Reduced Oct4 Expression Directs a Robust Pluripotent State with Distinct Signaling Activity and Increased Enhancer Occupancy by Oct4 and Nanog

1Pluripotency is promoted and stabilized by a pluripotent-cell-specific gene regulatory network (GRN). Central to this network are the transcription factors Oct4, Sox2, and Nanog. ESCs fluctuate between states of high Nanog expression, possessing high self-renewal efficiency and low Nanog expression, exhibiting increased differentiation propensity. Notably, Nanog is also expressed heterogeneously in the preimplantation embryo (;; low Nanog expression facilitates primitive endoderm differentiation, and high Nanog expression maintains epiblast identity. The importance of the Nanog-low state for proper embryonic development is underscored by the finding that embryos expanded from the eight-cell stage in the presence of drugs that inhibit MEK and fail to develop a primitive endoderm; conversely, the supplementation of embryos in culture with fibroblast growth factor (FGF) results in embryos in which epiblast identity is lost in favor of primitive endoderm. In ESCs, MEK inhibition upregulates Nanog. How FGF regulates Nanog is unknown, but the process may involve the allelic regulation of Nanog through downstream MEK targets. In addition to Nanog, other pluripotency factors, such as Esrrb, Rex1, Stella, Klf4, and Tbx3, are expressed heterogeneously in ESCs. Although the mechanisms generating ESC heterogeneity have begun to be characterized, how populations of ESCs expressing homogeneously high levels of transcription factors arise in the presence of serum is not well understood.

2Recently, ESCs cultured in the presence of MEK and GSK inhibitors have been reported to express a more homogenous level of pluripotency transcription factors. However, the maintenance of this “ground state” requires a continuous supply of MEK and GSK 3 inhibitors. Therefore, it is of considerable interest to determine whether a robust pluripotent state can exist in the absence of such pharmacological inhibition. Oct4 levels, generally considered to be uniform in undifferentiated ESCs, are critical not only in enabling self-renewal but also for the outcome of differentiation. However, these prior studies focused either on the complete elimination or overexpression of Oct4. Here, we demonstrate that Oct4 levels in ESCs cultured in serum occupy a broad unimodal peak; genetic exclusion of the highest Oct4 expressers enables the robust propagation of pluripotency with leukemia inhibitory factor (LIF) as the sole requirement. In this condition, the binding of Oct4 to key regulatory nodes of the pluripotency GRN is enhanced, and signaling responses are tilted in favor of self-renewal.

3Global chromatin localization and transient expression assays suggest that Oct4 positively regulates Nanog gene expression. A clear prediction from this is that cells that have reduced Oct4 levels should have reduced Nanog levels. Therefore, it was with surprise that we noted that multiple ESC lines in which homologous recombination had introduced a drug-resistance gene trap cassette to the Oct4 locus, expressed elevated levels of Nanog protein and messenger RNA (mRNA) when compared to Oct4+/+ ESCs (Figures 1A and 1B; see Table S1 [available online] for a list of cell lines used in this study). Previously, Nanog has been demonstrated to be expressed heterogeneously in undifferentiated ESCs. Therefore, this increase in Nanog protein could be due to either a general increase in the level of Nanog per cell or an alteration in the proportion of cells in the Nanog-negative compartment. Immunofluorescence analysis showed that Oct4+/− ESCs expressed Nanog more uniformly throughout the colonies than Oct4+/+ ESCs and at levels comparable to those present in Nanog-high nuclei in Oct4+/+ colonies (Figure 1C).

4Although Oct4 protein has been reported to be expressed at 65% wild-type (WT) levels in populations of Oct4+/− ESCs by immunoblot analysis, the range of expression in individual cells has not been addressed. Immunofluorescence analysis (Figure 1D) shows that the range of Oct4 levels in individual Oct4+/− cells strongly overlaps the distribution of Oct4 levels seen in WT cells. Not only do Oct4+/− ESC colonies have a reduced number of cells that express high levels of Oct4 (Figure 1D), but, interestingly, cells that lack Oct4 are also missing (Figure 1D), a point we return to below. Moreover, aggregation of Oct4+/− ESCs with WT embryos showed that the Oct4 protein levels in Oct4+/− cells were within the range observed in WT inner cell mass (ICM) cells (Figure 1F). Intracellular fluorescence-activated cell sorting (FACS) analysis in bulk ESC cultures showed that the levels of both Oct4 and Nanog present in individual Oct4+/− ESCs are within the normal range found in WT ESCs (Figure 1E). In addition, although there is no effect of altering the Nanog dosage upon the range of Oct4 expression levels, as recently reported, reducing Oct4 to the level present in Oct4+/− ESCs shifts the level of Nanog to the higher end of the range detected in WT ESCs. Altogether, these results indicate that the increase in Nanog expression in Oct4+/− cell lines is due to a reduction in the proportion of Nanog-low and Nanog-negative cells.

