Mapping the networks for pluripotency

Abstract

There has been an immense interest in embryonic stem cells owing to their pluripotent property, which refers to the ability to differentiate into all cell types of an embryo. In the maintenance of this pluripotent nature, transcription factors play essential roles, and signalling pathways also act to sustain the undifferentiated state. Recent studies have unravelled multiple forms of interconnection and crosstalk between these two regulatory aspects of pluripotency. With the discovery of epiblast stem cells, there is an emerging concept that different pluripotent states could exist, and knowledge of both transcriptional networks and signalling pathways has been vital in dissecting the properties of these different states. Similar to classical reprogramming methodologies, various combinations of transcription factor transduction and the modulation of intracellular signalling have enabled the interconversion between pluripotent states. These studies provide an insight into the defining characteristics as well as the plasticity of pluripotent cells.

1. Introduction

Pluripotency refers to the ability to differentiate into all cell types of the embryo when the appropriate stimulus is provided [1]. One of the earliest and most well-established stem cell models for pluripotency is mouse embryonic stem cells (mESCs), which were derived from the inner cell mass (ICM) of the mouse blastocyst nearly three decades ago [2,3]. The establishment of well-defined culture conditions allows mESCs to self-renew infinitely while maintaining a pluripotent state in vitro, providing a useful source of cells for molecular studies and differentiation into a variety of desired cell types. Importantly, mESCs satisfy all the criteria that define pluripotency, including the tetraploid complementation assay which tests for the ability of cells to give rise to an entire embryo when injected in 4n blastocysts [4,5].

Early studies have led to the discovery of the core pluripotency factors, namely, Oct4, Sox2 and Nanog in mouse embryos and mESCs [6]. Since then, much research effort has been invested to dissect the molecular functions of these core pluripotency factors in the maintenance of pluripotency. Recent years have seen many studies attempting to understand pluripotency in a genome-wide manner as well as on a systems level so as to provide a global understanding of the embryonic stem cell (ESC) state and differentiation. The signalling networks have also been elucidated. From initial studies that identified leukaemia inhibitory factor (LIF) and bone morphogenetic protein 4 (BMP4) as essential growth factors for maintenance of the mESC state [7], there has been progress in dissecting culture-mediated signalling pathways and in understanding how signalling can be integrated to the transcriptional network. Modulation of signalling has subsequently shown that extrinsic growth factors could be dispensable for sustaining the pluripotent state [8].

Although the mouse provides a model from which potential parallels can be drawn from humans, it was two decades after the discovery of mESCs that human ESCs (hESCs) were established in culture [9]. Although hESCs are also derived from the ICM, there are many striking differences from mESCs such as morphology, marker expression, transcription factor binding activities and culture requirements. Intriguingly, it was later found that hESCs resemble more closely mouse epiblast stem cells (EpiSCs), which were derived from post-implantation late epiblasts of the mouse [10,11]. Morphologically distinct from the compact dome-shaped mESC colonies, both hESCs and EpiSCs grow as a flat monolayer, and are sensitive to single-cell passaging. Therefore, hESCs, EpiSCs and mESCs manifest themselves as potentially distinct cell types (table 1).

Table 1.

Comparison of the different types of pluripotent cells.

type	source	morphology	culture conditions	pluripotency status	references
mESC	ICM	dome-shaped	LIF/BMP4	tetraploid complementation	[5,7]
			2i	contribution to germline	[8]
EpiSC	late epiblast	flattened monolayer	Activin A/bFGF	contribution to teratomas	[10,11]
hESC	ICM	flattened monolayer	Activin A/bFGF	contribution to teratomas	[9,12]

Apart from morphological and molecular differences, EpiSCs differ from mESCs on the level of pluripotency that can be attained by various standard assays [4]. EpiSCs can form teratomas in vivo, but they neither form chimeras nor contribute to the germline like mESCs [10,11]. Recent studies indicate that transcription factor transduction and manipulation of culture conditions can provide a powerful method to interconvert between different pluripotency states, such as to elevate the developmental capacity of EpiSCs [13–15]. Exploring this conversion process allows critical insights to be gained on the defining properties of these pluripotent states.

