• Open Access

Clonal Isolation of an Intermediate Pluripotent Stem Cell State§


  • Kuo-Hsuan Chang,

    Corresponding author
    1. Stem Cell Neurogenesis Group, Institute of Clinical Sciences, Imperial College London, London, United Kingdom
    2. Department of Neurology, Chang Gung Memorial Hospital Linkou Medical Center and College of Medicine, Chang Gung University, Taoyuan, Taiwan
    • Department of Neurology, Chang Gung Memorial Hospital Linkou Medical Center and College of Medicine, Chang Gung University, No. 5, Fusing St., Gueishan Township, Taoyuan County 333, Taiwan
    Search for more papers by this author
    • Telephone: 886-3-3281200, ext. 8348; Fax: +886-3-3288849

  • Meng Li

    1. Stem Cell Neurogenesis Group, Institute of Clinical Sciences, Imperial College London, London, United Kingdom
    Search for more papers by this author

  • Author contributions: K.H.C.: concept and design, execution of all experiments, data analysis and interpretation, obtained funding, and wrote the paper; M.L.: concept and design, data analysis and interpretation, obtained funding, and wrote and finalized the paper.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS January 22, 2013.


Pluripotent stem cells of different embryonic origin respond to distinct signaling pathways. Embryonic stem cells (ESCs), which are derived from the inner cell mass of preimplantation embryos, are dependent on LIF-Stat3 signaling, while epiblast stem cells (EpiSCs), which are established from postimplantation embryos, require activin-Smad2/3 signaling. Recent studies have revealed heterogeneity of ESCs and the presence of intermediate pluripotent stem cell populations, whose responsiveness to growth factors, gene expression patterns, and associated chromatic signatures are compatible to a state in between ESCs and EpiSCs. However, it remains unknown whether such cell populations represent a stable entity at single-cell level. Here, we describe the identification of clonal stem cells from mouse ESCs with global gene expression profiles representing such a state. These pluripotent stem cells display dual responsiveness to LIF-Stat3 and activin-Smad2/3 at single-cell level and thus named as intermediate epiblast stem cells (IESCs). Furthermore, these cells show accelerated temporal gene expression kinetics during embryoid body differentiation in vitro consistent with a more advanced differentiation stage than that of ESCs. The successful isolation of IESCs supports the notion that traverse from naïve ground state toward lineage commitment occurs gradually in which transition milestones can be captured as clonogenic entity. Our finding provides a new model to better understand the multiple pluripotent states. STEM CELLS 2013;31:918–927


Blastocyst-derived embryonic stem cells (ESCs) and postimplantation epiblast-derived epiblast stem cells (EpiSCs) share a number of cellular features, such as pluripotency and ability to self-renewal [1–3]. However, they use distinct signaling pathways for self-renewal. ESCs are sustained by LIF-Stat3 signaling, while EpiSCs are maintained by the activin-Smad2/3 pathway. ESCs can differentiate into EpiSCs when cultured in activin and fibroblast growth factor 2 (FGF2) in a chemically defined medium, while EpiSCs spontaneously revert to ESC-like cells when propagated in a medium containing leukemia inhibitor factor (LIF) and fetal calf serum (FCS) in the presence of mouse embryonic fibroblast (MEF) feeders or by transfection with defined factors in a feeder-free condition [4–7]. Furthermore, ESCs in LIF have been reported to exist in a metastable pluripotent state, with subpopulations fluctuating between the ESC and EpiSC stages, as indicated by the differential expression levels of Rex1 and Stella [8, 9]. This gene expression feature appears to be shared by bFGF, Activin and BIO-derived stem cells (FAB-SCs), a blastocyst-derived culture established in activin A, FGF2, and BIO in the presence of feeders [10]. However, such an intermediate pluripotent ground state(s) has not been captured in a stable cell line. In this study, we report that clonal intermediate pluripotent lines of activin- and LIF-responsive stem cells can be established from ESCs. These cells provided a unique opportunity to investigate pluripotency and lineage commitment.


