Epiblast Stem Cell-Based System Reveals Reprogramming Synergy of Germline Factors

1Specification of primordial germ cells (PGCs) in mice is accompanied by extensive epigenetic reprogramming, which is essential for generating the totipotent state. The key determinants of PGC specification, Blimp1/Prdm1 and Prdm14, induce repression of the somatic program and initiate epigenetic reprogramming in early germ cells, and they regulate this process together with their direct and indirect targets. Cell culture-based systems might be particularly useful for testing how the individual components contribute to complex reprogramming events in the germline, which in turn could improve our ability to control cell fates.

2PGC specification commences at embryonic day (E) 6.25 from postimplantation epiblast; these epiblast cells undergo major epigenetic changes after implantation, including DNA methylation and X inactivation. Epiblast stem cells (epiSCs) derived from E5.5–E6.5 epiblast inherit key properties from these cells and retain the potential to undergo reversion to embryonic stem cells (ESCs) or specification to unipotent PGCs. The alternative fates from epiSCs to ESCs or PGCs are quite distinct, but they share important common features, including reactivation of the inactive X chromosome, DNA demethylation, and re-expression of key pluripotency genes. Importantly, for Oct4 expression, there is a switch from the use of the proximal to the distal enhancer, the so-called enhanceosome locus of pluripotency. Thus, the key epigenetic modifications in postimplantation epiblast and epiSCs, which constitutes a robust epigenetic boundary, are reversed during reprogramming in both instances, although reversion of epiSCs to ESCs, or indeed of somatic cells to induced pluripotent stem cells (iPSCs), may also transit through a PGC-like state. EpiSCs can therefore be used to investigate aspects of epigenetic reprogramming and the roles of genes in early germ cells. The fact that epiSCs acquire additional DNA methylation during their derivation, which probably reduces their competence for PGC specification, is paradoxically an advantage for their use in such assays.

3EpiSCs self-renew in activin and basic fibroblast growth factor (bFGF), with a gene expression profile and epigenetic state that is distinct from mouse ESCs. EpiSCs can, however, revert to ESCs upon exposure to leukemia inhibitory factor (LIF)-Stat3 signaling on feeder cells, a process that is improved with the introduction of transcription factors, such as Klf4 or Nr5a2.

4Here we used epiSCs to explore the role of germline factors during reprogramming to ESCs. We found a potent combinatorial role for early germline factors, Prdm14-Klf2, that accelerate and enhance the ensuing process, including X reactivation and DNA demethylation, which are among the key reprogramming events that are also seen in early PGCs. Our approach might provide critical insight into the role of key germline factors, which in turn could be tested directly on germ cells while expanding our knowledge of complex reprogramming mechanisms in general.

5There are two major attributes of epiSCs that can be utilized to generate reporter lines for studying epigenetic reprogramming. First, female epiSCs exhibit an inactive X chromosome (Xi), which is unlike ESCs and PGCs. Second, the expression of Oct4 in epiSCs requires its proximal enhancer (PE), whereas it is the distal enhancer (DE) that drives Oct4 expression in both ESCs and PGCs. We therefore established two epiSC reporter lines to examine reprogramming by monitoring the status of X reactivation and by analyzing the activation of Oct4 DE in response to germline factors (see also later).

6To monitor the state of the X chromosome in epiSCs, we derived epiSCs from female E6.5 epiblast with a GFP reporter on the paternal X chromosome. The resulting XmXpGFP epiSC lines showed heterogeneous GFP expression resulting from random X chromosome inactivation in female postimplantation epiblast (Figures 1A and 1B). Next, we established a homogeneous population of GFP-negative epiSCs, termed XiGFP epiSCs (Figure 1B), in which the GFP transgene is located exclusively on the Xi, providing a basis to study the reactivation of the Xi by monitoring GFP expression.