Here, we discuss the transcriptional network and signalling requirements of ESCs, and how these features vary in different pluripotent cell types. We also provide an overview of the numerous interconversion experiments and discuss the insights gained from these studies. We conclude by highlighting the ensuing questions and potential applications.

2. Dissecting the transcriptional regulatory network in embryonic stem cells

(a). Core pluripotency factors: Oct4, Sox2 and Nanog

The core transcription factors, Oct4 (encoded by Pou5f1), Sox2 and Nanog, are central to the regulation of pluripotency. Their importance in early embryo development is highlighted by several studies. Disruption of Pou5f1 caused aberrant differentiation of the ICM to the trophectoderm lineage instead of the embryo [16], Nanog null ICM failed to develop into the epiblast [17], while Sox2 knockout in mice conferred early embryonic lethality [18].

Given the importance of the core transcription factors in maintenance as well as induction of pluripotency, it is imperative to identify their downstream target genes. Genome-wide chromatin immunoprecipitation (ChIP) studies revealed that Oct4, Sox2 and Nanog co-bind to the promoters of many genes in both mESCs and hESCs [19,20]. Notable categories of co-bound target genes include signalling intermediaries, microRNAs, chromatin-remodelling and histone modifying proteins, hence establishing a direct link between the core transcriptional network and other key aspects of pluripotency regulation. These studies also indicate that the core transcription factors maintain ESCs by activating other pluripotency factors, such as Esrrb and Zic3, as well as repressing developmental genes such as Cdx2 to suppress differentiation. Oct4, Sox2 and Nanog also activate themselves and each other, constituting a core transcriptional regulatory network with features of auto-activation and feed-forward loops [19–24].

(b). Expansion of the pluripotency network

To expand the pluripotency transcriptional network beyond the core transcription factors, one approach is to search for their binding partners by affinity-based purification and mass spectrometry. Using Nanog as a starting bait followed by iterative identification of interaction partners, an extensive protein interaction network was established for mESCs [25]. The protein interaction network consists of many proteins that are individually important for the maintenance of pluripotency, such as Oct4, Sall4 and Dax1. Other than direct protein–protein interaction, members of the network are also co-regulated at the transcriptional level, suggesting that these pluripotency factors are intricately connected on multiple levels.

Using a similar affinity purification strategy with improved tagging methodologies, two groups independently established an extensive Oct4-centred interactome [26,27]. Among the identified Oct4 interaction partners, a significant proportion are transcription factors with known functions in pluripotency such as Sox2, Nanog, Sall4, Klf4, Klf5, Zfp143, Dax1, Esrrb and Tcfcp2l1, many of which are also transcriptionally regulated by Oct4 itself. Another striking commonality in the Oct4 interactome and the Nanog interaction network is the prominence of chromatin-modifying complexes, specifically the SWI/SNF and NuRD complexes [26–28]. The SWI/SNF complex is essential for maintenance of pluripotency [29,30], while the NuRD repressor complex co-localizes with Oct4 and Nanog to suppress differentiation genes [28]. Notably, the NuRD and SWI/SNF complexes associate with Oct4 and some of its interaction partners [26–28], suggesting that these complexes are frequently recruited and possibly occupy a central role in the pluripotency network.

Another approach to identify novel pluripotency-associated genes is to perform large-scale RNAi screens in mESCs. In addition to Pou5f1, Sox2 and Nanog, knockdown of other genes such as Esrrb, Tbx3 and Tcl1 led to a loss of self-renewal in mESCs [31]. Two other independent groups performed genome-wide knockdown in mESCs and used a Pou5f1-GFP reporter assay to indicate a loss of ESC identity [32,33]. One of the novel hits identified from the RNAi screen is the chromatin regulator Paf1C, which was found to functionally overlap with the Oct4–Nanog circuitry [33]. In addition, the other RNAi screen also led to the identification of other novel pluripotency factors such as Cnot3 and Trim28a [32]. Through the binding at common genomic sites, c-Myc, Zfx, Cnot3 and Trim28 form a self-renewal module and the genes associated with this module are mainly involved in cell cycle regulation and cancer [32].