Cell Culture

E14tg2a ESCs [11] and E14tg2a-derived clonal IESC cell lines were maintained in gelatin-coated tissue culture plates in glasgow minimum essential medium (GMEM) containing 10% FCS, supplemented with LIF (prepared inhouse) or 12 ng/ml activin A (R&D system, Minneapolis/MN, http://www.rndsystems.com), respectively. IESC-derived EpiSCs (Epi-IESCs) and ESC-derived EpiSCs (EpiSCs) were cultured on a fibronectin (Millipore, Billerica/MA, http://www.millipore.com)-coated surface in N2B27 supplemented with 12 ng/ml activin A and 8 ng/ml FGF2 (R&D system, Minneapolis/MN, http://www.rndsystems.com) [5]. ESCs and IESCs were passaged as single-cell suspensions by trypsinization, while Epi-EpiSCs and EpiSCs were dissociated as small-cell clumps by mechanical scraping in 200 U/ml collagenase IV (Gibco, Paisley/UK, www.invitrogen.com/gibco). For Jak and Smad inhibitor treatment, 1 × 105 of cells were plated in N2B27 alone for 12 hours before treatments. Cells were exposed to 10 μM SB431542 (Sigma) and 1 μM Jak inhibitor (JAKI) (Calbiochem, Nottingham/UK, http://www.calbiochem.co.uk) for specific durations before harvesting. For embryo body (EB) differentiation, ESCs or IESCs were plated in bacteria-grade culture dishes in N2B27 (for neural induction) or FCS medium (for meso-endodermal induction). EBs were harvested every other day from day 2 to 14 for RNA extraction or fixed and sectioned on day 10 for immunocytochemistry.

Growth Curve Analysis

ESCs, IESCs, Epi-IESCs, and EpiSCs were plated at 1,000 cells per well in 24-well plates. Cells were detached with 0.025% trypsin and the cell number was counted using a hemocytometer (BDH, Dagenham, UK, http://www.bdh.com) every 24 hours. The population doubling time for each cell line was then determined.

Quantitative RT-PCR

Total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad/CA, http://www.invitrogen.com). One microgram of RNA was reversely transcribed into cDNA using SuperScript III Super Mix kit (Invitrogen, Carlsbad/CA, http://www.invitrogen.com) according to the manufacturer's instructions. Quantitative RT-PCR (qPCR) was performed using MESA GREEN qPCR Mastermix Plus (Eurogentec, Liège/Belgium, http://www.eurogentec.com), and the reaction was run in a Chromo 4 real-time PCR system (Bio-Rad, Hercules/CA, http://www.bio-rad.com). Briefly, 2 μl 1:10 diluted cDNA was mixed with 10 μl MESA GREEN PCR mix (Eurogentec, Liège/Belgium, http://www.eurogentec.com), 10 μM forward primer, 10 μM reverse primer, and 6 μl water. The amplification was done for 40 cycles (94°C 1 minute, 65°C 1 minute, and 72°C 1 minute). To verify the PCR product, melting curve was carried out in each reaction. Relative gene expressions were calculated using the 2−ΔΔCt method [12], and ΔCt = Ct (target gene) − Ct (β-Actin), in which Ct indicates cycle threshold (the fractional cycle number where the fluorescent signal reaches detection threshold). The primer sequences used in this study are shown in supporting information Table S1.

Chromatin Immunoprecipitation

Approximately 1 × 107 cells were crosslinked with 1% formaldehyde for 5 minutes at room temperature. After sonication, chromatin–DNA complex was precipitated with Dynabeads Protein G (Invitrogen, Carlsbad/CA, http://www.invitrogen.com)-linked anti-trimethyl Lys 4 histone H3 (07-473, Upstate, Lake Placid/NY, http://www.upstate.com), anti-trimethyl Lys 27 histone H3 (07-449, Upstate, Lake Placid/NY, http://www.upstate.com), or anti-histone H3 (Upstate, Lake Placid/NY, http://www.upstate.com). Eluted DNA was amplified by qPCR using the primers listed in supporting information Table S2.


The primary antibodies used for immunocytochemistry in this study were: rabbit anti-Oct4 (1:500, Abcam, Cambridge/UK, http://www.abcam.com), mouse anti-Sox2 (1:500, Millipore, Billerica, MA, http://www.millipore.com), mouse anti-SSEA1 (1:100, Developmental Studies Hybridoma Bank, Iowa City/IA, http://dshb.biology.uiowa.edu), rabbit anti-Otx2 (1:500, Millipore, Billerica, MA, http://www.millipore.com), mouse anti-β3-Tubulin (1:1,000, Covance, Princeton/NJ, http://www.covance.com), goat anti-Afp (1:100, Santa Cruz, Dallas/TX, http://www.scbt.com), and rabbit anti-Pax2 (1:100, Covance, Princeton/NJ, http://www.covance.com). The secondary antibodies were Alexa594-conjugated donkey anti-rabbit IgG (1:200, Invitrogen, Carlsbad, CA, http://www.invitrogen.com), Alexa488-conjugated donkey anti-goat IgG (1:200, Invitrogen, Carlsbad, CA, http://www.invitrogen.com), and Alexa488-conjugated donkey anti-mouse IgG (1:200, Invitrogen, Carlsbad, CA, http://www.invitrogen.com). Antibodies used for Western blot were goat anti-Lamin B (1:2,000, Santa Cruz, Dallas/TX, http://www.scbt.com), rabbit anti-LIF receptor (LIFR) (1:1,000, Santa Cruz, Dallas/TX, http://www.scbt.com), rabbit anti-Nanog (1:1,000, Abcam, Cambridge/UK, http://www.abcam.com), rabbit anti-Smad2/3 (1:1,000, Cell Signaling, Danvers/MA, http://www.cellsignal.com), and rabbit anti-p-Smad2 (1:1,000, Cell Signaling, Danvers/MA, http://www.cellsignal.com); the secondary antibodies used were goat anti-rabbit IgG-HRP (1:2,500, GE Healthcare, Little Chalfont/UK, http://www.gehealthcare.com) and donkey anti-goat IgG-HRP (1:2,500, Santa Cruz, Dallas/TX, http://www.scbt.com).