7We observed a stable Xi when the cells were cultured in the absence of feeder cells on fibronectin in serum-free medium containing activin and bFGF, as shown by the fact that we did not detect GFP-positive cells (Figures 1B and 1C). Only when we cultured these XiGFP epiSCs on feeder cells in the presence of serum supplement did we notice, albeit very infrequently, a small number (around 0.05%) of GFP-positive cells (; data not shown), which may arise because of LIF or unknown factors secreted by feeder cells or present in serum supplement. We therefore cultured the cells without feeders on fibronectin and without serum supplement to ensure a repressed X-GFP reporter at the outset.

8XiGFP epiSCs showed high expression of postimplantation epiblast genes Fgf5 and Xist and low expression of ESC- and PGC-associated genes, including Klf2, Prdm14, Stella, Rex1, Nr0B1, and Tsix (Figure 1D; ). Whereas Oct4 transcript and protein levels were similar in XiGFP epiSCs and female ESCs, both Sox2 and Nanog were detected at reduced levels in XiGFP epiSCs (Figures 1D and 1E; ). Furthermore, Oct4-positive XiGFP epiSCs showed nuclear domains of Xist (92%, n = 100), as was the case in female mouse embryonic fibroblasts (MEFs) (97%, n = 100) (Figure 1F), consistent with the presence of an Xi. These XiGFP epiSCs showed accumulation of the repressive histone 3 lysine 27 trimethylation (H3K27me3) chromatin mark together with nuclear foci of Enhancer of Zeste (Ezh2) (Figure 1G; ). In addition, monoubiquitinated H2A (ubH2A) colocalized with H3K27me3 domains in the majority of epiSC nuclei (Figure 1G). Notably, we did not detect binding of Oct4, Sox2, or Nanog to intron 1 of Xist in XiGFP epiSCs, unlike female ESCs (Figure S1A available online; ), which indicates dissociation of key pluripotency factors from Xist intron 1 (and perhaps other loci), probably during the formation of postimplantation epiblast.

9Thus, the presence of multiple Xi markers demonstrates stable X inactivation in XiGFP epiSCs under our culture conditions.

10Next, we asked whether reprogramming and X chromosome reactivation in epiSCs could be promoted by factors that are upregulated in early germ cells, including Blimp1, Prdm14, Stella, Tcfap2c, Klf2, Klf5, Nanog, Sox2, and Dnd1, because these genes are induced just before the repression of Xist and X reactivation in PGCs.

11Initially, we found that transient transfection of germline gene combinations did not induce XiGFP reporter expression in epiSCs (data not shown). We therefore generated epiSCs with stable expression of combinations of two or three candidate factors by piggyBac (PB) transposition (Figure 2A; ), starting with Prdm14, Stella, Klf2, Nanog, and Sox2. EpiSCs with stable expression of these factors at similar or higher levels than the endogenous transcripts in female X-GFP ESCs (Figure 2B) had increased Stella, Nanog, and Sox2 protein compared to vector control epiSCs (Figure S1B). The levels of Prdm14 in manipulated epiSCs were higher than those in ESCs cultured in serum and LIF (Figure 2B), but similar to the levels in ESCs maintained in 2i and LIF, with the latter having about four times higher levels of Prdm14 transcript (Figure S1C). Transcript and protein levels of Sox2, a direct Prdm14 target, showed an increase in epiSCs with a gain of function for Prdm14 (Figures 2B and S1E), but there was no change in Nanog (Figure 2B). Despite these changes, the X-GFP reporter remained repressed, indicating stability of the Xi (Figure 2D, left panel).

12To exclude a possibility that X chromosome reactivation had been initiated but did not proceed to biallelic expression of X-linked genes, we analyzed the levels of Xist and Tsix in epiSCs sorted for SSEA1 to eliminate any differentiated cells. Quantitative real-time PCR (Q-PCR) for Xist and Tsix showed that they remained highly expressed and fully repressed, respectively (Figure S1D). Moreover, the epiSCs continued to display nuclear H3K27me3 domains (Figure S1E), which further confirmed that the Xi was not reactivated.

