959, Abcam) (cells); Oct4, 1 μg/ml (Santa Cruz, N-19) (embryos) or 1 μg/ml (Santa Cruz, C-10) (cells); Sox2, 1 μg/ml (Santa Cruz, Y-17) (cells).

Whole-mount in situ was performed as described (Wilkinson et al., 1990); proteinase K treatment was empirically adjusted between 8-16 minutes according to embryo size and stage. The riboprobes used were: Oct4 (Scholer et al., 1990) and Nanog (Chambers et al., 2003). Embryos were sectioned in paraffin wax (7 μm).

Images were captured using Volocity (Improvision) software on a Zeiss Stemi SV11 dissecting microscope (for whole embryos), an Olympus IX51 (for cultured cells), or an Olympus BX61 (for sections). Image processing was performed using Adobe Photoshop software. In the case of whole embryos stained with Oct4, fluorescence was visualized using a Leica DM IRE2 inverted confocal microscope (Leica Microsystems). Image acquisition and processing were carried out using the Leica Confocal (Leica Microsystems) and Volocity (Improvision) software packages, respectively.

To quantify fluorescence signal per nucleus, we generated a semi-automated image analysis pipeline. Regions of interest in embryos were cropped and analysed in parallel with an internal negative control region (supplementary material Fig. S5). These RGB images were pre-processed [background subtraction in each channel and gamma correction in the blue channel (DAPI) to enhance weakly fluorescing nuclei] using our own plug-in in ImageJ (http://rsbweb.nih.gov/ij/). To distinguish individual nuclei, the blue channel was segmented using a previously published algorithm (Li et al., 2007) with the following parameters and their respective values for 2D and 3D images: sigma 0.1 (2D);0.15 (3D), minimum nucleus size 350 pixels (2D and 3D), and fusion threshold 2 (2D);1 (3D); see supplementary material Fig. S5). In the resulting image, each nucleus is labelled with a unique greyscale-based identifier. Using a Java application that we developed within eclipse (www.eclipse.org), average pixel intensities in the red and green channels of the preprocessed RGB image were calculated within the superimposed volume of each nucleus. Charts representing red and green average intensities were generated using the Jfreechart library (http://www.jfree.org/jfreechart/). Applications developed for this analysis can be downloaded at http://www.crm.ed.ac.uk/research/group/embryonic-stem-cell-differentiation.

Bisulphite sequencing was performed using the Imprint DNA modification kit (Sigma) together with primers meNanog-F2-S and meNanog-F2-AS (Imamura et al., 2006). Approximately 30 clones were sequenced; sequences characterized by incomplete bisulphite conversion were discarded.

MEFs or E14Tg2a cells (1×10⁷) were used. MF1 embryos, staged as described previously (Downs and Davies, 1993) were dissected free of extra-embryonic tissue (including the allantois containing primordial germ cells). Embryos were pooled [for E7.5 (LB-EHF), 27-30 embryos; for E8.5 (2-8 somites), 20-22 embryos] and dissociated in 0.5% trypsin/2.5% pancreatin/PBS. For E8.5 explants, 20-22 dissociated embryos were cultured in EpiSC conditions with or without 1 mg/ml doxycycline. After 24 hours, explants were trypsinized, replated on non-gelatinized tissue culture dishes for 15 minutes to deplete feeders and collected.