Microarray Analysis

Mouse Expression-Array-430.2.0 Genechips (Affymetrix, Santa Clara/CA, http://www.affymetrix.com) was used in this study. Each cell type was represented by three biologically independent replicates. Data were analyzed using DNA-Chip Analyzer (dChip; http://www.dchip.org).

Teratoma Formation and Histology

Nude mice were anesthetized with diethyl ether. IESCs were suspended in 10% FCS-GMEM at 1 × 107 cells per milliliter and 100 μl of the cell suspension was injected subcutaneously into the dorsal flank of nude mice. Four weeks after the injection, tumors were surgically dissected from the mice, fixed in 4% formaldehyde, and embedded in paraffin. Sections were stained with hematoxylin and eosin.

Morula Aggregation

Morula aggregation was performed according to the standard procedures as described in [13].


Establishment of Clonogenic Activin- and LIF-Responsive Cell Lines

EpiSCs can spontaneously revert to ESC-like cells in the presence of MEF feeders in medium containing LIF and FCS; this suggests that serum- or feeder-derived factors may facilitate the maintenance of a stem cell in transition between the two states [4, 14]. We hypothesized that activin A (referred to as activin thereafter) may support such a stem cell state and therefore cultured E14tg2a ESCs in a serum-containing medium in the presence of activin instead of LIF (Fig. 1A). Under this condition, an average of 29 undifferentiated primary colonies was formed in 7 days in a plate seeded with 2 × 103 cells. Sister cultures plated in LIF typically produced approximately 500 colonies. A total of 144 activin-derived primary colonies were obtained from five plates and independently expanded in activin. Within 10 days, 34 of these colonies were able to expand as morphologically undifferentiated cultures, while the rest differentiated and were discarded. To determine whether the expanded cultures retained LIF responsiveness, we subjected each of the 34 clones to a colony-forming assay, with the cells ranging from passages 3–5, in LIF, activin, and medium alone. Alkaline phosphatase (AP) staining was performed 5 days later to label undifferentiated cells. Twelve clones produced a similar number of AP+ colonies in LIF, activin, and medium alone, thereby suggesting that they may be factor independent (supporting information Fig. S1A). Nine clones produced abundant AP+ colonies in LIF, but few in activin and medium alone, indicating that they remain LIF dependent (supporting information Fig. S1B). Conversely, 13 clones generated AP+ colonies in both LIF and activin with similar efficiency, but few AP+ colonies in medium alone (supporting information Fig. S1C). Within this group, clones number 10 and 16 had the highest ratio between the number of AP+ colonies in activin and that in media alone. For this reason, these two clones were chosen for further studies and routinely propagated in activin. After 25 passages, cells from both clones were able to form colonies in LIF and in activin, at similar frequencies (Fig. 1B). These results demonstrate that a proportion of ESCs (0.13%) can, or can be induced to, respond to activin, while retaining responsiveness to LIF for self-renewal. These cells were designated as intermediate epiblast stem cells (IESCs), and the clones 10 and 16 were renamed as IESC1 and IESC2, respectively.

Figure 1.