5To examine dynamic aspects of heterogeneous transcription factor regulation, the previously described targeting vector conferring Nanog-directed expression of green fluorescent protein (GFP) and puromycin resistance was introduced into two different Oct4 mutant cell lines via homologous recombination (Figures S1A and S1B). The temporal response to puromycin withdrawal confirmed that Oct4+/− ESCs retain homogenous GFP expression under conditions in which typically heterogeneous GFP expression emerges in WT cultures (Figure 1G).

6The correlation between homogeneous Nanog expression and Oct4 heterozygosity in independent cell lines suggests a causative role for the Oct4 protein level in the generation of Nanog heterogeneity. To test this hypothesis, the effect of increasing the Oct4 level was examined with ZHTc-Nanog:GFP cells (Figures S1A and S1 B and Table S1). In addition to the Nanog:GFP reporter, these cells also contain a doxycycline-suppressible Oct4 transgene. If homogeneous Nanog expression was caused by reduced Oct4 levels, then raising the Oct4 level by titrating down the doxycycline concentration should restore Nanog heterogeneity. This was investigated by immunofluorescence (Figure 2A) and intracellular FACS (Figures 2B, S2A, and S2B). No changes in Oct4 or Nanog were seen during the first 2 days; however, by day 3, increasing Oct4 expression was detected at doxycycline concentrations of 0.3 ng/ml or less (Figure S2A). Interestingly, the level of Oct4 expression obtained with 0.03 ng/ml doxycycline approached, but did not exceed, the level observed in the Oct4+/+ control line E14Tg2a-Nanog:GFP. Decreased Nanog expression also became detectable by day 3, and the population resolved into a bimodal expression pattern by day 4 (Figures S2A and S2B). Importantly, bimodal Nanog expression was achieved at the doxycycline concentration (0.03 ng/ml) that elevated Oct4 toward WT levels and did not require Oct4 overexpression (Figures S2A and 2A). The restoration of Nanog heterogeneity was reflected by the increased coefficient of variation of Nanog:GFP at doxycycline concentrations of 0.1 ng/ml or less (Figure S2B). These results indicate that the lack of Nanog-low and Oct4-low cells in Oct4+/− cultures is not due to aberrant selection for cells with an enhanced self-renewal propensity but is a consequence of the Oct4 expression level; restoring the Oct4 level toward, but not beyond, WT levels is sufficient to restore Nanog heterogeneity. Dynamically, the increase in Oct4 level results in decreased Nanog expression in a subpopulation of cells, and, subsequently, Oct4-low cells emerge from this Nanog-low population (Figure 2B).

7The Nanog-low cells formed by doxycycline treatment of the ZHTc-Nanog:GFP cells expressed high SSEA1+, suggesting that they were undifferentiated (Figure S2C). Nevertheless, the question remained as to whether Nanog-low cells induced in ZHTc-Nanog:GFP cultures after Oct4 upregulation were already committed to differentiate. Therefore, as schematized in Figure 2D, SSEA1+-GFP− cells were sorted from ZHTc-Nanog:GFP cells cultured in reduced doxycycline and replated in culture with 1 μg/ml doxycycline. By day 4, 35% of the sorted GFP− cells had reverted to a GFP+ state (Figure 2D). This not only indicates that GFP− cells can remain undifferentiated but also confirms the crucial role of Oct4 in facilitating switching both from Nanog-high to Nanog-low states and also from Nanog-low to Nanog-high states.

8Immunofluorescence analysis showed that, similar to Nanog, Esrrb and Klf4 were expressed relatively homogeneously in Oct4+/− cultures (Figure S1C). Upon the restoration of Oct4 toward WT levels, both Esrrb and Klf4 became heterogeneously expressed (Figure 2C). Therefore, these findings identify Oct4 as a critical facilitator of the switch between heterogeneous and homogeneous states of transcription factor expression in ESCs.