Formaldehyde-crosslinked, sonicated chromatin was prepared as previously described (Navarro et al., 2010) and 20 μg used to prepare FAIRE and reference samples. For the FAIRE sample, protein-free DNA molecules of the crosslinked chromatin were purified by phenol-chloroform extraction and ethanol precipitated as described previously (Giresi et al., 2007). For the reference sample, crosslinking was reversed by overnight incubation at 65°C in TE/1% SDS, phenol-chloroform extracted and ethanol precipitated. FAIRE and reference samples were analysed in parallel using a LightCycler 480 (Roche) and LightCycler 480 SYBR Green 1 Master (Roche). Primer sequences were: Nanog, TGGCCTTCAGATAGGCTGAT (sense) and CAAGAAGTCAGAAGGAAGTGAGC (antisense); Sox2, AGGGCTGGGAGAAAGAAGAG (sense) and CCGCGATTGTTGTGATTAGTT (antisense); Oct4 distal enhancer AGAGTGCTGTCTAGGCCTTA (sense) and CCAGAACTCTCAACCTCCCT (antisense); Oct4 proximal enhancer, GGGAAGCAGGGTATCTCCAT (sense) and TCCCCTCACACAAGACTTCC (antisense).

To examine the relationship between loss of pluripotency and expression of core pluripotency regulators, quantitative RT-PCR (qPCR) was performed on RNA from whole embryos throughout the period covering the reported disappearance of pluripotency (Damjanov et al., 1971). Owing to the natural variation in developmental stage between littermates, precise assessment of the stage of embryo development was made using the morphological criteria of Downs and Davies (Downs and Davies, 1993) (Fig. 1). Expression of Sox2 [which continues in developing neurectoderm (Avilion et al., 2003)] increases throughout this period, whereas Nanog and Oct4 expression declines, becoming undetectable in the somatic cells of three- to five-somite embryos (Nanog) and 12- to 15-somite embryos (Oct4) (Fig. 1A; supplementary material Fig. S1).

The temporal correlation between the reported loss of pluripotency and Nanog expression prompted further analysis of its spatiotemporal expression pattern. As judged by immunofluorescence, in situ hybridization and GFP reporter expression from the endogenous Nanog gene (Chambers et al., 2007) (Fig. 1B; supplementary material Fig. S1A), Nanog is strongly expressed in the proximal-posterior region from the start of gastrulation, consistent with published data (Hart et al., 2004; Hatano et al., 2005). Expression is lost as cells delaminate to move through the primitive streak (supplementary material Fig. S1A). Nanog is excluded from the distal/anterior epiblast (Fig. 1B; supplementary material Fig. S1A,B) and subsequently disappears entirely from the epiblast at the onset of somitogenesis. Expression is re-established in primordial germ cells (PGCs), as documented previously (Chambers et al., 2003; Yamaguchi et al., 2005). In the embryo proper, the last detectable expression of GFP is localized to the distal tip at 0-1 somites (Fig. 1B). This region was therefore tested for pluripotency from headfold to early somite stages. Kidney capsule grafts from presomite embryos produced large teratocarcinomas containing derivatives of all three germ layers and EC cells (two out of four tumours were teratocarcinomas), consistent with published data (Beddington, 1983) (Fig. 1C; supplementary material Fig. S2). However, the equivalent tissue from two- to five-somite embryos produced small growths devoid of EC cells and endoderm (n=23) (Fig. 1C). qPCR analysis of the distal tip at two to five somites revealed expression of Oct4 and Sox2, but confirmed that Nanog transcripts were absent (supplementary material Fig. S1E). Therefore, pluripotency disappears at early somitogenesis, coincident with extinction of Nanog but before Oct4 becomes undetectable.

We next investigated the concordance of EpiSC derivation and teratocarcinoma formation. EpiSC lines have previously only been derived up to E6.5 (Brons et al., 2007). Upon explantation of tissue from E7.5-E8.0 pre-somite embryos in EpiSC conditions, Nanog:GFP was readily detected in explants. These explants could be passaged and in all cases (8/8) gave rise to EpiSC lines that expressed Oct4 and Nanog protein (Fig. 1D). By contrast, explants from two- to four-somite embryos did not express Nanog:GFP (0/6). Upon passaging, these explants produced morphologically differentiated cells that did not express Oct4 or Nanog (Fig. 1D). Thus, the ability to derive EpiSC lines is lost concomitant with the loss of teratocarcinoma-forming potential, indicating that EpiSC derivation is an effective assay for post-implantation pluripotency.