13Thus, our data show that X inactivation is remarkably stable in epiSCs expressing combinations of germline factors, which indicates that LIF-Stat3 signaling may be required to trigger reprogramming and X reactivation.

14Next, we examined the impact of LIF and serum (henceforth called LIF-Stat3) on XiGFP expression during reversion of epiSCs to ESCs on feeder cells (Figure 2C; ). By counting the number of GFP-positive colony patches every day, we found that epiSCs with exogenous Prdm14 and Stella or vector control did not show activation of the X-GFP reporter despite culture for 2 weeks in LIF-Stat3 (Figures 2D and 2E). In contrast, epiSCs expressing Klf2 and Nanog or Prdm14 and Nanog showed GFP expression after 7–8 days (Figure 2E). However, epiSCs expressing Prdm14 and Klf2 produced GFP-positive colonies after just 3–4 days, resulting in about 500 GFP-positive colonies from 50,000 plated cells on day 6 (Figures 2D and 2E). This effect is striking considering that X reactivation is a late event during reprogramming. The addition of Nanog had no additional impact (Figure 2E), although Prdm14 and Nanog have been proposed to cooperate in ESCs.

15Because Prdm14 activates Sox2, we asked whether epiSCs with exogenous Sox2 and Klf2 together could also accelerate activation of the X-GFP reporter, but this combination resulted in a slower response compared with Prdm14 and Klf2 (Figure 2E). This suggests that other critical targets are activated or repressed by Prdm14 and might have a role during reversion to ESCs.

16Initially, epiSCs with Prdm14 and Klf2 transferred to LIF-Stat3 showed mosaic GFP expression (Figure S1F), but the GFP signal frequently spread out to the whole colony (Figure 2F). These GFP-positive colonies were picked and expanded, resulting in lines with homogeneous GFP expression (Figure 2F), which indicates a reactivated X chromosome.

17Thus, our data show that Prdm14 and Klf2 trigger particularly rapid X reactivation in epiSCs upon transfer to LIF-Stat3, raising the possibility of cooperation between the two factors.

18To test whether Prdm14 and Klf2 cooperate to promote reprogramming, we examined XiGFP epiSCs expressing these genes individually or together (Figure 3A). Prior to transfer to LIF-Stat3, XiGFP epiSCs with different factor combinations retained a characteristic epiSC gene expression profile, with low expression of Stella, Rex1, Nr0B1, and Tsix and high expression of Xist and Fgf5 (Figure S2A). Importantly, epiSCs with Prdm14 alone did not show XiGFP reporter expression after 11 days in LIF-Stat3 (Figure 3B), and even after 21 days, only a slight effect was seen (Figure S2B). This indicates that Prdm14 alone is not sufficient to induce a major effect on reprogramming and X reactivation. By contrast, epiSCs with Klf2 alone produced GFP-positive cells but in a relatively protracted manner on day 6–7, and more robustly on day 9, compared to a highly accelerated rate of reprogramming by Prdm14 and Klf2 (Figure 3B). This demonstrates a combinatorial role for Prdm14 and Klf2 in X reactivation. The difference in the timing of X reactivation was not due to faster division of epiSCs carrying Prdm14 and Klf2 compared to cells with Klf2 alone, as monitored by growth curves (Figure S2C), although any putative contribution of cell division to X reactivation will require tracking the number of cell divisions in real-time.

19Therefore, although Prdm14 does not promote reprogramming on its own, it synergizes with Klf2 to trigger particularly rapid X reactivation.