The existence of more than one transcriptional module in ESCs has previously been uncovered. By using the ChIP-Seq technology, the genome-wide binding sites of 13 transcription factors and two co-regulators were mapped in mESCs [34]. Based on co-localization analyses, these pluripotency transcription factors can be clustered into two groups. The first cluster consists of the core pluripotency factors Oct4, Sox2 and Nanog as well as signalling effectors Smad1 and STAT3, while the second cluster includes c-Myc, n-Myc, E2F1 and Zfx [34]. The first cluster was found to co-localize extensively with the enhancer-associated transcriptional co-activator p300 at non-promoter regions, bringing forth the idea of an ESC-specific enhanceosome. In another independent study, a biotinylation ChIP-chip strategy was used to study the promoter binding profiles of nine transcription factors: Oct4, Sox2, Klf4, c-Myc, Nanog, Dax1, Rex1, Nac1 and Zfp281 [35]. Similar to Chen et al. [34], Kim et al. also detected high-density binding at many targets, supporting interconnectivity among the studied factors. Interestingly, the latter study discovered a correlation between binding density of pluripotency factors and expression levels, whereby active genes are frequently bound by multiple factors while inactive genes tend to be bound by fewer factors [35].

While the transcriptional network has been considerably explored in mESCs, relatively little is known about key pluripotency players in hESCs. Using a POU5F1-GFP construct as reporter for an undifferentiated state, a genome-wide RNAi screen was performed to uncover novel regulators of hESC pluripotency [36]. Several known regulators of ESC pluripotency such as OCT4, ZIC3 and NANOG emerged as top hits. The identified candidates were enriched for transcription and translation factors as well as components of biochemical complexes that have not been previously implicated in hESC maintenance, including the INO80 chromatin remodelling complex as well as the mediator and TAF transcriptional regulatory complexes [26,36]. In addition, PRDM14 has been highlighted as a novel pluripotency regulator in hESCs. PRDM14 directly regulates the expression of POU5F1 and co-localizes extensively with OCT4, SOX2 and NANOG, showing clear integration with the core transcriptional network. PRDM14 also enhances the reprogramming efficiency of human fibroblasts and can replace KLF4 when used in conjunction with OCT4 and SOX2. Therefore, genome-wide RNAi screens can advance the establishment of pluripotency networks and provide novel pluripotency factors that may be useful for reprogramming.

The availability of genome-wide tools has allowed the rapid expansion of identified pluripotency factors and enabled an improved understanding of the transcriptional network underlying pluripotency. The core pluripotency factors have been used as baits for the identification of novel protein interaction partners, which have highly co-regulated expression and exhibit extensive co-localization on target genes. In addition, genome-wide RNAi studies have led to discovery of novel genes and complexes that are shown to participate in the maintenance of an undifferentiated ESC state. Collectively, these large-scale studies demonstrate that the pluripotency transcriptional network is highly interconnected on multiple levels and extends to other aspects of ESC regulation such as epigenetics and signalling.

(c). Systematic approaches to understanding pluripotency

In order to gain a global understanding of how various levels of regulation can be integrated in ESCs, Lemischka and colleagues measured genome-wide changes that occur upon specific RNAi knockdown of Nanog [37]. This systems-level study provides a novel approach to reveal dynamic changes on multiple regulatory levels, such as transcriptional, epigenetics and proteomics, which occur during ESC differentiation. When Nanog was downregulated, there was extensive perturbation of promoter epigenetic status, RNA polymerase II binding, gene expression profile and nuclear protein abundance [37]. Further analysis of the different categories of discordances that can occur among these regulatory levels revealed that a large proportion of the changes in nuclear protein levels did not correlate with the mRNA levels, suggesting that translational and post-translational regulation could have prominent roles in ESCs. In contrast, changes in the expression levels of chromatin modifiers were shown to be mainly regulated at the transcriptional level, suggesting that transcription factors could have a primary role in directing epigenetic changes. However, it is unclear if these findings are specific to ESCs or are broadly applicable to all cell types.