Characterization of IESCs. (A): Experimental procedure for the generation and validation of IESCs. Thirty-four clones were established from 144 original colonies. Colony-forming assay was performed for each clone in LIF, Activin, and media alone. Twelve clones produced abundant alkaline phosphatase (AP+) colonies in LIF but few in Activin or media control. Nine clones produced AP+ colonies in all three conditions while 13 clones generated abundant AP+ colonies in either LIF or Activin condition at similar efficiency with few AP+ colonies in media alone. (B): Bar graph illustrating the number of AP+ colonies formed in activin, LIF, and medium alone by passage 25 IESC1 and IESC2. (C): Bright-field image of IESCs, ESCs, and Epi-IESCs. Scale bar = 200 μm. (D): Growth curves of ESCs, IESCs, Epi-IESCs, and EpiSCs. One thousand cells were plated into each well of six-well plates at day 0. (E): Immunofluorescent staining for Oct4 (green), Sox2 (white), SSEA1 (Red), and Otx2 (yellow) in ESCs, IESCs, Epi-IESCs, and EpiSCs. Scale bar = 100 μm. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; ESC, embryonic stem cell; EpiSC, epiblast stem cell; IESCs, intermediate epiblast stem cells; LIF, leukemia inhibitory factor.

IESCs Are Developmentally Proximal to ESCs and Readily Differentiate into EpiSCs

When plated at low density, IESC1 cells (IESC1s) and IESC2 cells (IESC2s) formed mostly dome-like colonies with differentiated cells at the periphery (Fig. 1C). Unlike EpiSCs, IESCs are resistant to trypsinization and can be passaged as single-cell suspensions. Monolayers of IESCs are morphologically similar to ESCs. IESC1s and IESC2s showed slower growth rate than ESCs with a population doubling time of 18.86, 18.93, and 16.62 hours, respectively (Fig. 1D). Immunocytochemical analysis revealed homogenous expression of Oct4, Sox2, and SSEA1 (Fig. 1E). However, antibody staining revealed that the expression of Otx2, which is induced in EpiSCs [3], was not present in IESCs (Fig. 1E). These observations suggested that IESCs are developmentally proximal to ESCs and distinct from EpiSCs.

To further confirm the above hypothesis, we transferred IESCs to feeder-free EpiSC culture conditions in activin and FGF2 in N2B27 medium [5]. This treatment resulted in prominent and rapid morphological changes, as demonstrated by the appearance of compact, flat colonies resembling EpiSCs (Fig. 1C). By passage 5 (approx. 2 weeks), the cultures became morphologically homogeneous, with the majority of cells acquired Otx2 expression while maintaining Oct4, Sox2, and SSEA1 expressions (Fig. 1E). This expression pattern is characteristic of EpiSCs derived from the embryo or mouse ESCs [1, 3, 5] Furthermore, these cells showed a similar population doubling time to EpiSCs (around 36 hours, Fig. 1D). We therefore designated the derived cells as Epi-IESCs. These findings indicate that IESCs are developmentally upstream of EpiSCs.

IESCs Have a Global Gene Expression Profile Intermediate to that of ESCs and EpiSCs

We then performed genome-wide expression profiling using the Affymetrics MOE430.2 GeneChip to further elucidate the similarities and differences among IESCs, ESCs, and Epi-IESCs We found that the expression levels of many pluripotent genes, including Rex1, Stella, Pecam1, Klf4, Nanog, Fbx015, and Tbx3, were lower in IESCs (passage 20) compared to ESCs, but higher compared to Epi-IESCs (Fig. 2A). Conversely, the expression levels of gene markers associated with epiblasts, such as Fgf5, Nodal, Lefty1, and Foxa2, were found to be higher in IESCs than ESCs, but lower than Epi-IESCs. Furthermore, whole transcriptome hierarchical cluster analysis placed IESCs between ESCs and Epi-IESCs (Fig. 2B).

Figure 2.

Expression profiling of IESCs. (A): Heat map illustrating the expression level of 26 frequently referred pluripotency and early lineage gene markers. (B): Hierarchy clustering of 41,174 probes expressed by ESCs, IESC1s (passage 15), and Epi-IESC1s (passage 10). High expression level is indicated in shades of red and low expression in shades of green. (C, D): Quantitative RT-PCR analysis of pluripotency (C) and early lineage marker genes (D) in ESCs, IESCs, Epi-IESCs, and EpiSCs. Data were normalized to β-Actin and ESCs. (E, F): Chromatin immunoprecipitation-PCR data showing the association of H3K4Me3 (E) and H3K27Me3 (F) with promoters (-P) and coding (-C) sequences of Oct4, Nanog, Pecam1, Rex1, Stella, Klf4, Otx2, Fgf5, Nkx2.2, and Myf5 in ESCs, IESCs, and Epi-IESCs. Data shown are the average ± SD. Abbreviations: ESC, embryonic stem cell; EpiSC, epiblast stem cell; IESCs, intermediate epiblast stem cells.