9Oct4 has previously been shown to be necessary for ESC self-renewal, given that the elimination of Oct4 results in trophectodermal differentiation, whereas overexpression causes a proportion of cells to differentiate into cells expressing endoderm and mesoderm markers. How Oct4 mediates this differentiation remains unknown. A priori, it might be expected that Oct4+/− cells would differentiate more rapidly. However, the foregoing results showing that Oct4+/− cells have a reduced proportion of Nanog-negative cells, along with the model that differentiation occurs more effectively through Nanog-low cells, prompted an investigation of this point. Upon LIF withdrawal, Oct4+/− cell lines showed delayed signs of morphological differentiation (Figure 3A), delayed downregulation of the pluripotency markers Rex1 and FGF4, and reduced upregulation of the differentiation markers Brachyury and Sox17 (Figure 3B) in comparison to Oct4+/+ ESCs.

10Neural differentiation was also delayed in Oct4+/− ESCs (Figures 3C and 3D). The persistence of large numbers of Oct4-expressing cells in Oct4+/− cultures (Figure 3D) indicates that these cells were unable to undergo efficient neural differentiation not because they were being diverted into nonneural lineages but because they remained undifferentiated.

11Next, the ability of Oct4+/− ESCs to differentiate into epiblast stem cells (EpiSCs) was investigated by replating Oct4+/− and Oct4+/+ ESCs in activin and FGF. Oct4+/− ESCs showed a slower reduction in the expression of Nanog:GFP, Pecam (Figure 3E), and ESC-specific transcripts and a delayed, diminished induction of EpiSC transcripts (Figure 3F). To functionally analyze commitment, we replated cells into GMEMβ-FCS-LIF after one or two passages in N2B27-activin-FGF2. Cultures of Oct4+/− cells retain a greater ability to form alkaline phosphatase-positive undifferentiated ESC colonies than Oct4+/+ cultures (Figure 3G). These results indicate that Oct4+/− ESCs are compromised in their ability to efficiently commit toward differentiation and that this delay is in the initial exit from the ESC state.

12Nevertheless, EpiSC lines could be established from Oct4+/− ESCs by six passages in activin-FGF. Once established, cells were passaged for an additional 2 weeks in the presence of a JAK inhibitor in order to ensure the complete elimination of any residual ESCs; effective JAK inhibition was confirmed with ESCs (Figure 3H). Despite prior JAK inhibition, Oct4+/− EpiSCs were able to form occasional colonies when replated in GMEMβ-FCS-LIF (Figure 3I), suggesting that, in the presence of reduced Oct4 protein, the stringent separation between ESC and EpiSC states is compromised. Oct4+/− EpiSCs also displayed a greater propensity to generate induced pluripotent stem cells than WT cells, as shown by the increased colony number formed after the expression of Esrrb and prior to replating in GMEMβ-FCS-LIF (Figure 3I).

13Our previous studies suggested that differentiation proceeds efficiently after transient downregulation of Nanog. This predicts that the retarded differentiation kinetics of the Oct4+/− cells could be rescued by restoring a Nanog-low compartment to the population. Therefore, we used ZHTc-Nanog:GFP cells to vary the Oct4 level (as outlined in Figure S3A) and examined differentiation kinetics. Transient treatment of ZHTc-Nanog:GFP cells for 4 days with the doxycycline concentration shown in order to elevate Oct4 expression toward WT levels (0.03 ng/ml; Figure S2A) produced a decrease in Nanog mRNA (Figure S3B) and the heterogeneous expression of Nanog:GFP (Figure S3A). Continued induction of the Oct4 transgene during the differentiation assays was prevented by reapplying doxycycline at 1 μg/ml. After 8 days in the neural differentiation protocol, cells in which heterogeneous Nanog expression was initially imposed showed increased expression of neural markers in comparison to cells maintained in 1 μg/ml doxycycline (Figures S3C and S3D). The response of cells to LIF withdrawal was also assessed. Cells in which heterogeneous Nanog expression was initially induced showed a more differentiated morphology (Figure S3E) and an enhanced expression of differentiation markers (Figure S3F) in comparison to cells maintained in 1 μg/ml doxycycline. These results indicate that the slower differentiation of Oct4+/− cells is attributable to a lack of heterogeneous Nanog expression and that this can be rescued by restoring the full range of Oct4 levels observed in WT ESCs.