Another study examined how the pluripotency network can be offset by the systematic overexpression of 50 different transcription factors [38]. Most of the transcription factors that result in substantial transcriptional changes upon overexpression are generally related to differentiation and not present in ESCs, so it is intriguing that the overexpression of key pluripotency transcription factors Klf4 and Sox2 can also cause substantial perturbation to the transcriptome. Given that pluripotency factors form highly interconnected protein complexes and the extent of binding affects transcriptional activity [34,35], it was suggested that these transcription factors may be dose-sensitive and distort the stoichiometric balance in the pluripotency network when overexpressed. Among the overexpressed transcription factors, Cdx2 induction led to the most widespread impact on the transcriptome. Further investigations revealed that Cdx2 interacts with the NuRD repressor complex and may interfere with the binding of pluripotency transcription factors to target genes [38], therefore providing a possible mechanism for how differentiation transcription factors mediate their downstream effects.

(d). Differences among mouse embryonic stem cells, human embryonic stem cells and epiblast stem cells

The core pluripotency factors, Oct4, Sox2 and Nanog, are recognized as key nodes in the transcriptional networks of both mESCs and hESCs. However, a comparison of Oct4 and Nanog target genes revealed a limited overlap between mESCs and hESCs [19,20]. These differences in transcription factor binding activities were partly attributed to species-specific variation between the mouse and human, a postulation that was later supported by the finding that Oct4 and Nanog bound targets are significantly enriched for species-specific transposable elements that may have diversified the binding site repertoire between the mouse and human cells [39]. On the other hand, these Oct4 target genes appeared to be more similar between EpiSCs and hESCs as compared with mESCs [11]. The resemblance of hESCs to EpiSCs, which are derived from post-implantation epiblasts of mouse origin, suggests that hESCs may be derived from a developmental stage more similar to EpiSCs than mESCs.

Apart from differences in Oct4 and Nanog binding profiles, EpiSCs and mESCs can also be distinguished by a distinct marker expression [10,11]. EpiSCs have significantly reduced expression of various mESC-associated genes like Rex1, Stella, Tbx3 as well as Klf4, and elevated expression of late epiblast genes, including Fgf5 and Brachyury. Intriguingly, hESCs express some mESC-associated markers such as KLF4 and REX1 [40] but not other mESC markers such as SSEA1 [9]. The transcriptional disparities among mESCs, EpiSCs and hESCs cannot be fully explained by species or developmental stages alone, and suggest that these three pluripotent cell types represent unique pluripotent states. Consistent with this hypothesis is a recent finding that PRDM14 is specifically required to maintain hESCs, but not mESCs and EpiSCs [36].

3. Signalling pathways and the ground state hypothesis

(a). Culture-mediated signalling in mouse embryonic stem cells

ESC maintenance has classically been known to be dependent on external stimuli from culture conditions. Although mESCs were initially cultured on mouse embryonic fibroblasts (MEFs) with foetal calf serum (FCS), it was discovered that FCS can be effectively replaced by BMP4 [7], and the cytokine LIF is sufficient for feeder-free culture [41]. BMP4 and LIF have subsequently been shown to act in various ways to maintain the undifferentiated state of mESCs.

BMP4 induces transcription factor Smad1 and inhibitor of differentiation (Id) proteins, which inhibit neural-specific differentiation [7]. LIF, on the other hand, mediates its action by multiple signalling pathways for cellular proliferation, including the LIF-STAT3, mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3-kinase (PI-3K) pathways. Direct inhibition of the MAPK pathway using MEK inhibitors prevents differentiation [42]. On the contrary, inhibition of the PI-3K pathway induces differentiation, which can be overcome by using MAPK kinase (MEK) inhibitors, demonstrating that PI-3K maintains the undifferentiated state via suppression of the MAPK pathway [43]. In the absence of LIF, BMP4 redirects cells to non-neural differentiation, and the basis of this differentiation propensity is due to activation of the fibroblast growth factor (FGF)–extracellular signal-regulated kinase (ERK) MAPK pathway [44]. Hence, both culture factors BMP4 and LIF converge upon the MAPK pathway as a differentiation cue.