To validate the microarray data, we examined the expression profiles of IESC1s and IESC2s against a panel of frequently referred pluripotency and early lineage gene markers by qPCR (Fig. 2C). Consistent with the microarray data, Rex1, Stella, Klf4, Pecam1, and Nanog transcript levels were evidently lower in IESCs and Epi-IESCs compared to ESCs, while the levels of Oct4 were similar in IESCs, ESCs, Epi-IESCs, and EpiSCs. In contrast, the levels of a number of genes expressed in the epiblast and early germ layers, such as Fgf5, Brachyury, Nodal, Pitx2, Otx2, Gata6, and Foxa2, were lower in IESCs than Epi-IESCs or EpiSCs but higher than ESCs (Fig. 2D).

To investigate whether chromatin remodeling is associated with the above observed gene expression, we performed chromatin immunoprecipitation (ChIP) analysis to determine histone modifications indicative of gene activation (H3K4Me3) and repression (H3K27Me3) in ESCs, IESCs, and Epi-IESCs. The chromatin status in the promoters and the coding regions was examined for Oct4, Nanog, Pcam1, Rex1, Stella, Klf4, Otx2, and Fgf5. As controls, we also included a bivalent gene Nkx2.2 [15] and a myoblast-specific gene Myf5 [16]. Consistent with previous findings, Nkx2.2 was associated with both active H3K4Me3 and repressive H3K27Me3 in all cultures, while Myf5 showed a preferential binding to H3K27Me3. We found that while the association of H3K4Me3 with Oct4 and Nanog was comparable in ESCs, IESCs, and Epi-IESCs (Fig. 2E), its association with Pecam1, Rex1, Stella, and Klf4 was more pronounced in ESCs than in IESCs and Epi-IESCs. In contrast, the association of H3K4Me3 with Otx2 and Fgf5 was more apparent in IESCs and Epi-IESCs than in ESCs. Consistent with the H3K4Me3 ChIP data, the association of the repressive H3K27Me3 with pluripotent factors Pecam1, Rex1, Stella, and Klf4 was preferentially found in IESCs and Epi-IESCs. In contrast, its association with Fgf5 and Otx2 was enriched in ESCs than in IESCs and Epi-IESCs (Fig. 2F).

We then investigated the ability of IESCs to reverse back to an ESC-like state. To this end, we cultured IESC1s in LIF for three passages (IESC1-Ls) and examined their gene expression profile by microarray. Interestingly, the expression of Rex1, Stella, Pecam1, Klf4, Nanog, Fbx015, and Tbx3 in IESC1-Ls reverted to levels similar to those in ESCs (supporting information Fig. S2A). Accordingly, hierarchical clustering analysis grouped IESC1-Ls together with ESCs in the same cluster (supporting information Fig. S2B). This finding suggests that the IESC intermediate state could be reverted to ESC-like state by LIF. Together, our global gene expression and epigenetic profiling of IESCs support the notion that they represent stem cells in between ESC and EpiSC and are capable of reverting back to an ESC-like state.

Activin-Smad2/3 and LIF-Stat3 Pathways Are Active and Coupled to Pluripotent Gene Expression in IESCs

The activin-Smad2/3 pathway controls EpiSCs, but not ESCs, by directly regulating Nanog [3, 14]. The maintenance of IESCs by activin suggests that Nanog is a possible target of activin/Smad2/3 signaling. Therefore, we determined Nanog protein expression by Western blotting after a 24-hour activin treatment of IESC1s, IESC2s, ESCs, and Epi-IESC1s (Fig. 3A). Activin-induced Smad2 phosphorylation (p-Smad2) was observed in all cell types (Fig. 3B). This was coupled to a pronounced increase in Nanog protein level in Epi-IESCs but not in ESCs. Interestingly, induction of Nanog was evident in both IESC1s and IESC2s. Conversely, the treatment of Epi-IESCs, IESC1s, and IESC2s by an activin inhibitor SB431542 for 3 and 12 hours resulted in a gradual decrease in Nanog transcript level, while the same treatment had no effect on ESCs (Fig. 3C). Furthermore, activin rescued Nanog expression in IESCs and Epi-IESCs previously treated for 12 hours with SB431542. We next examined the expression levels of activin receptor and Smad families determined by microarray. Surprisingly, we found relatively little changes in the expression levels of activin and Smad2/3/4 (Fig. 3D). Together, these data suggest that IESCs acquired the characteristics of EpiSC in signaling usage rather than alteration in the expression levels of activin-Smad2/3 pathways.

Figure 3.