20We noticed that XiGFP epiSCs that express Prdm14 and Klf2 formed compact, dome-shaped colonies faster than cells expressing Klf2 alone when cultured in LIF-Stat3 (Figure S2D). Indeed, Stella, which is expressed in ESCs and PGCs, was induced on day 3 in LIF-Stat3 when both Prdm14 and Klf2 were present, but not in cells with Klf2 alone or vector control (Figure 3C). In epiSCs with Prdm14-Klf2, X-GFP-positive cells appeared within Stella-expressing colonies (Figure 3C), suggesting that Stella expression may precede X reactivation, consistent with the temporal sequence observed in PGCs. After 4 days in LIF-Stat3, larger clusters of X-GFP-positive cells were located within Stella-positive colonies (Figure 3C). However, their expression did not always coincide (Figure 3C), which could be due to heterogeneity and dynamic changes in Stella expression in ESCs. Q-PCR analysis confirmed specific Stella induction in epiSCs with Prdm14 and Klf2 on day 4 in LIF-Stat3, in contrast to cells expressing Klf2 alone, Prdm14 alone, or vector control (Figure 3D). Bisulfite sequencing revealed partial DNA demethylation of the Stella locus in sorted GFP-positive cells from these cultures on day 4 in LIF-Stat3 (Figure S2E). Furthermore, Rex1, Nr0B1, and Nr5a2 were induced earlier, when both Prdm14 and Klf2 were expressed in epiSCs (Figure 3D). These data suggest that Prdm14 and Klf2 together accelerate epigenetic reprogramming in epiSCs.

21To gain further insight on the specificity of the response of epiSCs to Prdm14 and Klf2, we examined the response to other Klf family members, namely Klf4 and Klf5, that can promote reprogramming of fibroblasts to iPSCs; both are expressed in ESCs, although only Klf5 is detected in PGCs. We found that although Klf2, Klf4, or Klf5 alone could induce reprogramming of epiSCs, the efficiency of the process was higher when Prdm14 was also present (Figure S2F). However, Prdm14 was most potent when combined with Klf2, compared to the response with either Klf4 or Klf5 (Figure S2F) or with other pluripotency factors such as Nanog (Figure S2G). The specific effect of Prdm14 with Klf2 is unexpected, because Klf2, Klf4, and Klf5 have redundant functions in mouse ESCs. This suggests that re-entry into naive pluripotency may require different factors or factor combinations than maintenance of a naive pluripotent state.

22To further determine the rate and efficiency of reprogramming of epiSCs by Prdm14 and Klf2, we monitored it with the Oct4-ΔPE-GFP reporter, which contains Oct4 DE only. This reporter is repressed in epiSCs, but it is activated in the course of reversion to ESCs and during PGC specification. We introduced Prdm14 and Klf2 in Oct4-ΔPE-GFP epiSCs and found that this reporter was also repressed in cells cultured in activin and bFGF (Figures S3A and S3B). However, upon transfer to LIF-Stat3, we detected an extremely rapid and efficient activation of the reporter, with GFP-positive cells appearing as early as day 2 and with an efficiency of reporter activation of 5% after 6 days in LIF-Stat3 (Figures 3F and S3C). The speed of the response in the presence of Klf2 alone did not match with that observed with Prdm14-Klf2 (Figures 3E and 3F). These GFP-positive cells, called reverted ESCs (rESCs), could be maintained thereafter and exhibit homogeneous Oct4-ΔPE-GFP expression (Figure S3D). Thus, these data further suggest that Prdm14-Klf2 trigger particularly fast and efficient epigenetic reprogramming. Interestingly, Klf2 expression follows Prdm14 induction in early germ cells (Figure S3E; ), suggesting that the two factors may also act similarly to promote reprogramming of PGCs. Indeed, the combination of Prdm14-Klf2 was particularly potent for induction of PGC-like cells, which is not seen with wild-type epiSCs (Figures S3F and S3G; ) or when the two factors are tested individually. Additional factors or factor combinations may be identified in the future that may further enhance PGC specification.

23To exclude the possibility that the effect of Prdm14-Klf2 is mediated through factors present in serum or secreted by feeder cells, we examined reprogramming in serum- and feeder-free 2i/LIF conditions. Reprogramming by Prdm14 and Klf2 was again more efficient compared with Klf2 alone, resulting in more compact GFP-positive colonies on day 6–9 (Figures S3H–S3J). There was also specific induction of Stella, Rex1, and Nr5a2 as early as day 2 (Figure S3K). These observations suggest that Prdm14 and Klf2 enhance epigenetic reprogramming also upon Erk inhibition in serum- and feeder-free 2i/LIF conditions.