Among the multiple pathways downstream of LIF, the most extensively investigated pathway has been LIF–STAT3. Expression of a dominant-negative form of STAT3 led to differentiation [41], while activation of the STAT3-ER construct is sufficient for maintenance of the undifferentiated state in the absence of LIF, when cells are maintained at high density [45]. In contrast to MAPK, STAT3 positively acts to maintain mESC pluripotency. STAT3 target genes were found to contribute to the maintenance of pluripotency, especially in the suppression of mesoderm and endoderm differentiation [46].

(b). Crosstalk with the transcriptional network in mouse embryonic stem cells

Culture-mediated signalling is intricately tied up with the transcription factor network. An obvious connection is the effective substitution of culture supplements by overexpression of transcription factors. Nanog overexpression maintains substantial cellular Id levels in the absence of LIF or BMP [7], and overcomes the requirement of these growth factors for self-renewal [7,47]. Klf2, on the other hand, can maintain the undifferentiated state of mESCs and prevents BMP4-induced differentiation in the absence of LIF [48]. Furthermore, inducible expression of STAT3 upregulates Klf4 [48], while addition of MAPK inhibitors also upregulates Tbx3 and Nanog [49]. Klf4 and Tbx3, the two upregulated pluripotent transcription factors, have been shown to be sufficient for LIF-independent mESC proliferation. Furthermore, Tbx3 overexpression maintains expression of Nanog [49], providing a clear demonstration of how extracellular stimuli could induce expression of the core pluripotency transcription factors. In addition, the transcription factors STAT3 and Smad1, which are downstream effectors of the LIF and BMP4 signalling pathways, respectively, co-localize at common target genes with the core transcription factors Oct4, Sox2 and Nanog [34]. Therefore, the two key signalling pathways in mESCs can be functionally integrated into the transcriptional network of mESCs.

(c). Differences among mouse embryonic stem cells, human embryonic stem cells and epiblast stem cells

Signalling requirements vary in different pluripotent cell types. Both hESCs and EpiSCs maintain the undifferentiated state through the Activin/nodal signalling pathway, instead of the LIF–STAT3 pathway [10–12]. Smad2/3, downstream effectors of the Activin/nodal signalling pathway, directly bind and induce Nanog expression in both hESCs and EpiSCs [50,51], thereby promoting self-renewal. In contrast to mESCs, it was shown using FGF/ERK inhibitors that hESCs and EpiSCs depend on the FGF–ERK pathway to inhibit differentiation [50]. FGF signalling also upregulates integrin instead of E-cadherin signalling in hESCs [52], and disrupted integrin signalling could account for the sensitivity of hESCs and potentially EpiSCs to single-cell passaging [53]. However, the signalling architecture of hESCs and EpiSCs are not totally the same, as FGF/ERK inhibition causes a drop in Nanog expression in hESCs but does not affect Nanog expression in EpiSCs [50]. Therefore, different signalling systems exist in mESCs, hESCs and EpiSCs.

(d). The ground state hypothesis

Maintenance of an undifferentiated state in mESCs is facilitated by appropriate modulation of intracellular signalling which is stimulated by extrinsic culture conditions. The convergence of both LIF and BMP on blocking the pro-differentiation MAPK pathway suggests that another alternative to maintain mESCs could be to directly inhibit the MAPK pathway. In support of this hypothesis, Smith and co-workers [8] reported that the inhibition of the MEK/ERK pathway can suppress mESC differentiation, while the addition of an inhibitor for glycogen synthase kinase 3 (GSK3) allows cells to regain wild-type proliferation rates. The maintenance of an undifferentiated state with independence from activation of ERK MAPK signalling is defined as the ground state of ESC self-renewal [54]. In contrary, cells that are poised for differentiation and rapidly differentiate when MAPK inhibitors are used, for example hESCs and EpiSCs, are considered to be in a primed state [54], thereby clearly establishing two distinct states of pluripotency.