14The above results suggest that differentiation induced by Oct4 elevation occurs via production of a differentiation-prone Nanog-low population. If so, differentiation should be blocked by enforced Nanog expression. This was tested with episomally supertransfectable cells carrying a loxP-flanked Nanog transgene (Nanog/T cells). In comparison to control cells, which differentiated upon Oct4 transfection (Figure S4A), Nanog/T cells formed morphologically undifferentiated colonies (Figure S4A). To establish that this differentiation blockade was due to transgenic Nanog expression, we deleted the loxP-flanked Nanog ORF by Cre transfection. Cre-mediated Nanog excision results in CAG-driven GFP expression. Although most green colonies formed by Cre treatment of empty vector transfectants had an undifferentiated morphology (Figures S4B and S4C), the majority of green colonies derived from Oct4 transfectants were completely differentiated (Figures S4B and S4C). Therefore, Oct4 induces differentiation by the production of a Nanog-low cell population.

15In comparison to Oct4+/+ cell lines, clonal self-renewal assays show that independent Oct4+/− ESC lines have an enhanced ability to form uniformly undifferentiated colonies in response to saturating LIF concentrations (Figures 4A and 4B). Further clonal self-renewal assays showed that Oct4+/− ESCs have ∼30-fold greater sensitivity to LIF (Figure 4C). FACS analysis of the effect of LIF concentration on Nanog expression showed that, although Nanog was heterogeneous in WT cells even at the highest LIF concentration, 10 U/ml LIF was sufficient to maintain a unimodally high Nanog:GFP profile in Oct4+/− ESCs (Figure 4D). These results indicate that Oct4+/− ESCs show enhanced responsiveness to LIF stimulation.

16In N2B27 supplemented with LIF and bone morphogenetic protein (BMP), both Oct4+/+ and Oct4+/− cells formed self-renewing colonies, whereas, in N2B27 alone or in N2B27 supplemented with BMP alone, the formation of undifferentiated colonies was severely impaired (Figure 4E). This was also true for WT cells plated in the presence of LIF alone. In contrast, Oct4+/− cells formed alkaline phosphatase-positive undifferentiated colonies in LIF alone (Figure 4E).

17To determine whether pluripotency could be sustained during the clonal propagation of Oct4+/− cells in LIF alone, we obtained derivative cells expressing GFP constitutively. These cells were plated at clonal density, expanded, and replated at clonal density in defined medium in recombinant LIF without BMP. After two passages, parental cells completely differentiated and could not be sustained (Figure 4F). In contrast, Oct4+/− ESCs efficiently formed uniformly alkaline phosphatase-positive undifferentiated colonies (Figure 4F). Similar results were obtained when Oct4 was targeted in AGFP ESCs. When GFP+:Oct4+/− ESCs were injected into the blastocoel cavity of a host C57Bl/6 host embryo, midgestation embryos were obtained in which GFP+:Oct4+/− cells were widely distributed (Figure 4G) and detectable in the germline at E11.5 (Figure 4H). When similarly injected embryos were allowed to develop to term, adult coat color chimeras were obtained (Figure 4I). Altogether, these data indicate that ESCs with reduced Oct4 levels are released from BMP dependence.

18Oct4 binds to an octamer consensus site centered −176 bp upstream of the Nanog transcription initiation site that has been shown to positively regulate Nanog in transient reporter assays. However, given that no data exist regarding the action of Oct4 in the context of the endogenous Nanog locus, the parsimonious hypothesis that Oct4 suppresses Nanog via the −176 bp octamer site was tested. Gene targeting introduced a mutant version of the GFP-ires-pac-pA vector in which the octamer site in the Nanog promoter was destroyed (Figure 5A), thus producing mTg2a-Nanog:GFP ESCs (Figures S5A and S5B). In puromycin, cells carrying the Oct site mutation show a slight shift to lower GFP levels (Figure 5B). Removing puromycin from mTg2a-Nanog:GFP cultures resulted in the increased emergence of GFP-negative cells in comparison to Tg2a-Nanog:GFP cells (Figures 5B and 5C). These effects, mirrored at the mRNA level (Figure 5D), indicate that Oct4 regulates the endogenous Nanog locus via the −176 bp site positively, not negatively.

19Thus far, the results suggest that Oct4+/− cells exist in a state of enhanced stability that shares features of the naive ground state. To examine the potential connections between the Oct4+/− state and the ground state, we induced Oct4 in ZHTc-Nanog:GFP cells in the presence of inhibitors of ERK and GSK3. The induction of Nanog heterogeneity by Oct4 was blocked by treatment with inhibitors of FGFR, MEK, or GSK3 (Figure 5E). These data suggest that elevating Oct4 from heterozygous levels induces heterogeneous Nanog expression by modulating signaling responses.

