SEARCH

SEARCH BY CITATION

Keywords:

  • Embryonic stem cells;
  • Neural induction;
  • Pluripotent stem cells;
  • Neural differentiation

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Formation of the neural plate is an intricate process in early mammalian embryonic development mediated by cells of the inner cell mass and involving a series of steps, including development of the epiblast. Here, we report on the creation of an embryonic stem (ES) cell-based system to isolate and identify neural induction intermediates with characteristics of epiblast cells and neural plate. We demonstrate that neural commitment requires prior differentiation of ES cells into epiblast cells that are indistinguishable from those derived from natural embryos. We also demonstrate that epiblast cells can be isolated and cultured as epiblast stem cell lines. Fgf signaling is shown to be required for the differentiation of ES cells into these epiblast cells. Fgf2, widely used for maintenance of both human ES cells and epiblast stem cells, inhibits formation of early neural cells by epiblast intermediates in a dose-dependent manner and is sufficient to promote transient self-renewal of epiblast stem cells. In contrast, Fgf8, the endogenous embryonic neural inducer, fails to promote epiblast self-renewal, but rather promotes more homogenous neural induction with transient self-renewal of early neural cells. Removal of Fgf signaling entirely from epiblast cells promotes rapid neural induction and subsequent neurogenesis. We conclude that Fgf signaling plays different roles during the differentiation of ES cells, with an initial requirement in epiblast formation and a subsequent role in self-renewal. Fgf2 and Fgf8 thus stimulate self-renewal in different cell types. STEM CELLS 2010;28:1772–1781


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Stem cells are found in many mammalian tissues and organs. These cells exhibit divergent abilities for self-renewal and differentiation and are required for organ homeostasis and cell replacement when body function is challenged, for example, during injury or disease states. Much attention has been paid to the biology of stem cells, which is far from being understood, and their use as a potential source in transplantation therapies and models for disease as well as drug discovery. Several types of stem cells have been isolated and some have even been cultured. Of the latter, pluripotent stem cells hold the greatest developmental potential and is most promising for therapeutic applications. A detailed understanding of the mechanisms controlling self-renewal and differentiation of these cells is a prerequisite to realizing their full potential. Mammalian pluripotent stem cells are located in the inner cell mass (ICM) of blastocysts and have been used as a source of embryonic stem (ES) cell lines for both mouse and human models. As ES cells have many features in common with their in vivo counterparts, the process of embryogenesis can be used as a guide to understand and manipulate ES cell commitment in vitro.

The first step in neurulation is the formation of the neural plate, which becomes apparent in the early embryo by a thickening of the ectoderm opposite to the primitive streak. Neural plate formation involves at least two distinct steps. First, the ICM forms two epithelial layers, that is, the epiblast and the hypoblast. Although cells of the epiblast and ICM are considered to retain features of pluripotential cells, they have distinctly different properties. In terms of function, the ICM is capable of forming chimeras when aggregated with early embryos or injected into the blastocoele, whereas postimplantation epiblast cells form chimeras only very rarely [1, 2]. Expression patterns also differ between the epiblast and ICM. For example, the abundant expression of Fgf5 and the lack of expression of Rex1 distinguish epiblast from ICM cells [3]. Finally, although expression of Pou5f1, also known Oct4, is necessary for both cells of the ICM and epiblast, it is driven by two different enhancers in ICM/ES cells and epiblast cells [4]. In both ICM and ES cells, Oct4 expression is driven primarily from the distal enhancer. However, in postimplantation epiblast cells, Oct4 expression is driven by the proximal enhancer [4]. Recent evidence has also shown that Fgf4 plays a role in epiblast formation through the ERK1/2 signaling pathway [5].

Following its course during embryonic development, the epiblast will subsequently undergo gastrulation to form the three germ lineages: ectoderm, mesoderm, and endoderm. The ectoderm overlies the notochord and develops into the neural plate or neuroectoderm, which will eventually give rise to the central nervous system. Coincident with cell differentiation to form the neural plate is loss of pluripotency and downregulation of Oct4 expression [4]. In the neural plate, upregulation of Sip1 (also known as Zeb2 or Zfhx1b) expression results in the inability of the cells to give rise to mesoderm, and overexpression of Sip1 in human ES cells is sufficient to limit the mesendodermal-inducing effects of Activin-Nodal signaling [6, 7]. As Sip1 expression also results in direct repression of E-cadherin expression, loss of E-cadherin is among the earliest events in neural cell lineage commitment [8–10]. Of significance is the finding that the most commonly used markers for neural induction, Pax6 and Sox1, begin to be highly expressed at somitogenesis, which occurs at least half a day after neural plate formation [11–13].