The combination of both the MEK/ERK and GSK3 inhibitors, also known as 2i conditions, has dispensed with the requirement for LIF and BMP4 in mESC culture. 2i conditions have also enabled the derivation of mESCs from the non-permissive nonobese diabetic (NOD) mouse strain [55], as well as the derivation of germline-competent rat ESCs in combination with LIF [56,57]. Knockout rats have been generated for p53, a key tumour suppressor in cancer [58], providing a promising start to the development of rat models for disease. These studies suggest a potential universality in application of the 2i culture conditions and the pluripotent ground state.

From an understanding of the key signalling pathways mediating maintenance of mESCs, including notably the MAPK and the LIF–STAT3 pathways, there has been progress in efforts to modulate these signalling pathways. Utilization of the MEK and GSK inhibitors in 2i conditions has created a powerful culture system that allows derivation of pluripotent cells from refractory mouse strains as well as from different species. Equally importantly, 2i conditions have been one of the foundations for recent interest in enhancing the developmental capacity of other pluripotent cell types such as EpiSCs.

4. Interconversion between pluripotent states

(a). Interconversion between epiblast stem cells and mouse embryonic stem cells

It has been suggested that EpiSCs and mESCs represent two different states of pluripotency as the former does not satisfy the more stringent criteria of pluripotency [10,11]. Although mESCs represent the most established stem cell model for pluripotency, EpiSCs may provide a more relevant model for human studies given their similarities to hESCs. It is informative to investigate the different properties of these two cell types, and how they can be modulated to resemble each other (figure 1a).

Figure 1.

Interconversion of EpiSCs and hESCs to an mESC-like state. (a) mESCs convert into EpiSC-like cells when cultured in EpiSC media containing bFGF and Activin (curved light grey arrow). EpiSCs can be converted into an mESC-like state via overexpression of Klf4 or Nanog in conjunction with application of growth factors and/or 2i chemical inhibitors (straight dark grey arrow). Alternatively, some studies have also shown that purely chemical means is sufficient (curved dark grey arrow). (b) hESCs can be converted into an alternative pluripotent state in a similar manner to EpiSC (straight dark grey arrow). However, the process requires constitutive transgene expression, although Forskolin can prolong the maintenance of transiently transfected cells. hESCs can also be converted into trypsin-tolerant mESC-like cells using a chemical-based inhibitor cocktail (curved dark grey arrow).

(i). Mouse embryonic stem cell conversion to epiblast stem cell

As mESCs are derived from an earlier stage in the developmental process, conversion to an EpiSC-like state should be comparable to progress down the developmental programme. Indeed, mESCs convert to EpiSC-like cells within several passages when cultured in Activin A/bFGF conditions [14]. These EpiSC-like cells downregulated several mESC markers such as Nanog, Rex1, Nr0b1 and Klf4, and underwent X chromosome inactivation. Hence, the switch from mESC to EpiSC culture conditions is sufficient to result in molecular changes towards an EpiSC fate. In an independent study, the conversion can be achieved within 2 days by either FGF addition or LIF/STAT3 pathway inhibition, and further accelerated by a combination of both treatments [50].

(ii). Epiblast stem cell conversion to mouse embryonic stem cell-like state using transcription factors

While mESCs can spontaneously differentiate to an EpiSC-like state, the conversion in the reverse direction requires some degree of reprogramming and a high extent of selection. Tapping on the similarity to iPSC reprogramming, transcription factor mediated approaches have been a prominent choice. As EpiSCs do not express Klf4 or Klf2 as opposed to mESCs, it may be intuitive to reintroduce these key pluripotency transcription factors in an attempt to recapitulate mESC expression levels [14,48]. Transfection with Klf4 under 2i/LIF conditions generated converted cells, termed as Epi-iPSC, which exhibited mESC morphology and marker profile, and could contribute to germline-competent chimeras [14]. Given the high functional redundancy among members of the Kruppel-like family [59], Klf2 can substitute for Klf4 in the generation of Epi-iPSCs under 2i/LIF conditions [48]. Interestingly, transcription factor transduction was only effective in directing the conversion when combined with 2i/LIF conditions, but not with the EpiSC culture condition [14]. mESC characteristics were also maintained even when the transgene was excised [14], suggesting that the intrinsic molecular machinery had already been reprogrammed.