20Fgf4 signaling is implicated in the downregulation of Nanog in ESCs and in the induction of heterogeneity in the ICM. Consistent with this, adding recombinant FGF to Tg2a-Nanog:GFP ESCs increased the proportion of Nanog:GFP-low cells in the culture (Figure 5F). Given that Fgf4 is regulated by Oct4, the lack of Nanog heterogeneity in Oct4+/− ESCs could be due to reduced Fgf4 or altered Fgf receptor expression by Oct4+/− ESCs. However, FGF4 transcripts were elevated in Oct4+/− ESCs with no difference detectable in Fgfr2 transcript level (Figure S5C). To further assess the dynamic effects of FGF, we targeted GFP-ires-pac-pA to Nanog in Fgf4−/− ESCs (Figure S5D). FGF-Nanog:GFP ESCs express Nanog:GFP unimodally (Figure S5E), Nanog:GFP-low cells being induced by FGF addition (Figure S5E). However, adding FGF to parallel cultures of either OKO-Nanog:GFP or ZHTc-Nanog:GFP ESCs did not affect Nanog:GFP, suggesting that the lack of heterogeneity in Oct4+/− ESCs is due to an inability of Oct4+/− ESCs to respond appropriately to FGF (Figure 5F). To determine whether this lack of response was due to an inability of Oct4+/− ESCs to elicit early events downstream of FGF receptor, we examined the kinetics of ERK phosphorylation. When ESCs were cultured in the absence of serum or cytokines and stimulated with FGF, both Oct4+/+ and Oct4+/− ESCs showed a comparable profile and duration of ERK phosphorylation (Figure 5G).

21In contrast to FGF stimulation, LIF stimulation increased phospho-ERK more strongly in Oct4+/− ESCs (Figure 5G). Phospho-STAT3 was detected heterogeneously in Oct4+/+ ESCs cultured in 100 U/ml LIF and predominantly in Nanog-expressing cells (Figure 5H). In contrast, almost all Oct4+/− cells express phospho-STAT3 (Figure 5H). Moreover, phospho-STAT3 became heterogeneous when ZHTc-Nanog:GFP cells were shifted from 1,000 ng/ml to 0.03 ng/ml doxycycline (Figure 5I). Furthermore, when parallel cultures of Oct4+/+ and Oct4+/− ESCs were starved and restimulated with LIF, phospho-STAT3 was induced to a far higher level in Oct4+/− ESCs (Figure 5J). These data suggest that the difference in transcription factor heterogeneity in Oct4+/− ESCs may be due to enhanced LIF sensitivity mediated by increased phospho-STAT3.

22Previously, Wnt signaling has been reported to act synergistically with LIF in order to support ESC self-renewal. Interestingly, Oct4+/− ESCs express increased Wnt3a mRNA (Figure 6A) with Oct4 induction downregulating Wnt3a (Figure 6A). Therefore, the potential of Wnt3a to regulate Nanog heterogeneity was assessed by FACS. Wnt3a blocked the emergence of a heterogeneous Nanog expression profile from Tg2a-Nanog:GFP cultures in a dose-dependent manner (Figure 6B). Wnt3a also prevented Oct4-induced Nanog heterogeneity in Oct4+/− ESCs (Figure 6C). Moreover, the ability of FGF to induce Nanog heterogeneity in Oct4+/+ ESCs was highly attenuated by Wnt3a (Figure 6D).

23To further investigate molecular distinctions between Oct4+/+ and Oct4+/− ESCs, we performed microarray analysis on two Oct4+/+ and two Oct4+/− ESC lines (Figures 6E and S6). This identified 163 genes that differed consistently between the two Oct4+/+ and two Oct4+/− ESC lines (|log2FC| ≥ log2(1.25), false discovery rate [FDR] ≤ 0.1; see Experimental Procedures). Functional enrichment analysis identified the biological processes “Wnt receptor signaling pathway” (FDR = 0.0609) and “embryonic pattern specification” (FDR = 0.0397) as the only gene ontology terms enriched in the candidate list. In addition to Wnt3a, Wnt6 was increased in Oct4+/− ESCs, whereas the Wnt antagonist Sfrp2 was decreased (Figure 6F). Furthermore, several pluripotency transcription factors were elevated in Oct4+/− ESCs (Figure 6G).