To examine the relationship between pluripotency and Nanog expression in more detail, a staged series of Nanog:GFP embryos was isolated, separated into distal/anterior (Nanog-nonexpressing) and proximal/posterior (Nanog-expressing) regions, and explanted in EpiSC culture conditions (Fig. 2). The presence of bright GFP-positive PGCs at the base of the allantois allowed us to confirm exclusion of these cells from posterior explants. Without exception, prior to somitogenesis, anterior and posterior explants showed Nanog expression at 24-48 hours (Fig. 2A) and produced stable EpiSC lines expressing characteristic markers (n=12 for each; Fig. 2D,E; supplementary material Fig. S3C). Levels of marker expression in all EpiSC lines tested in the present work were within the limits of variation observed previously for EpiSC lines (Brons et al., 2007; Guo and Smith, 2010; Guo et al., 2009; Hayashi and Surani, 2009). In addition, prior to somitogenesis, anterior and posterior explants showed equivalent teratocarcinoma-forming capacity (three out of four anterior versus six out of nine posterior-derived tumours were teratocarcinomas; supplementary material Fig. S3A,B). In striking contrast, not a single explant from embryos that had initiated somitogenesis, including the two late headfold stage embryos shown in Fig. 2B, where somitogenesis is incipient, reactivated Nanog or formed EpiSC lines, from anterior or posterior tissue (0/11 of each). In total, 23 EpiSC lines from presomitogenesis-stage embryos were established and passaged three to nine times (supplementary material Table S1). No lines were isolated from somite-stage embryos (n=21). Therefore, there is a dramatic drop in pluripotency that occurs simultaneously in anterior and posterior tissue between headfold and somitogenesis stages. Together, these results indicate that Nanog re-expression in vitro correlates with EpiSC derivation and teratocarcinoma formation, and thus with pluripotency. Furthermore, the ability of cells to activate Nanog expression within 24 hours of explantation in EpiSC conditions (Fig. 2A) demonstrates plasticity of gene expression within the anterior epiblast at this stage.

Two possibilities may explain the observation that Nanog-nonexpressing epiblast cells are pluripotent prior to somitogenesis. Either these cells represent a latent pluripotent state that requires Nanog reactivation to manifest pluripotency, or Nanog is not essential for epiblast pluripotency. To distinguish between these possibilities, Nanog^–/– ES cells were used to produce E6.5-E7.5 chimaeras (Fig. 3A,B), which were then explanted and tested for EpiSC growth (Fig. 3A,B). We analysed Nanog^–/– cell lines TβC44cre6 and RCNβH-B(t), alongside a control derivative line RCNβH-B(t)R, in which one of the non-functional Nanog alleles has been repaired by homologous recombination (Chambers et al., 2007). Discrimination of ES cell-derived from wild-type host cells is possible because RCNβH-B(t) and RCNβH-B(t)R express GFP constitutively, whereas TβC44cre6 expresses GFP from a targeted Nanog-null allele. Nanog^–/– cells formed chimaeras indistinguishable from wild-type or Nanog^–/+ cells (Fig. 3B). Explanted colonies of Nanog^–/– cells were picked and expanded to produce two independent EpiSC lines that lacked Nanog but showed robust Oct4 and Sox2 expression (Fig. 3B; supplementary material Fig. S4A).

To exclude the formal possibility that initial derivation of chimaera-derived Nanog^–/– lines required co-culture with wild-type cells, pure Nanog^–/– ES cell populations were cultured in EpiSC conditions, previously shown to permit in vitro EpiSC derivation (Guo et al., 2009). Two independent in vitro-derived Nanog^–/– EpiSC lines that expressed EpiSC-specific markers were readily obtained (Fig. 3C,D; supplementary material Fig. S4B). Nanog^–/– EpiSC produced teratocarcinomas following kidney capsule grafting (Fig. 3E,F). Thus, in contrast to the necessity for Nanog in the acquisition of the ‘ground state’ pluripotency found in ES cells (Silva et al., 2009), Nanog is neither required to transit to nor to maintain EpiSC pluripotency, although Nanog expression marks both pluripotent states in vitro.