Klf2 and Klf4 target several pluripotency genes in common with Nanog, and can regulate Nanog expression [59], therefore Nanog could potentially mediate the same reprogramming process. Indeed, stable transfection of Nanog induced a similar reversion of EpiSCs to a mESC-like phenotype under 2i/LIF conditions at a 10-fold higher efficiency than Klf4 transfection [17]. Unlike Klf4, however, Nanog-mediated EpiSC reprogramming can occur under LIF/BMP4 conditions without the need for 2i. These Nanog-induced Epi-iPSCs exhibit an mESC-like expression profile, including upregulation of Klf4. Transient transfection of Nanog, however, did not yield any stable Epi-iPSCs and chimera formation was only observed upon co-transfection with Klf4. Therefore, Klf4 and Nanog may function synergistically in the conversion of EpiSC to a mESC-like state.

(iii). Epiblast stem cell conversion to mouse embryonic stem cell-like state using chemical approaches

Similar success has been achieved by other chemical-assisted reprogramming approaches that do not involve the exogenous introduction of specific transcription factors. mESC-like cells can also be generated from EpiSCs under purely 2i/LIF conditions without the need for transcription factors, and these cells have displayed chimera formation as well as germline transmission [50]. One possible explanation is that inhibition of the FGF/ERK pathway induces Klf2 expression in EpiSCs [50]. A more diverse inhibitor cocktail has also been established for the conversion of EpiSC to a mESC-like state. A combination of parnate and transforming growth factor (TGF)-ß inhibitor, used in conjunction with 2i/FGFR inhibitor, can reprogramme EpiSCs to cells that can contribute to germline [60]. Parnate can induce a global increase in the activating H3K4 methylation and is postulated to activate previously silenced pluripotency genes [61], highlighting the importance of epigenetic modification in facilitating cellular conversions.

Unlike other studies that require 2i conditions, Surani and co-workers [15] reported that EpiSCs could be reprogrammed into mESC-like cells in the presence of LIF/FCS and MEF feeder cells. These reprogrammed epiblast ESC-like cells (rESC) show gradually increasing similarity to mESCs in their gene expression profile with more passages, including an upregulation of E-cadherin. E-cadherin, a transmembrane molecule involved in intercellular adhesion, was previously shown to be stimulated by LIF/BMP4 and correlated with differentiation capability [62]. Furthermore, knockout of E-cadherin traps cells in a partially reprogrammed state [63]. Taken together, these studies indicate a potential requirement for E-cadherin in attainment of pluripotency.

LIF is an essential component of traditional mESC culture medium, and is frequently added to 2i cultures. Therefore, it would be informative to determine the effect of LIF, in particular the LIF–STAT3 pathway, in reprogramming. It was shown that STAT3, while insufficient on its own, can increase the efficiency of EpiSC reprogramming by several fold when used in combination with Klf4 or Nanog transfection, indicating that the enhancing effects of STAT3 extends beyond direct activation of these pluripotency factors [13]. A transient activation of STAT3, followed by transfer to 2i/LIF culture can recapitulate the effects of continuous STAT3 activation [13], suggesting that a short initial activation of LIF–STAT3 is sufficient to prime EpiSCs for reprogramming.

Both chemical and transcription factor approaches have proven to be viable in EpiSC reprogramming. However, a common ingredient in most of these methods is the usage of inhibitor cocktails and LIF. These inhibitors induce conversions to mESC-like cells by blocking critical differentiation pathways in ES cells, while LIF can additionally increase the efficiency of reprogramming.

(b). Interconversion between human embryonic stem cells and mouse embryonic stem cell-like human cells

Given that hESCs cannot be maintained under 2i conditions [64], a pressing question is whether hESCs represent a primed pluripotency state similar to EpiSCs, and therefore could be reverted back to the mESC-like ground state by appropriate culture conditions or reprogramming. Attempts to convert hESCs to a mESC-like state draw many parallels with the EpiSC to mESC-like switch. For this purpose, a combination of transcription factor transduction and chemical approaches has been used (figure 1b).