Fgf signaling plays a significant role in neural induction. Treatment of cells with inhibitors of Fgf receptors results in clear defects in neural induction [14]. Of the many different Fgfs, research has primarily focused on Fgf8 as the endogenous neural inducer. Fgf8 is expressed prior to primitive streak formation in the posterior proximal epiblast as well as in the visceral endoderm [15–17]. Removal of Fgf8 in transgenic mouse embryos results in severe gastrulation defects, such as aberrant formation and patterning of the neuroectoderm [18]. In chick embryos, exposure of Fgf8 for a period of 2 hours induces expression of early preneural genes, whereas exposure of Fgf8 for 5 hours enables the epiblast to respond to bone morphogenetic protein (BMP) antagonist signals, which then stabilize the expression of the neural marker Sox3 [19]. A search for genes differentially expressed during this time frame of Fgf8 exposure, that is, between 2 and 5 hours, led to the identification of Churchill [6]. Churchill interacts with Sip1 to promote neural commitment through inhibition of Activin-Nodal signaling, thereby inhibiting mesendodermal specification [6, 20]. Therefore, Fgf8 represents the best candidate for an endogenous inducer of neural character [19].

Exposure of ES cells to specific culture conditions results in the rapid induction of neural cells [21, 22]. Expression analysis of differentiating cultures suggests that ES cell differentiation is a multistep process directly comparable with embryogenesis [23, 24]. Transient Fgf5 expression shortly after initiation of differentiation suggests that ES cells undergo epiblast formation [3, 23-26]. As Fgf5 expression declines, neural commitment ensues, as marked by Sox1 expression [23, 24]. These earliest neural cells are distinct from neural stem cells [24, 27, 28]. However, although ES cell neural induction is a rapid process that follows embryogenesis, it also occurs stochastically, with some cells committing earlier than others [23]. Therefore, cultures of differentiating ES cells generally contain cell types of all degrees/levels of differentiation at any one time.

It is well-established that Fgf signaling is essential for the self-renewal of epiblast stem cells and human ES cells. Indeed, Fgf2 is widely used in combination with either feeder cells or other growth factors for their epiblast stem cell and ES cell maintenance in culture. Recently, it has been found that inhibition of Fgf signaling, in combination with other factors, is capable of promoting mouse ES cell self-renewal [29]. However, the role of Fgf signaling during pluripotent cell commitment to neural induction is rather controversial. One study using mouse ES cells and defined culture conditions reported that treatment with an Fgf inhibitor resulted in a significant decrease in the number of Sox1-positive neural cells [22]. However, another publication demonstrated that removal of either Fgf4 or Erk2 from mouse ES cells resulted in defects in somatic lineage commitment, with the cells becoming stuck in an intermediate stage between the ES and epiblast cell states [5]. Recent work from our group has shown that inhibition of Fgf increased the efficiency of neural differentiation of both mouse and human pluripotent stem cells [30]. In contrast, one recent report showed that inhibition of Fgf signaling blocked the commitment of mouse and human pluripotent stem cells to neural lineage [31]. Another recent publication found that although Fgf signaling significantly enhanced neural specification, it is not necessary [32]. One possible reason for these variable reports is that Fgfs may have different effects at different time points during early embryonic development. Therefore, a culture system capable of resolving the earliest stages of neural commitment would help clarify the role played by Fgf signaling in neural commitment. Of significant interest is a comparison of the effects of Fgf2, which promotes epiblast stem cell self-renewal, with those of Fgf8, which is considered the endogenous neural inducer.