24To confirm complete reprogramming of rESCs, we tested their potential to contribute to somatic lineages. The cells contributed to chimeras after transgene excision (Figures 4A and 4B) and not before, presumably because the PB transgenes were not silenced. These Oct4-ΔPE-GFP-positive rESCs (Figure 4C) had a similar gene expression profile to ESCs (Figure 4G), integrated into the inner cell mass upon 8-cell injection (Figure 4D), and contributed to coat-color chimeras (Figure 4E) and to E13.5 genital ridges (Figure 4F). Interestingly, rESCs with Prdm14 and Klf2 transgenes had lower transcript levels of Dnmt3b, T, Lefty1, and Fgf5 than ESCs cultured in serum and LIF but similar levels as ESCs maintained in 2i and LIF (Figure 4G). Therefore, Prdm14 and Klf2 seem to generate rESCs that acquire a naive pluripotent state with lower levels of differentiation genes, which may have similarities with ESCs cultured in 2i. We confirmed the loss of nuclear foci of Xist, Ezh2, and H3K27me3 in rESCs (Figures S4A–S4C), indicating a fully reactivated X chromosome.

25Hence, our data suggest that these rESCs have acquired a naive pluripotent state that remains stable in the absence of ectopic transgene expression.

26To gain insight into the mechanism of acceleration of reprogramming by Prdm14 and Klf2, we analyzed the global gene expression changes, starting with epiSCs cultured in activin and bFGF, and after 2–4 days in LIF-Stat3 culture (Figures 5A and S5A). Unsupervised hierarchical clustering revealed two main clusters, corresponding to epiSCs in activin and bFGF, and after culture in LIF-Stat3, respectively (Figure 5A). Strikingly, epiSCs maintained in activin and bFGF with Prdm14 alone were similar to cells with both Prdm14 and Klf2, whereas epiSCs with Klf2 alone clustered together with vector control. The effect of Prdm14 is also reflected in the changes in epiSC morphology (Figure 5B).

27We analyzed the transcriptome of epiSCs for additional changes induced by Prdm14 and found that 1,433 genes were induced and 1,310 genes were repressed in epiSCs with Prdm14 compared to vector control (FDR < 0.005; Table S1). Reanalysis of Prdm14 ChIP-Seq data for ESCs revealed that 1,135/1,433 upregulated genes (p value 1.38 × 10−26) and 1,088/1,310 downregulated genes (p value 6.16 × 10−42) are targets of Prdm14 (Table S1), indicating a direct effect on the transcriptome and suggesting that Prdm14 can act both as an activator and a repressor. Gene ontology analysis showed that gastrulation, embryonic morphogenesis, and tissue morphogenesis genes were predominantly repressed, whereas cytoskeletal genes were induced (Figure 5C). We confirmed that genes associated with early lineage specification were repressed (Figure 5D) and many of them are Prdm14 targets, such as Nodal, Foxa2, Gata6, Hhex, Eomes, Foxh1, and Otx2. Whereas some genes were repressed in epiSCs with Klf2, most of them were repressed only when Prdm14 was present (Figure 5D). Thus, whereas Prdm14 represses extraembryonic endoderm differentiation in ESCs, it may also reduce differentiation toward early somatic lineages in epiSCs. Indeed, we found that Prdm14 overexpression inhibited differentiation of epiSCs into endoderm but not into neuroectoderm (Figure S6), which is consistent with the predominant repression of mesendodermal genes in activin and bFGF (Figure 5D).

28Among the repressed genes in epiSCs by Prdm14, we also found Thy1, which is repressed during the early phase of somatic cell reprogramming; Snai1, a Prdm14 target and an effector of epithelial-to-mesenchymal transition; Tcf3, a Prdm14 target and a negative regulator of pluripotency; and components of the Wnt pathway (Figure 5D).