Previously, Nanog has been proposed as the key intermediary downstream of activin that is responsible for maintaining self-renewal of both human ES cells and mouse EpiSC (Vallier et al., 2009). However, as shown above, EpiSCs can be established and maintained without Nanog. To further investigate the signalling requirements of EpiSCs, epiblast tissue was explanted in the presence of signalling pathway inhibitors. In the presence of an activin inhibitor (but not a MEK/Erk inhibitor) epiblast tissue lost expression of Nanog:GFP (Fig. 4A). Moreover, similar to Nanog^+/– Nanog: GFP explants, Nanog^–/– EpiSC lines treated with activin inhibitor failed to maintain Nanog:GFP reporter expression at 48h (Fig. 4B,C). By contrast, Nanog^–/– cells cultured in MEK/Erk inhibitor maintained Nanog:GFP expression at 48 hours. After 118 hours, Nanog:GFP expression and EpiSC morphology were lost in both conditions. Together, these results indicate that, in EpiSCs, Nanog is a target of activin signalling (Greber et al., 2010) but Nanog is not necessary for activin-dependent self-renewal (Vallier et al., 2009).

We next examined the spatiotemporal expression of Oct4. Until the late bud stage, Oct4 is widely expressed (Scholer et al., 1990; Yeom et al., 1996). However, from early headfold stage, levels of Oct4 mRNA and protein are consistently lower in the anterior part of the embryo containing the prospective forebrain, than more posterior regions (Fig. 5A,B) (Downs, 2008). Image analysis of Oct4 immunofluorescence shows a range of fluorescence intensity levels per nucleus, with consistently lower levels appearing in the anteriormost region at all headfold stages (Fig. 5C; supplementary material Fig. S5). Fluorescence intensity declined in both regions with increasing embryonic age (Fig. 5C). We tested Nanog reactivation in explanted tissue from two microdissected regions expressing different levels of Oct4: region 1 [low Oct4], corresponding to prospective forebrain, and the adjoining region 2 [higher Oct4] (Fig. 5C-F). The efficiency of Nanog reactivation in EpiSC culture was consistently lower in region 1 than region 2 through headfold stages (Fig. 5F). Moreover, the efficiency of Nanog-reactivation in both regions 1 and 2 declined with developmental stage (Fig. 5F) such that by early somitogenesis (zero to two somites) only an extremely low proportion of colonies expressed Nanog. Interestingly, the frequency of Nanog reactivation in each region at a given stage correlates with the frequency of cells showing immunofluorescence above a consistent threshold fluorescence intensity value (indicated by the dashed line in Fig. 5C), suggesting a minimum threshold of Oct4 for Nanog reactivation. The larger sample size at zero to two somites in these experiments compared with those in Fig. 2C bring increased resolution to the time period in which pluripotency is lost. Together, these data suggest that sufficiently high Oct4 levels may be important in embryonic pluripotency.

To determine the time over which this drop in Nanog reactivation occurs, we measured the duration of the headfold stage. Downs and Davies (Downs and Davies, 1993) and our independent observations indicate that late headfold stage embryos never constitute a majority of embryos dissected at 6-hourly intervals from E6.75-E8.25, suggesting that the duration of this stage is short (supplementary material Fig. S6). We cultured individually staged E6.5-E7.5 (late streak to late bud stage) embryos for 7-16 hours. All embryos developed according to the schedule expected from fresh dissections of embryos from timed matings (supplementary material Fig. S6A,C). This indicates that the culture conditions allow accurate reproduction of normal development over this period. Embryos cultured from early headfold stage developed to late headfold stage within 3 hours, and by 4 hours had formed one or two somites (n=3) (supplementary material Fig. S6B,C). Late headfold stage embryos had formed two or three somites within 4 hours (n=5) (supplementary material Fig. S6B,C). Therefore, the period of Oct4 downregulation critical to pluripotency loss at headfold stage is no longer than 4 hours.