Activin-Smad2/3 and LIF-Stat3 signaling in IESCs. (A): Illustration of the experimental scheme. (B): The phosphorylation of Smad2 (p-Smad2) accompanied by an increase of Nanog expression in IESC1, IESC2, and Epi-IESC1. Cells were lysed after 24-hour activin treatment. (C): Quantitative RT-PCR (qPCR) analysis for Nanog expression by ESCs, IESCs, and Epi-IESCs cultured with or without the Smad inhibitor SB431542. (D): The changes in the levels of Activin receptors (Acvrs) and Smad2/3/4 with the change in the cell state were assessed by comparison against the level of each gene in ESC. (E): qPCR analysis for the effects of Jak inhibitor JAK1 on Socs3 and Klf4 expression by ESC, IESCs, and Epi-IESCs. (F): qPCR analysis for Socs3 and Klf4 expression in response to LIF by ESC, IESCs, and Epi-IESCs. Data were normalized to β-Actin and are the average ± SD values of no treatment at the same time points. (G): The changes in the levels of the components of LIF/STAT3 signaling with the change in the cell state were assessed by comparison with the level of each gene in ESC. (H): Decreased LIFR expression in Epi-IESC1 and Epi-IESC2. Abbreviations: ESC, embryonic stem cell; EpiSC, epiblast stem cell; IESCs, intermediate epiblast stem cells; LIF, leukemia inhibitor factor; LIFR, LIF receptor.

LIF-induced self-renewal of ESCs involves the downstream activation of Stat3 and its targets, such as Socs3 and Klf4 [3, 17, 18]. To determine whether Stat3 signaling is active in IESCs, we administered a pulse of LIF to IESCs and determined its effect on the expression of Stat3 target genes Socs3 and Klf4 (Fig. 3A). We found that, as observed in ESCs, LIF treatment lead to a robust induction of Socs3 and Klf4 in IESC1s and IESC2s (Fig. 3E). However, little change was observed in Epi-IESCs. In keeping with the above finding, exposure of IESCs to a JAK inhibitor JAKI for 12 hours in N2B27 resulted in a 3-, 2.5-, and 2-fold decrease in their levels of Socs3 and Klf4 transcripts, respectively, as observed in ESCs. However, treatment with LIF for 4 hours resulted in a full recovery of the Socs3 and Klf4 levels in both ESCs and IESCs (Fig. 3F). JAKI treatment had no effect on Epi-IESC1s with regard to Socs3 and Klf4 expression. Furthermore, we found that the expression levels of Lif and Stat3 in LIF-independent Epi-IESC1s were only slightly decreased compared to the levels in ESCs (Fig. 3G). However, the levels of Lifr expression were dramatically decreased in Epi-IESC1s. Western blotting analysis further confirmed the low-level expression of LIFR in Epi-IESCs (Fig. 3H). This downregulation may abolish LIF-mediated activation of Stat and its downstream targets, which explains the unresponsiveness of Epi-IESCs to LIF.

IESCs Undergo Multilineage Differentiation In Vitro at a Faster Pace than ESCs

ESCs can differentiate into all three germ-layer derivatives in EBs in an orderly fashion that mimics embryonic development [18, 19]. Similar to ESCs, we found that IESCs readily form EBs in both defined and serum-containing media and that the resultant EBs had dimensions similar to those formed by ESCs (Fig. 4A). During the course of a 14-day differentiation period, IESC1 and IESC2 EBs displayed temporally regulated expression of early lineage-specific genes, such as Pax6, Nestin, and β3-Tubulin (neuroectoderm), Brachyury, Goosecoid, and Pax2 (mesoderm), and Bmp4 and Afp (endoderm). Interestingly, the induction of these early lineage marker genes and the subsequent decline in their levels were detected earlier in IESCs, Epi-IESCs, and EpiSCs than in ESCs (Fig. 4B–4D). For instance, the level of Goosecoid in IESCs, Epi-IESCs, and EpiSCs began to increase at day 4, while the increase of Goosecoid transcript in ESCs was observed on day 8 (Fig. 4C). Consistent with the transcript analysis, immunostaining of day 10 EBs revealed a higher number of β3-Tubulin+, Pax2+, and Afp+ cells in IESC1s cultures than those in ESCs (Fig. 4E–4G). These data suggest that IESCs can participate in multilineage differentiation. Importantly, these cells can be more easily “primed” toward somatic differentiation than ESCs.

Figure 4.

IESCs enter lineage differentiation sooner than ESCs. (A): Embryo bodies (EBs) generated by ESCs and IESCs. Scale bar = 200 μm. (B): Quantitative RT-PCR (qPCR) analysis of neuroectodermal marker gene expressions during 14-day EB differentiation of ESC, IESCs, Epi-IESCs, and EpiSC. (B–D): qPCR analysis of neuroectodermal, mesodermal, and endodermal marker gene expressions during 14-day EB differentiation of aforementioned cell lines. Data shown are the average ± SD. (E–G): β3-Tubulin+, Pax2+, and Afp+ cells in EBs after 10-day differentiation. Scale bar = 50 μm. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; ESC, embryonic stem cell; EpiSC, epiblast stem cell; IESCs, intermediate epiblast stem cells.