24Next, we investigated the genome-wide binding profiles of Oct4 and Nanog in both Oct4+/− and Oct4+/+ ESCs. For both Oct4 and Nanog, the majority of loci are shared between Oct4+/+ and Oct4+/− ESCs (Figures S7A and S7B). However, striking differences in the binding intensity exist between Oct4+/+ and Oct4+/− ESCs. At genes encoding pluripotency regulators (Esrrb, Tbx3, Oct4, Klf5, and Wnt6), specific loci with increased Oct4 and/or Nanog occupancy were identified (Figure 7A, 7B, 7C, S7C, S7D, and S7E). Such an increase was not found at other loci (e.g., Figure 7D). Next, we asked whether genes with increased Oct4 occupancy are associated with gene expression. Genes near Oct4-binding events that are enhanced in Oct4+/− ESCs are significantly enriched in the set of genes with higher expression in Oct4+/− ESCs (p = 1.7 × 10−37; Figure 7E), whereas Oct4-binding events reduced in Oct4+/− ESCs occur near genes with higher expression in Oct4+/+ ESCs (p = 1.2 × 10−8, Figure S7F). This change in gene expression is reflected in epigenetic differences. Genes where the promoter shows reduced levels of the repressive histone modification H3K27me3 in Oct4+/− ESCs in comparison to Oct4+/+ ESCs show higher expression in Oct4+/− ESCs (Figure 7F; p = 1.9 × 10−19). Altogether, this indicates that qualitative differences of Oct4 occupancy along with epigenetic differences are involved in regulating the robustness of the pluripotent state via effects on the pluripotency GRN.

25Surprisingly, even though Oct4 protein is reduced in Oct4+/− ESCs, the binding sites with the strongest difference between Oct4+/+ and Oct4+/− ESCs show increased occupancy in Oct4+/− ESCs. These Oct4+/−-specific Oct4 binding sites occur largely at genomic loci associated with p300 and H3K4me1, but not H3K4me3 (Figures 7G, 7H, and 7I). p300 and H3K4me1 are classical marks for enhancer regulatory sites, whereas H3K4me3 is highly abundant at proximal promoters. In addition, the sites that show enhanced Oct4 occupancy in the Oct4+/− ESCs are generally further from the transcription start sites (Figure S7G). Moreover, loci showing enhanced occupancy by Oct4 also exhibit enhanced co-occupancy by Nanog (Figure S7H).

26These data point toward an increase in Oct4 occupancy at enhancer regulatory regions in the Oct4+/− ESCs. The proximal and distal enhancers of Oct4 (Figure 7C) are of particular interest. The distal enhancer, which is associated with preimplantation embryos and ESCs, shows increased binding of Oct4 and Nanog in Oct4+/− ESCs. The proximal enhancer, which is associated with postimplantation epiblast and EpiSCs, shows decreased binding of Nanog. Thus, reduced Oct4 protein levels in Oct4+/− ESCs appear to induce a shift in Oct4 binding preferences toward a higher occupancy of key regulatory elements, thereby strengthening the regulatory network that maintains robust pluripotency.

27Although the heterogeneous expression of pluripotency transcription factors has the capacity to generate multiple outcomes from a single functionally defined cell type, the factors controlling heterogeneity are only beginning to be elucidated. Nanog autorepression has recently been demonstrated with this contributing to Nanog heterogeneity. In addition, Satb1, Satb2, and Ezh2 have been reported to influence Nanog heterogeneity. In this study, we identify an origin for transcription factor heterogeneity within the circuitry of the pluripotency GRN. Multiple Oct4+/− ESC lines have a unimodal high Nanog expression profile and lack Nanog-low undifferentiated cells. Heterogeneity is caused by a higher range of Oct4 levels present in WT ESC cultures because increasing the Oct4 concentration range toward the WT distribution switches Nanog to a bimodal heterogeneous profile.

28The fact that Oct4+/− cultures not only lack cells expressing the highest Oct4 concentrations present in Oct4+/+ cultures but are also essentially devoid of the differentiated Oct4-negative cells that are readily detectable in WT cultures is of no small interest here. These findings have two important implications. First, the full range of Oct4 concentrations present in the Oct4+/− population is within the distribution of expression levels observed both in WT ESC cultures and in the WT ICM, indicating that the Oct4 protein levels in Oct4+/− cells are physiologically relevant. Second, the appearance of differentiated Oct4-negative cells in ESC cultures is part of the normal network configuration assembled by WT Oct4 levels and does not require excessive Oct4 expression.

29The more homogeneous expression of transcription factors present in Oct4+/− ESCs results in delayed differentiation in comparison to WT ESCs. Importantly, restoring Nanog heterogeneity in Oct4+/− ESCs by transiently increasing the Oct4 level re-established normal differentiation. Furthermore, enforced expression of Nanog prevents differentiation induced by Oct4 elevation. Altogether these findings offer an explanation for the original observation that an increase in Oct4 can cause differentiation. Rather than being a consequence of Oct4 action at elevated concentrations, our results show that Oct4 acts at physiological levels to provide a population of Nanog-low cells responsive to differentiation cues.