To test the hypothesis that the reduction in Oct4 levels results in loss of embryonic pluripotency, the consequences of re-expressing Oct4 in somitogenesis-stage embryos were examined using mice carrying an Oct4 transgene (TgOct4) that is ubiquitously expressed in response to doxycycline administration (Hochedlinger et al., 2005). Explants from anterior (forebrain) and posterior (primitive streak) regions of somitogenesis-stage TgOct4/+;rtTA/+ embryos up to E15.5 were tested for EpiSC colony formation and Nanog reactivation in the presence or absence of doxycycline (Fig. 6A). Doxycycline efficiently induced Oct4 expression, which was undetectable in untreated explants (Fig. 6B; supplementary material Fig. S7A). In explants from E8.5-E10.5, Nanog:GFP was induced in doxycycline-treated explants within 24 hours (Fig. 6B; supplementary material Fig. S7B) and both Nanog and Sox2 were robustly expressed by 48 hours (Fig. 6C; supplementary material Fig. S7C,E). Moreover, image analysis shows that Nanog is only reactivated in Oct4-expressing cells, confirming that Nanog reactivation is cell-autonomous (supplementary material Fig. S7D). Interestingly, cells with the highest intensity Oct4 immunofluorescence did not show Nanog reactivation, suggesting both a lower and upper threshold of Oct4 expression for optimal Nanog reactivation (supplementary material Fig. S7D). The majority (62%) of colonies expressed Nanog protein in 48-hour explants at E8.5 (n=68; Fig. 6B), and, of these, 25% of Oct4-expressing cells were Nanog protein-positive (n=11241; supplementary material Fig. S7D). The frequency of Nanog reactivation declined thereafter until E11.5 (3% of colonies; n=143) (Fig. 6B). At E13.5-E14.5, we observed no Nanog-positive colonies at 48 hours, but after a more prolonged period (120 hours) a few colonies had become Nanog positive (Fig. 6B). By E15.5, we found no Nanog-positive colonies in doxycycline-treated explants, even after 120 hours (Fig. 6B). Thus, Nanog reactivation in EpiSC culture is initially highly sensitive to raised Oct4 levels, but reactivation gradually becomes refractory to these conditions.

Despite the delay in reactivation of Nanog at E13.5, EpiSC lines expressing characteristic markers could be derived from embryos up to this stage (Fig. 6D,E; supplementary material Fig. S8A). A total of 34 E9.0-E13.5 EpiSC lines have been established and passaged up to 30 times (supplementary material Table S1). As seen in presomite stage EpiSC lines, marker gene expression levels varied (Fig. 6E), but were within the range of previously reported expression levels (Brons et al., 2007; Guo and Smith, 2010; Guo et al., 2009; Hayashi and Surani, 2009). Crucially, a representative E10.5 EpiSC line showed activin and MEK/ERK-dependent, Jak/STAT signalling-independent propagation (supplementary material Fig. S9). Moreover, teratocarcinoma-forming capacity was retained after prolonged culture in the presence of a Jak inhibitor (Fig. 6F; supplementary material Fig. S8B). To detect the presence of self-renewing cells in the tumour, secondary EpiSC were isolated. After four passages, these lines showed characteristic EpiSC morphology and expressed Oct4, Nanog and Sox (supplementary material Fig. S8C). To test whether Oct4 induces teratocarcinoma-forming potential without prior EpiSC line establishment, freshly dissected TgOct4/+;rtTA/+ E8.5 (five- to 11-somite stage) forebrain regions were grafted to the kidney capsule of wild-type hosts. Graft recipient animals were maintained with or without doxycycline (Fig. 6F). After 4 weeks, two out of three doxycycline-treated grafts had formed large teratocarcinomas containing derivatives of all three germ layers (Fig. 6F; supplementary material Fig. S8B), whereas untreated tissue formed only small, differentiated tumours. Taken together, these results indicate that threshold levels of Oct4 are necessary to maintain pluripotency in the presomitogenesis embryo, as well as to mediate its re-acquisition in somitogenesis-stage, non-pluripotent cells.