Teratoma Formation and Chimera Contribution of IESCs

To further investigate the differentiation capacity of IESCs, we grafted IESC1 and IESC2 subcutaneously into immunocompromised mice. Four weeks later, cells from both IESC lines developed tumors. Hematoxylin and eosin stain of these tumor sections revealed the presence of hair follicle and neural tissue (ectoderm, Fig. 5Aa, 5Ab), bone and cartilage (mesoderm, Fig. 5Ac, 5Ad), and gut epithelium and acinar gland (endoderm, Fig. 5Ae, 5Af), representing the three major germ layers.

Figure 5.

Teratoma formation and chimera contribution of IESCs. (A): Hematoxylin and eosin staining of teratomas derived from IESCs confirmed multilineage differentiation. Hair follicle and neural tissue (ectoderm, a, b), bone and cartilage (mesoderm, c, d), and gut epithelium and acinar gland (endoderm, e, f) representing the three major germ layers. (B): GFP-labeled IESCs and control ESCs were aggregated with morula embryos and their integration into blastocysts was examined 24 hours later. Scale bars = 200 μm. (C): Summary of the morula aggregation data. Abbreviations: ESC, embryonic stem cell; GFP, green fluorescent protein; ICM, inner cell mass; IESCs, intermediate epiblast stem cells.

ESCs can colonize blastocyst and contribute to the germline while EpiSCs are unable to form chimeras [3]. We then assessed the ability of IESCs to colonize blastocyst chimeras by morula aggregation. IESC1s and ESCs engineered to constitutively express green fluorescent protein (GFP) were used in this study for the direct visualization of the contribution of IESC derivatives. Of the 56 blastocysts obtained in the IESC aggregation group, 19 (33.93%) expressed GFP, and all GFP+ cells were found in the inner cell mass (ICM, Fig. 5B, 5C). This efficiency is slightly lower than that of ESC controls where 31 of the 58 (53.45%) blastocysts contained GFP+ cells in the ICM. However, when recovered at midgestation stage (9.5 d.p.c.), most of the highly chimeric embryos were developmentally retarded (supporting information Fig. S3), while the normal looking embryos had low level of IESC contribution. Therefore, while IESCs share characteristics of ESCs in their ability to colonize blastocysts, their developmental stage is not compatible with the host tissue to warrant further development.


In this study, we report, for the first time, the establishment of clonogenic stem cell lines that bridge ESCs and EpiSCs (Fig. 6). These cells depend upon activin-Smad2/3 signaling for self-renewal while also retain LIF-Stat3 responsiveness. They are primed toward multilineage differentiation in vitro while capable of integration into the ICM. Our study provides evidence to support the notion that the transition from the naïve ground state toward lineage commitment occurs gradually and that transition milestones can be captured as stable clonogenic entities.

Figure 6.

A proposed model for the pluripotent state of IESCs. IESCs exist in an intermediate cell state between ESCs and EpiSCs. Notably, IESCs express lower levels of Rex1 and Stella and higher levels of Fgf5 and Otx2 than ESCs. Pluripotent cells in such an intermediate state are biased toward differentiation. Abbreviations: ESC, embryonic stem cell; EpiSC, epiblast stem cell; ICM, inner cell mass; IESCs, intermediate epiblast stem cells; LIF, leukemia inhibitory factor.

ESCs cultured in the traditional serum-LIF condition include a small proportion of cells with low expression levels of Nanog, Rex1, and Stella, which are primed to differentiation [8, 9, 19]. The gene expression patterns of Stella and Rex1 ESCs is akin to that of postimplantation epiblasts [8, 9]. For all the three markers, fluorescence-activated cell sorting (FACS)-purified fractions of marker positive or marker-negative subpopulations can reestablish the parental heterogeneous equilibration [8, 9]. This suggests that in the serum-LIF condition, mouse ESCs are truly in a metastable state in which a dynamic balance exists between the ground and primed states. In this study, we found that the serum-activin condition can support clonal self-renewal of IESCs in a primed state. The early embryonic gene expression patterns suggested that IESCs are in an intermediate cell state between ESCs and EpiSCs. Although IESCs are poised to differentiation, they display gene expression and functional characteristics different from those of Nanog, Stella, or Rex1 ESC subpopulations. Expression levels of a series of pluripotent markers, including Rex1, Stella, Fbx15, and Nr0b1, were low in IESCs, but high in Nanog ESCs [19]. Stella ESCs seldom form EBs, whereas IESCs could efficiently generate EBs. Rex1 ESCs are unable to contribute to embryonic tissues. Considering their gene expression patterns, IESCs are possibly a subpopulation of the Nanog, Rex1, or Stella ESCs captured by culturing in the serum-activin condition. This also implies that a fraction of the Nanog, Rex1, and/or Stella ESCs in routine serum-LIF culture condition may be in a similar state to IESCs. An interesting experiment would be to compare the efficiency of IESC generation from FACS-sorted cells expressing high and low levels of Nanog, Stella, or Rex1.