29Notably, several early epiblast or ESC-associated genes, such as Sox2, Gbx2, Esrrb, Fbxo15, Gdf3, Dppa2, and Dppa4 as well as the FGF inhibitor Spry3 and several members of the dual specificity phosphatase (Dusp) family, were induced (Figure 5D); Dppa4 is particularly interesting, because its levels correlate with the efficiency of epiSC reversion. Altogether, these data suggest that Prdm14 enables repression of early germ layer-associated genes as well as induction of early epiblast markers in epiSCs.

30Next, we compared the reciprocal effects of gain of function of Prdm14 in epiSCs with Prdm14 knockdown in ESCs. We found only 37/1,443 induced genes, such as Sox2, repressed upon Prdm14 knockdown, and 94/1,310 repressed genes, such as Dnmt3b and Snai1, induced (Table S2). The limited overlap in gene expression changes may suggest cell context-dependent effects of Prdm14.

31Among the epigenetic modifier targets, we found that Prdm14 repressed Dnmt3b and Hdac6 in epiSCs (Figures 5D and 5E). Dnmt3b, which is predominantly responsible for DNA methylation in postimplantation epiblast cells, is also repressed in PGCs. However, we did not detect changes in GLP (Ehmt1) and Uhrf1 (Figure 5E) that are efficiently repressed in early PGCs; GLP was suggested to be responsible for the genome-wide decrease in H3K9me2. Consistently, there was no decrease in H3K9me2 levels in epiSCs with gain of Prdm14 (Figure S5B). Nonetheless, there was rapid DNA demethylation in epiSCs during reprogramming as judged by re-expression of Stella and Rex1 (Figures 3C and 3D), which are methylated in epiSCs.

32A particularly interesting observation was that many differentiation-associated genes, which were repressed in epiSCs with Prdm14, remained highly expressed or were even induced in epiSCs with Klf2 when cultured in LIF-Stat3 (Figure S5C). Thus, although Klf2 in epiSCs may induce differentiation upon transfer to LIF-Stat3, such tendency is blocked efficiently by Prdm14 and may contribute to fast and efficient reprogramming.

33Therefore, Prdm14 may enhance the competence for reprogramming by repression of differentiation and induction of early epiblast markers.

34The extensive gene expression changes induced by Prdm14 alone in epiSCs, although surprising, were not sufficient for successful reprogramming in LIF-Stat3, suggesting that a combinatorial role of Prdm14-Klf2 is critical, which is reflected in the gene expression changes (Figure 6A; Table S3). Gene ontology analysis revealed that among genes specifically induced with Prdm14-Klf2 were transcription factors, DNA binding, and negative regulation of differentiation genes (Figure S7A). Among the upregulated genes were reprogramming and ESC-associated genes, such as Nr5a2, Esrrb, Dppa2/3/4/5, Fbxo15, and Tcfap2a (Figure 6A). The downregulated genes included those involved in gastrulation, primary germ layer formation, and tissue morphogenesis (Figure S7A), such as the epiblast marker Fgf5 and the X-inactivation regulator Satb1 (Figure 6A; ).

35Reanalysis of ChIP-Seq data for Prdm14 and Stat3 and ChIP-on-chip for Klf2 showed that 578/611 Klf2-bound genes (95%) were also bound by Prdm14, whereas 238 of them (39%) are bound by both Prdm14 and Stat3 (Figure S7B and Table S4). Among Prdm14-Klf2-bound genes were the reprogramming factor Nr5a2 and Oct4 DE (Figure S7C). To test whether Prdm14 might help to recruit Klf2 to these targets, we performed ChIP for Klf2 on the Nr5a2 promoter and Oct4 DE. This revealed increased Klf2 recruitment in epiSCs in the presence of Prdm14 and Klf2, compared to cells with Klf2 alone on day 2 of culture in LIF-Stat3 (Figure 6B), indicating that Prdm14 may promote the binding of Klf2 to enhance reprogramming.