A recent study demonstrated that hESC can be converted to a naive, or mESC-like state by the constitutive expression of OCT4 and KLF4 or KLF4 and KLF2 under 2i/LIF conditions [65]. These naive hESCs gained a mESC-like colony morphology, displayed pre-X-inactivation status and survived single-cell passaging as well as inhibition of Activin/nodal signalling. However, the naive hESCs rapidly differentiated when the exogenous transcription factors were removed, unless Forskolin was added, which prolonged the maintenance of the cells [65]. Forskolin acts partly by inducing KLF4 and KLF2, emphasizing the importance of KLF factors in maintenance of the naive state.

There have been other efforts towards achieving certain characteristics of mESCs. In the presence of LIF, the combination of MEK and p38 inhibitors can convert hESCs to a mESC-like cell type [53]. These converted hESCs are more tolerant to single-cell passaging, a property that was correlated to an increased expression of E-cadherin and reduced dependence on integrin signalling. Although further characterization of the converted hESCs is required, this study suggests that a switch in cell adhesion systems may be required for hESCs to acquire certain mESC-like properties [53]. Besides cell adhesion molecules, atmospheric oxygen levels can also possibly affect maintenance of pluripotency. When hESCs were shifted from the physiological condition of 5 per cent oxygen to the atmospheric condition of 20 per cent oxygen, X chromosome inactivation occurred [66]. While X-inactivation could be an in vitro artefact, this finding presents the exciting possibility that culturing cells under physiological oxygen conditions could result in a closer approximation to the mESC-like ground state.

Further work is needed to elucidate if a ground state of pluripotency exists in hESCs. Studies demonstrating the conversion of EpiSCs to mESC-like cells have been informative regarding the factors affecting hESCs and possible manipulation techniques. For instance, there are several common features shared between the conversion of hESCs to mESC-like state and the switch from EpiSC to mESC-like cells, including the contribution of Klf4 and E-cadherin. However, the more extensive use of inhibitors and requirement of sustained overexpression of transcription factors in hESC reprogramming seem to suggest that hESCs are less permissive to reprogramming than NOD strain mice or rats [65]. Currently, there is also a lack of evidence to show that mESC-like hESCs have expanded developmental potential. Therefore, for definitive conclusions to be drawn, more research should be conducted in human stem cell models in addition to experiments on the EpiSC to mESC-like switch. In particular, it would be interesting to find out if hESC with mESC-like features can be directly derived using 2i conditions.

5. Future outlook

Our view of pluripotency has been extensively shaped by recent developments. Multiple different approaches have elucidated a highly interconnected transcription factor network. To better discern the contribution of each factor within the network, further in-depth functional studies would be required. High-throughput methodologies coupled with a multidisciplinary approach will also be useful for studying the interactions among the components of the pluripotency network and providing a systems-level overview. The modulation of intracellular signalling by chemical inhibitors has allowed the derivation of pluripotent cells across species. Furthermore, interconversion studies have shown that with the aid of inhibitors, transcription factor mediated reprogramming can be accomplished without permanent transgene integration. The progress towards discovery of new chemical inhibitors paves the way for safer and more effective manipulation of cell fate. The feasibility of interconversions indicates that pluripotent cells are highly plastic depending on the culture conditions used. For example, cells cultured in 2i/LIF media convert to a mESC-like state. This leads to the crucial question regarding what are the true defining characteristics of each cell type.

An in-depth genetic understanding of pluripotency is highly useful in therapeutics, disease and development. Knowledge of the effect of culture conditions, and in particular the conversion of hESC to ground state pluripotency, could facilitate easier manipulation and culturing. Pluripotency is also a key feature of the developing embryo, and pluripotent stem cell models may provide a glimpse of developmental progression and lineage specification events in vivo. Ultimately, unravelling the regulatory networks in pluripotency and moving forward in instructing cell fate will allow for more directed differentiation into useful cell types and will offer new insights in developmental biology.