To investigate whether the loss and reacquisition of pluripotency is underpinned by chromatin changes, we examined nucleosome positioning at the regulatory regions of Oct4 and Nanog (Fig. 7A). Analysis of nucleosome-depleted chromatin using formaldehyde assisted isolation of regulatory elements (FAIRE) (Giresi et al., 2007) showed prominent changes between presomite and early somite stages. Despite the mixture of cell types (including those expressing low level Oct4 and Nanog) present in embryos at these stages, we detected a profound reduction in the overall accessibility of chromatin at both Oct4 and Nanog regulatory regions in whole embryos (Fig. 7B), correlating with the observed changes in transcript levels. This chromatin closure was reversed in two- to eight-somite stage Oct4-expressing explants after 24 hours in EpiSC culture (Fig. 7C). Although both Nanog and Oct4 chromatin reopened following Oct4 induction, Oct4 regulatory elements did not show as extensive reopening as Nanog (Fig. 7C). This is noteworthy given that Nanog expression is extinguished prior to Oct4 and suggests that genes of the pluripotency network do not all recover simultaneously. Moreover, activin/FGF culture conditions alone can provoke chromatin reopening at Nanog in the absence of ectopic Oct4 expression and Nanog reactivation (Fig. 7C). This suggests that activin/FGF can stimulate nucleosomal displacement at the Nanog promoter but that binding of Oct4 to the Oct/Sox motif is required to stimulate Nanog transcription.

The failure to reactivate Nanog in post-pluripotent tissue and the decreasing efficiency of reactivation in presence of elevated Oct4 expression suggested that progressive stabilizing changes may occur at Nanog. We therefore examined the extent of CpG methylation in E7.5-E14.5 embryos at the Nanog promoter (Fig. 7D). This region has been shown to exhibit much higher levels of methylation in MEFs compared with embryonic stem cells (Imamura et al., 2006). We found that methylation levels were low at both presomite and early somite stages (Fig. 7D), but increased as development progressed. Moreover, analysis of an established EpiSC line derived from E13.5 Oct4-overexpressing forebrain tissue showed that DNA methylation levels at the Nanog promoter resemble those of pre-somite E7.5 embryos (Fig. 7D). These findings show that Oct4-induced Nanog reactivation and acquisition of pluripotency in somitogenesis-stage explants is accompanied by lowered methylation at Nanog, whereas the progressive decrease in its efficiency late in development correlates with increasing Nanog promoter methylation.

Here, we examine the role of two core pluripotency factors, Nanog and Oct4, in pluripotency during post-implantation development. We show that Nanog expression in vitro is a reliable indicator of pluripotency but is dispensable for the maintenance of post-implantation pluripotency. We have identified a novel transition state when cells are no longer pluripotent but can rapidly reacquire pluripotency upon re-expression of Oct4. Our work suggests that two mechanisms operate to ensure that pluripotent cells are eliminated from the somitogenesis-stage embryo. The first is a rapid chromatin-based shutdown that is susceptible to perturbation by variation in the levels of pluripotency network components, and the second is a slower, more permanent, DNA methylation-based stabilization of the non-pluripotent state (Fig. 7E).