30Our results establish unequivocally that a state of homogeneous transcription factor expression and robust self-renewal can exist without the need for pharmacological signaling pathway inhibition. The reasons for this are manifold. First, Oct4+/− ESCs exhibit an increased sensitivity to LIF. This can be seen by enhanced STAT3 phosphorylation, enhanced colony-forming capacity at low LIF concentrations, and enhanced maintenance of unimodally high Nanog:GFP expression at low LIF concentrations. This increased sensitivity to LIF is accompanied by a resistance to the induction of Nanog heterogeneity and differentiation in response to FGF. Interestingly, Oct4+/− ESCs respond to FGF by increasing ERK phosphorylation, suggesting that the lack of a full response to FGF in these cells is due to either the interception and elimination of effective signaling downstream of ERK or the dominance of an opposing signal. FGF4 mRNA expression is elevated in Oct4+/− cells, which may reflect dysregulation of the FGF signaling system in these cells. The combined inhibition of ERK and GSK3b has been spectacularly successful for the derivation of ESCs from recalcitrant strains of mice and from other species. Perhaps the inability to derive ESCs from the majority of non-129 mouse strains is due to the expression level of Oct4 being beyond the range that can be stabilized in culture by LIF.

31Oct4+/− cells also exhibit elevated expression of Wnt6 and Wnt3a and decreased expression of Sfrp2, which is indicative of a shift in the Wnt signaling system. Wnt3a has previously been reported to act in synergy with LIF on ESCs; although, without LIF, Wnt3a cannot maintain an undifferentiated ESC state. Indeed, in the presence of Wnt3a, LIF concentrations as low as 6 U/ml can support undifferentiated colony formation by WT ESCs. Here, we show that, at this LIF concentration, the majority of Oct4+/− ESCs colonies are completely undifferentiated. Supplementation of ESC cultures with Wnt3a has also been reported to block the transition from ESCs to EpiSCs but, again, only when LIF is also added. Not only do we observe a retarded transition to EpiSCs by Oct4+/− ESCs, but we also find that Oct4+/− ESCs can be clonally propagated in LIF-N2B27 in the absence of either BMP or signaling inhibitors.

32Although the differentiation of Oct4+/− ESCs to an EpiSC state is delayed, Oct4+/− EpiSCs can be established. An interesting property of Oct4+/− EpiSCs is the capacity to form undifferentiated ESC-like colonies when replated in LIF-fetal calf serum (FCS). It has been reported previously that EpiSCs can revert to an ESC-like state after culture on feeder cells. However, this reversion is slow, requiring >20 days. In contrast, Oct4+/− EpiSCs form ESC-like colonies in the absence of feeder cells within 6 days. This suggests that the distinction between the ESC and EpiSC pluripotent states is less clear-cut in cell populations expressing a reduced Oct4 concentration range. This may indicate that the higher Oct4 concentrations present in WT ESCs are important in imparting unidirectionality upon developmental progression.

33It is of interest that Oct4+/− ESCs show increased STAT3 phosphorylation in response to LIF. This may be related to the altered expression of Wnt signaling ligands in Oct4+/− ESCs, given that evidence has been presented that Wnt3a may increase STAT3 expression. Although Oct4 upregulation and a decrease in LIF concentration result in a similar differentiation phenotype, no connection has previously been made between Oct4 and STAT3 signaling. Here, we show that the degree of STAT3 phosphorylation varies with levels of Oct4 and surprisingly, in Oct4+/− ESCs, phosphorylated STAT3 can be detected in almost every cell. This contrasts with WT ESCs and suggests that insufficient or attenuated STAT3 signaling may contribute to ESC heterogeneity. These results suggest that Oct4+/− ESCs are primed to respond to LIF via enhanced STAT3 phosphorylation. The possibility that enhanced STAT3 signaling may have a dominant effect over signals antagonistic to ESC self-renewal such as ERK has been suggested. Therefore, our findings that Oct4+/− EpiSCs have a higher reprogramming efficiency than WT EpiSCs may, at least in part, be due to the enhanced phosphorylation of STAT3 in Oct4+/− cells.