Here, we report on the creation of an ES cell-based system to investigate the commitment of cells along the neural cell lineage. No method currently available exists to simultaneously isolate multiple neural induction intermediates that support either continuous cell culture or molecular characterization. Other published methods for analyzing neural differentiation in vitro rely on techniques that analyze a bulk heterogeneous culture (such as reverse transcription polymerase chain reaction [RT-PCR]) or otherwise isolate one particular subpopulation (such as Sox1-green fluorescent protein [GFP]-positive cells) [22]. Here, we report on the creation of a system to identify, isolate, and characterize multiple neural induction intermediates, enabling the study of ES cell neural differentiation in a very precise manner. We demonstrate that epiblast intermediate cells formed during ES cell differentiation can be propagated as epiblast stem cells that are indistinguishable from epiblast stem cells derived from embryos. We show that Fgf signaling is required for ES cell differentiation into epiblast cells and that Fgf2 blocks neural commitment of epiblast intermediate cells in a dose-dependent manner. Exposure of epiblast stem cells to Fgf2 is shown to lead to a transient blockade of differentiation, whereas inhibition of Fgf signaling promotes the rapid differentiation into neural cells as well as some non-neural cells. In contrast, Fgf8 is shown to promote a more homogeneous neural induction mediated by epiblast cells as well as a transient self-renewal of early neural cells. Therefore, we conclude that Fgf2 and Fgf8 exert different roles in the self-renewal and neural commitment of pluripotent stem cells.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

ES Cells Culture

ES cells were cultured in the absence of feeder cells on gelatinized plates in ES cell medium, which consisted of knockout Dulbecco's modified Eagle's medium (DMEM), 16% knockout serum replacement, 4% fetal calf serum, 1% nonessential amino acids, 1% penicillin/streptomycin/glutamine (Invitrogen, Carlsbad, California, http://www.invitrogen.com), 1% β-mercaptoethanol (fresh 100× stock solution made by diluting 7 μl of 14.3 M β-mercaptoethanol in 10 ml phosphate buffered saline (PBS)), and 2,000 units/ml of leukemia inhibitory factor.

Monolayer differentiation was performed by modifying the original procedure described by Ying et al. [22]. Briefly, cells were trypsinized and plated at low density in ES cell medium. The next day, the ES cell medium was replaced by N2B27 medium, which consisted of 50% neurobasal medium with N2, 50% DMEM/F12 with B27 supplemented with 0.15% bovine serum albumin, and 1% β-mercaptoethanol (fresh 100× stock solution made by diluting 7 μl of 14.3 M β-mercaptoethanol in 10 ml PBS). Fgf2 and Fgf8 (Peprotech, Rocky Hill, New Jersey, http://www.peprotech.com) were added at the indicated concentrations starting at day 0 of differentiation. SU5402 (Calbiochem, San Diego, California, http://www.calbiochem.com) and PD0325901 (Axon Medchem, Groningen, The Netherlands, http://www.axonmedchem.com) were used at concentrations of 10 and 1 μM, respectively. Induction into endodermal and mesodermal lineages was performed by adding 10 ng/ml of Activin A (Peprotech, Rocky Hill, New Jersey, http://www.peprotech.com) and of BMP4 (Peprotech, Rocky Hill, New Jersey, http://www.peprotech.com), respectively.

Epiblast Cell Culture

Embryo-derived epiblast stem cells were obtained from embryonic day 5.5 (E5.5)-stage embryos. After whole embryos were placed in culture, epiblast colonies were manually picked and expanded. Epiblast medium, which was a mixture of DMEM/F12, 20% knockout serum replacement, 1% nonessential amino acids, 1% penicillin/streptomycin/glutamine, 1% β-mercaptoethanol (fresh 100× stock solution made by diluting 7 μl of 14.3 M β-mercaptoethanol in 10 ml PBS), and 5 ng/ml of Fgf2, was conditioned overnight by irradiated mouse embryo fibroblasts. After adding an additional 5 ng/ml of Fgf2, the medium was used to feed epiblast cells. Epiblast cells were cultured on wells coated with fetal calf serum (FCS) for 20 minutes at room temperature and washed twice with PBS. For differentiation, epiblast stem cells differentiated in N2B27 medium supplemented as indicated.

Fluorescence-Activated Cell Sorting Analysis

Cells were dissociated with trypsin and immunostained with anti-SSEA1-PE (R&D Systems, Minneapolis, Minnesota, http://www.rndsystems.com) and anti-E-cadherin (ECCD2; Calbiochem). Allophycocyanin-conjugated donkey anti-rat secondary antibody (Jackson ImmunoResearch) was used for detection of ECCD2. Flow cytometric analysis and cell sorting were performed on a FACSAria cell sorter (BD Biosciences, San Jose, California, http://www.bdbiosciences.com/home.jsp). Data analysis was done using FlowJo software (Treestar, Ashland, Oregon, http://www.treestar.com).