The ability of Nanog-null epiblast cells to form EpiSCs rigorously establishes that Nanog is neither a necessary mediator of activin signalling in EpiSC, nor is it required for EpiSC maintenance or derivation. Interestingly, established ES cell lines, which also have no absolute requirement for Nanog, nevertheless show decreased self-renewal efficiency in the absence of Nanog (Chambers et al., 2007). Whether similar quantitative effects of Nanog on self-renewal or differentiation of EpiSCs exist remains an unanswered question.

The efficient reactivation of pluripotency in the time window between the onset of somitogenesis (E8.0) and late organogenesis (E10.5) following re-expression of Oct4 demonstrates that pluripotency, and therefore teratocarcinogenesis, could result from the reactivation of expression of a single gene during this time. Reactivation of Oct4 in 3-week-old animals does not result in teratocarcinoma formation (Hochedlinger et al., 2005), suggesting that the competence of cells to respond to elevated Oct4 by re-establishing the pluripotent state is limited to earlier stages. E8-E10 in mouse corresponds approximately to human Carnegie stages 9-14 (20-33 days post ovulation) (Kaufman, 1992). Such an early developmental origin for human neonatal teratomas would be consistent with the occasionally very large size of such tumours and appearance in ultrasound scans as early as 13.5 weeks in human foetuses (Holcroft et al., 2008). The efficient formation of teratocarcinomas from primary somatic tissue in mouse, which has been unambiguously separated from locations occupied by PGCs, indicates that a somatic origin for these tumours in humans is possible. This therefore calls into question the prevailing notion that human extragonadal teratocarcinomas have an obligatory origin from misplaced PGCs (Oosterhuis and Looijenga, 2005).

The window of rapid and efficient reactivation of pluripotency is qualitatively distinct from other reports of reprogramming differentiated cells to pluripotency. Oct4-induced reprogramming to pluripotency has been achieved for a variety of cell types (Kim et al., 2009a; Kim et al., 2009b; Tsai et al., 2011). However, these reprogramming processes take a minimum of 18 days (Tsai et al., 2011). This difference cannot be explained by the time taken to generate iPS cells rather than EpiSC, as generation of iEpiSC from mouse embryonic fibroblasts using Oct4, Myc, Klf4 and Sox2 together only occurs after 2 weeks and only exceeds iPS cell efficiency by a factor of 4 (Han et al., 2011). This reinforces the conclusion that the E8.0-E10.5 stage marks a distinct transition state prior to the stable shutdown of pluripotency.

We observe that Oct4 and Nanog levels are reduced in early post-implantation embryos compared with ES cells, and decline further during post-implantation development. Furthermore, several transcription factors (e.g. Esrrb, Klf4 that are thought to function combinatorially with Oct4, Sox2 and Nanog in ES cells and in pre-implantation embryos, are downregulated during post-implantation development (Han et al., 2010). Interestingly, the period from bud stage to early somitogenesis is characterized by the downregulation of many pluripotency associated genes (Mitiku and Baker, 2007). The regionalization of Nanog and Oct4 during gastrulation further points to a progressive dismantling of the pluripotency network. We observe a rapid loss of pluripotency, in under 4 hours, in cells where Oct4 protein levels are diminishing but still present, suggesting that threshold concentrations of this factor, rather than its absolute presence, determine pluripotency. Therefore, the gradual decrease in expression levels of individual pluripotency transcription factors may be rapidly converted into dramatic chromatin condensation at pluripotency gene regulatory elements over a short developmental period. This could manifest itself as a sudden collapse in the overall activity of the pluripotency gene regulatory network and, consequently, of pluripotency.

The disappearance of nucleosome-depleted regions may be an effect of this reduction in pluripotency factor concentrations, but could also result from an increase in histone modifications associated with repression. The observation that both distal and proximal Oct4 enhancers and the Nanog promoter show chromatin opening following Oct4 transgene induction is consistent with the observations that Oct4 binds these regulatory regions in EpiSCs (Tesar et al., 2007). However, the accessibility of chromatin at the Nanog promoter also increases after explantation without Oct4 induction, when Nanog is not expressed. This suggests that chromatin re-opening is not simply a consequence of transcriptional activity, nor is it obligatorily linked to an increase in Oct4 concentration. The apparent increase in accessibility of the Sox2 promoter in activin/FGF in the absence of Oct4 induction may reflect its activity in neural progenitors, as Fgf2 can promote the growth of neural stem cells from embryos of this stage (Tropepe et al., 1999).