A striking feature of IESCs is their unique responsiveness to signaling pathways. IESCs efficiently form undifferentiated colonies in both LIF and activin culture conditions. This dual responsiveness to LIF-Stat3 and activin-Smad2/3 signaling pathways was further confirmed by the analysis of short-term treatment with JAK and activin inhibitors. The inactivation of the Smad2/3 pathway in EpiSCs has been shown to promote their neuroectodermal differentiation, which can be prevented by constitutive Nanog expression [20]. In contrast, activin–Smad2/3 signaling does not seem to promote Nanog expression and self-renewal in ESCs [14]. The short-term exposure to activin cannot induce the expression of Nanog in ESCs, despite the phosphorylation of Smad2 by activin. Our data suggest that Smad2/3 signaling is possibly controlled directly by Nanog expression in IESCs, as in the case of EpiSCs. Considering its critical role in the maintenance of pluripotency, the induction of Nanog could explain the activin-responsiveness of IESCs. Although Nanog is not a downstream target of Smad2/3 signaling in mouse ESCs, it is unclear whether Nanog in ESCs may be responsible for the lack of responsiveness to Smad2/3 signaling. Similar to the findings of Yang et al. [7], our data demonstrated the downregulation of the LIFR in Epi-IESCs. These discoveries indicate that IESCs possess signaling-responsive features of both ESCs and EpiSCs and unravel the regulation of the self-renewal of cells in this intermediate pluripotent state.

We found that nine out of the 34 primary clones analyzed exhibited colony formation in the absence of both activin and LIF. These apparent factor-independent cells might have undergone karyotypic transformation. It is also possible that they were supported by LIF or other factors secreted by surrounding differentiated or differentiating cells.

Stem cells in different pluripotent states have distinct intracellular signal transduction pathways and display distinct levels of potential for cell commitment. For example, BMP4 has been reported to be essential for the self-renewal of mouse ESCs, while it promotes non-neural lineage differentiation and suppresses neural commitment in EpiSCs [21]. In mice, ERK signaling has been shown to exert its effects by inhibiting both reversion to a preimplantation ESC-like state and forward differentiation to the neural lineage [14]. Considering the potential heterogeneity of multiple pluripotent states in ESC populations, it is plausible that only some ESCs produce the required response to a given differentiation condition and commit to the desired cell phenotypes, while other ESCs may respond differently to generate other cell types. This notion is supported by the finding that neural fate commitment by EpiSCs is much more reliable and efficient than that by ESCs [22]. Thus, the synchronicity of the cell state before/during differentiation could play a critical role in lineage/cell-fate commitment.


In summary, we report, for the first time, the establishment of stable clonogenic stem cell lines IESCs that bridge ESCs and EpiSCs. We provide evidence that IESCs self-renewal via activin-Smad2/3 signaling while also retain LIF-Stat3 responsiveness. They display unique gene expression and epigenetic profiles compatible to a state in between ESCs and EpiSCs and can progress to EpiSCs in culture when exposed to FGF2. IESCs can undergo major germlayer differentiation in vitro in EBs and can colonize the blastocyst embryo in vivo. Our work demonstrates that transition from naïve ground state toward lineage commitment occurs gradually in which transition milestones can be captured as a clonogenic entity in cell culture. The isolation of intermediate-state pluripotent cells in this study provides a new model to better understand the multiple pluripotent states.


We thank Austin Smith, Peter Andrews, Xinsheng Nan, and Emily Gale for helpful discussion and comments on the manuscript. We are grateful to Bill Mansfield and Jennifer Nichols for instruction on morula aggregation and David Chambers for microarray. This study was sponsored by a Ph.D. training fellowship from Chang Gung Memorial Hospital, Taipei, Taiwan (CMRPG 360041-43 and 3A0691-93), grants from the UK Medical Research Council (G117/560 and U120005004), and the National Science Council, Executive Yuan, Taiwan (NSC 100-2314-B-182A-076-MY1-2).


The authors have declared that no competing interests exist.