34LIF has been reported to be upstream of transcription factor expression. Therefore, we propose that enhanced STAT3 phosphorylation leads to the more homogenous expression of pluripotency factors such as Nanog, Klf4, and Esrrb. Under such conditions, the profile of Oct4 binding shifts to sites more distal to transcription start sites. A prime example is the Oct4 gene. The distal Oct4 enhancer, which is more functional than the proximal enhancer in ESCs and the ICM than in EpiSCs and the postimplantation epiblast, is more fully occupied by Oct4 in Oct4+/− ESCs where the Oct4 concentration is reduced. These observations suggest that the robust network configuration of Oct4+/− ESCs reflects an early developmental state.

35Theoretical models have been proposed to explain heterogeneous Nanog expression. Both involve negative feedback on Nanog, and one model proposes Oct4 as a negative regulator of Nanog. Our work establishes the veracity of this conjecture but reveals that a state of nondynamic unimodal Nanog expression exists at diminished Oct4 levels.

36Our findings can be incorporated into a model in which Oct4 plays two diverse roles (Figure 7J). At the lower range of Oct4 concentrations found in WT ESCs and typified by Oct4+/− ESCs, Oct4 binds strongly to chromatin at key regulatory elements, directing the homogenous expression of pluripotency transcription factors and the expression of Wnt signaling components. This increased Wnt signaling activity enhances cell sensitivity to LIF, further re-enforcing the pluripotency GRN and making cells more refractory to differentiation signals such as FGF. This state can be maintained in Oct4+/− ESCs because the upward pressure of the re-enforced pluripotency transcription factors on the Oct4 loci cannot enforce Oct4 protein elevation because of the deletion of one of the Oct4 alleles. However, in Oct4+/+ cells, this movement into the higher Oct4 concentration range can occur. Then, the Wnt signaling system is no longer stimulated by Oct4, and the pluripotency transcription factors are expressed at a lower level. ESCs may exit this Oct4-high state by diminishing Oct4 expression as a consequence of the weakened transcription factor network strength, thereby reverting to an Oct4-low concentration range or by responding to a commitment signal (e.g., FGF) and differentiating.

37A priori, one might have expected the pluripotency GRN to have been optimally configured to support robust ESC self-renewal. In this respect, the demonstration that such robust ESC self-renewal is anchored through a network configuration orchestrated by reduced Oct4 concentrations is highly unexpected. This indicates that the control of pluripotent cell identity is naturally arranged in ESCs and ICM cells to provide heterogeneity. Consequently, stem cell self-renewal and rapid responsiveness to differentiation cues can simultaneously reside in a single stem cell population in which individual cells experience the same environmental stimuli. When one considers the situation in the embryo, it is perhaps less surprising that the network is not optimized for extended self-renewal, given that progression beyond the pluripotent state is required for orderly development. It will be of considerable interest to determine whether similar strategies are used in other stem cell systems to likewise facilitate rapid responses to changing physiological environments.

38ESCs were cultured in GMEMβ with 10% FCS and 100 U/ml LIF on gelatinized tissue culture flasks at a density of 2.5–10 × 104/cm2. Differentiation was induced by LIF withdrawal with a neural differentiation protocol or into EpiSCs, as described previously. Recombinant Wnt3a (R&D Systems) and recombinant Fgf2 (R&D) were used as indicated. JAK inhibitor I (Calbiochem) was used at 0.6 mM and PD0325901 (Axon Medchem) and Chiron 99021 (Axon) were used as described previously. Introduction of the Nanog:GFP reporter by homologous recombination was performed and verified as described previously.

39The colony-forming assay was performed by seeding 600 cells in a well of a six-well plate (60 cells/cm2). Cells were plated in GMEMβ-FCS with or without LIF or in N2B27 supplemented with recombinant LIF (Sigma-Aldrich) and/or recombinant BMP4 (R&D) as indicated and cultured in a 37°C, 7% CO2 incubator. After 7 days, cell staining was performed with an Alkaline Phosphatase Detection Kit (Sigma-Aldrich).

40For ESC quantifications, cells were segmented on the basis of DAPI with the nuclei detection algorithm of PerkinElmer’s Acapella Image Analysis Suite. For each nucleus, mean fluorescence for Oct4 and Nanog were computed.

41For embryo quantification, the fluorescence signal in each nucleus of whole-mount immunostained blastocysts were quantified with an automated pipeline as described previously, except that preprocessed DAPI-stained nuclei of RGB stack images were segmented with the FARSIGHT Toolkit (http://www.farsight-toolkit.org/) and incorrectly segmented nuclei were manually rectified with the FARSIGHT Nucleus Editor.