Our results demonstrate that nucleosome depletion at the Nanog proximal promoter is a crucial readout of a pluripotent state. Our finding that increased nucleosome occupancy at the Nanog promoter is reversible by increased Oct4 expression is interesting in light of the recent report that Oct4 binding maintains a nucleosomes-depleted region at the Nanog promoter in human EC cells (You et al., 2011). This study provides evidence that enforcing Oct4 expression in differentiated cancer cell lines restored regions of nucleosome depletion but only in the absence of DNA methylation. The same study reported that increased nucleosome occupancy during EC differentiation precedes DNA methylation, which also inhibits restoration of nucleosomes-depleted regions by forced Oct4 expression. Our data show that similar events occur in the more physiological context of mouse embryogenesis. This offers a potential explanation for the decreasing ability of enforced Oct4 expression to resuscitate pluripotency in explanted tissue and support a model in which the long-term stabilization of Nanog suppression is achieved by progressively increasing levels of methylation at Nanog between E8.5 and E15.5.

The high levels of Nanog in the proximal/posterior part of the epiblast, together with its disappearance on E7.5 and the remnant of expression in the node coincide with areas of Nodal expression and suggest that, just as in EpiSCs, Nanog expression is induced by Nodal/activin signalling in the post-implantation embryo. Thus, the presence of Nodal antagonists in the anterior visceral endoderm, which restrict Nodal signalling to the posterior of the embryo (Perea-Gomez et al., 2002), may also restrict Nanog to the proximal/posterior region. On dispersal of cells at explantation, the endoderm and epiblast separate, and thus in EpiSC culture conditions, epiblast cells are removed from Nodal antagonists and subjected to more uniform activin signalling. This could account for the rapid reactivation of Nanog in presomitic distal/anterior explants and the ability of inhibitors of activin/Nodal signalling to block reactivation in anterior explants.

During somitogenesis, when pluripotency has been extinguished from the epiblast, FGF and Nodal are used for organ-specific patterning roles. For example, Fgf8 patterns the mid/hindbrain junction (Crossley et al., 1996), while Nodal directs left/right asymmetry (Kawasumi et al., 2011). These events may dictate that by early somitogenesis stage the activation of the pluripotency network as a response to Nodal and FGF signalling is detrimental, and therefore requires efficient shutdown just prior to organogenesis.

It is likely that activity of the pluripotency network is not required by E7.5, as cell fates at this stage are highly regionalized. In particular, the region corresponding to Nanog-negative anterior epiblast is exclusively fated for neurectoderm (Forlani et al., 2003; Lawson et al., 1991; Lawson and Pedersen, 1992). The fact that these cells retain pluripotency and reactivate Nanog upon explantation indicates plasticity in gene expression during this pre-commitment stage coupled to an ability to reassert the activity of the pluripotency gene regulatory network upon appropriate environmental alteration represented by activin/FGF. This plasticity is lost at the onset of somitogenesis and marks the natural end of pluripotency at a time when Oct4 levels are continuing to decline. The subsequent resuscitation of pluripotency by the activation of a single transcription factor establishes that critical commitment events can hinge on modest changes in the activity of a single gene product, in this case, Oct4.

We thank John Agnew, Douglas Colby, Carolyn Manson, Renée McLay, Lynsey Robertson, Frances Stenhouse and Jan Ure, for technical assistance, and Clare Blackburn, Josh Brickman, Huw Jones, Keisuke Kaji, Pablo Navarro and Jenny Nichols for comments on the manuscript.