36We found that Prdm14 and Klf2 induced several germline-associated genes in epiSCs after transfer to LIF-Stat3, most notably Blimp1 (Prdm1), Stella (Dppa3), Fragilis (Ifitm1/3/5), and Nanos3 (Figure 6A), indicating that epiSC reprogramming by Prdm14-Klf2 may require progression through a germ cell intermediate. If so, this process might be restricted in the absence of Blimp1, a crucial determinant of PGC specification. We tested this hypothesis by using Blimp1-knockout epiSCs in our assay, but found that reprogramming was not affected as shown by the earlier appearance of Stella-expressing colonies, as well as specific induction of Rex1 and Nr5a2 and repression of Fgf5 in cells with Prdm14-Klf2 compared to those with Klf2 alone (Figures 7A–7C). In addition, rESCs reprogrammed by Prdm14 and Klf2 had retained DNA methylation in the imprinted loci Peg1, Peg3, and Snrpn (Figure 7D).

37Therefore, Prdm14 and Klf2 promote X reactivation and DNA demethylation (excepting loss of imprints), which also occurs in PGCs. However, Prdm14-Klf2 may act to induce characteristics of naive pluripotency without necessarily passing through a Blimp1-positive germ cell intermediate. This function may be conserved not only in the reversion of epiSCs to rESCs but also in re-establishing a primary background of pluripotency in PGCs, which is vital to this lineage (Figure 7E; ).

38Blimp1, Prdm14, and their targets drive PGC specification and reprogramming, which culminates in an epigenetic ground state. With an epiSC-based system to explore the role of germline factors in epigenetic reprogramming, we discovered a powerful synergistic effect of Prdm14-Klf2 that accelerates and enhances reprogramming of epiSCs to rESCs. Notably, Prdm14 alone has little effect on epiSC reprogramming, but it potentiates the action of Klf2; neither Klf4 nor Klf5 have an equivalent effect when combined with Prdm14.

39Whereas Prdm14 accelerates reprogramming of epiSCs by Klf2, it has surprisingly little effect by itself, which suggests that Prdm14 might prime epiSCs for reprogramming. Indeed, we show global changes in the transcriptome of epiSCs in response to Prdm14 alone, which include induction of Esrrb and repression of Tcf3 and Nodal. Prdm14 apparently induces changes in epiSCs toward an earlier postimplantation epiblast-like state, with the induction of Dppa4, which might contribute to the effectiveness of Klf2 in the process. In addition, by repressing early lineage-specification genes, Prdm14 may decrease the propensity of epiSCs toward heterogeneity and increase their responsiveness to LIF-Stat3 signaling and reprogramming factors. Prdm14, which binds predominantly to distal regulatory elements, might also be important for priming by inducing a permissive state in enhancers in preparation for interaction with the cognate promoters.

40Prdm14 enhances recruitment of Klf2 to key loci, such as Nr5a2 and the distal Oct4 enhancer, suggesting cooperation between the two factors. The binding sites of Prdm14 and Klf2 are in close proximity on these targets, suggesting that Klf2 could be recruited directly by Prdm14. However, because Prdm14 predominantly binds far away from the transcription start site, it might bring distal regulatory elements in close proximity to promoters; the resulting changes in the chromatin topography and potential chromatin-modifying activities of Prdm14, potentially through its PR/SET domain, may promote its combinatorial role with Klf2. Because the current analysis of Klf2 binding sites was performed with a tiling array interrogating only about 400 genes, the full extent of the overlap of binding between Prdm14 and Klf2 remains unknown and the combinatorial effect could involve cobinding on genes outside the scope of that assay. Therefore, genome-wide binding profiles of Klf2 in the absence and presence of Prdm14 during epiSC reversion are needed to identify Klf2 targets that are Prdm14 dependent and independent. Notably, reprogramming occurs only after transfer of epiSCs to LIF-Stat3, indicating a functional interaction of Stat3 with Prdm14 and/or Klf2.

41The epiSC-based assay we describe exhibits a rapid effect on X reactivation as seen in early germ cells. Prdm14-Klf2 can potentially act to regulate the expression of Xist and Tsix; Prdm14 itself binds directly to Xist intron 1 and to Rnf12, an X-linked Xist activator, whereas Klf2 might promote Tsix expression resulting in repression of Xist. Although Oct4, Sox2, and Nanog are not enriched on intron 1 of Xist in epiSCs, Prdm14 could promote recruitment of the pluripotency factors during reversion of epiSCs to rESCs, e.g., by altering the local status of the chromatin.

42The derivation of epiSCs from epiblast is coupled with acquisition of DNA methylation on loci such as Stella and Rex1, which could contribute to the refractory nature of epiSC reprogramming. DNA demethylation by Prdm14-Klf2 occurs rapidly after epiSCs are transferred to LIF-Stat3 as judged by the activation of Stella and Rex1. The rapid loss of DNA methylation cannot be easily accounted for by a passive mechanism, because expression of the maintenance methylase Dnmt1 or its vital cofactor Uhrf1 were unaffected; Uhrf1 is repressed in PGCs. Thus, rapid DNA demethylation might occur through alternative mechanisms, potentially involving hydroxylation of 5-methylcytosine by enzymes, such as Tet1 and Tet2, which are detected at the time of reprogramming in PGCs, and in ESCs. A recent study indicates that Stella and Rex1 are targets of Tet1/Tet2. Further studies are needed to determine the precise mechanism of DNA demethylation.

43Besides the role of Prdm14 in PGCs, PRDM14 is obligatory for the maintenance of pluripotency in human ESCs, whereas Prdm14 is not expressed in epiSCs. Knockdown of PRDM14 induces human ESC differentiation, as also reported for mouse ESCs. Furthermore, PRDM14 binds to the proximal Oct4 enhancer in human ESCs, whereas it is the distal enhancer that is bound by Prdm14 and gets activated during the reversion of epiSCs to rESCs. Notably, naive human ESCs, with two active X chromosomes, have recently been generated by introduction of KLF2 and KLF4 in 2i/LIF conditions, and functional cooperation between PRDM14 and KLF2 should be investigated in this context.

44Prdm14-Klf2-induced epigenetic changes in epiSCs during reversion to rESCs show some features in common with germ cell reprogramming, including X reactivation and DNA demethylation. The different cellular context, together with Blimp1 in PGCs, may be crucial for germ cell-specific features of reprogramming, especially the erasure of imprints. Indeed, we show that Prdm14-Klf2-induced reversion occurs efficiently in Blimp1-knockout epiSCs. In addition, Blimp1 is required neither for derivation and maintenance of ESCs nor for reversion of epiSCs to rESCs, but it is required for PGC specification (; S.B., H.G. Leitch, M.A.S., et al., unpublished).

45EpiSCs show a significant loss of competence for PGC specification compared with postimplantation epiblast or ESCs and a poor ability to undergo reversion to rESCs. Studies similar to the effects of Prdm14-Klf2 that promote both induction of PGC-like cells from epiSCs and reversion to rESCs could also advance our knowledge of the molecular basis of competence.

46Cell culture-based systems that recapitulate events in early germ cells might be mutually informative for epigenetic reprogramming in general, while increasing our knowledge of the complex events in early germ cells that are vital for totipotency and a prerequisite for pluripotency in vivo.

47Animal studies were authorized by a UK Home Office Project License and carried out in a Home Office-designated facility.

48Pooled stable transfectants were plated at a density of 10,000/30,000/50,000 cells per well of a 6-well tissue culture plate in standard ESC medium supplemented with LIF (1000 U/ml; ESGRO; Chemicon) and 20% fetal bovine serum (GIBCO) on mitomycin C-treated MEFs. The medium was first replaced after 48 hr and subsequently every 24 hr. Unless otherwise indicated, the number of GFP-positive colony patches per well was counted every day with an Olympus IX71 inverted microscope and results are shown as mean ± standard deviation (SD) of three independent experiments.