Real-Time RT-PCR

Total RNA was extracted using the RNeasy Mini Kit (QIA-Qiagen, Hilden, Germany, http://www.qiagen.com) and digested on column with DNase. RNA quality and concentration were determined using the Bioanalyzer RNA 6000 (Agilent Technologies, Santa Clara, California, http://www.agilent.com). Complementary DNA synthesis was performed with the High Capacity cDNA Archive Kit (Applied Biosystems, Carlsbad, California, http://www.appliedbiosystems.com) following the manufacturer's instructions but with a scaling down of the reaction volume to 20 μl. Transcript levels were determined using the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Carlsbad, California, http://www.appliedbiosystems.com) and the ready-to-use 5′-nuclease Assays-on-Demand. Zic1, Zic3, FoxD3, Sox10, and all reactions depicted in Figure 7B were performed with SYBR green. Primer sequences for SYBR green reactions are available on request.

Immunostaining

Cryosections and cells were fixed with paraformaldehyde and permeabilized with Triton X-100. After blocking, cells were stained overnight as indicated. Antibodies raised against Oct4, SSEA1, and Nanog were obtained from Santa Cruz, Developmental Studies Hybridoma Bank, and Cosmo Bio, respectively. Secondary staining with AlexaFluor-conjugated antibodies (Invitrogen, Carlsbad, California, http://www.invitrogen.com) was performed the following day.

Microarray Data

Total RNA was extracted using the RNeasy Mini Kit (QIAGEN, Hilden, Germany, http://www.qiagen.com) and digested on column with DNase. RNA was subjected to DNA microarray analysis essentially as described before [33–36]. Briefly, biotin-labeled cRNA was obtained from 3 μg of total RNA with the GeneChip One-Cycle labeling kit (Affymetrix, Santa Clara, California, http://www.affymetrix.com). Fifteen micrograms of cRNA was fragmented and hybridized to Affymetrix 430 2.0 GeneChip arrays (Affymetrix, Santa Clara, California, http://www.affymetrix.com) at 45°C for 16 hours. DNA chips were washed, stained, and scanned using an Affymetrix Fluidics device and GCS3000 scanner, and the images obtained were analyzed using the GCOS software. Normalization was calculated with RMA algorithm implemented in BioConductor [37]. Data postprocessing and graphics were done with in-house developed functions in Matlab.

Morula Aggregations

Cells were aggregated and cultured with denuded postcompacted eight-cell stage mouse embryos as reported with a slight modification [38]. Briefly, eight-cell stage embryos were flushed from mice ([C57BL/6 × C3H] F1 female × CD1 male) at 2.5 days postcoitum (dpc) and placed in M2 medium. Clumps of loosely connected cells (10–20 cells/clump) obtained after a brief period of trypsinization were selected and transferred into microdrops of potassium simplex optimized medium with 10% FCS under mineral oil; each clump was placed in a depression in the microdrop. At the same time, batches of 30–40 embryos were briefly incubated in acidified Tyrode's solution until dissolution of their zona pellucida. A single embryo was placed on top of each clump. All aggregates were assembled in this manner and cultured overnight at 37°C, 5% CO2. After 24 hours in culture, the majority of aggregates had formed blastocysts.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

A System for Dissecting Neural Differentiation

Induction of the neural plate from the ICM is marked by two major transitions during embryonic development. The first step involves the formation of epiblast cells, which will subsequently give rise to all three germ cell layers. The differential use of Oct4 enhancers distinguishes ICM/ES cells from epiblast cells [4]. An 18-kb genomic DNA fragment, referred to as GOF18, is capable of reproducing the native or endogenous expression pattern of Oct4 [4]. Deletion of the Oct4 proximal enhancer results in a significant reduction of Oct4 expression in the epiblast, but has relatively little effect on Oct4 activity in ICM and ES cells. We have previously utilized an ES cell line containing a transgenic GFP construct driven by a modified Oct4 promoter (OG2) in which the proximal enhancer had been deleted (Fig. 1A) [39–41]. Epiblast stem cells were derived from OG2 embryos, and their lack of GFP expression confirmed the specificity of the OG2 promoter (Fig. 1B).

thumbnail image

Figure 1. Strategy for dissecting neural induction. (A): Schematic of OG2 transgenic GFP construct. (B): Relative GFP levels in ES cells and epiblast stem cells by fluorescence-activated cell sorting analysis. (C): SSEA1 staining of neural plate from E7.75-stage mouse embryo. (D): Diagram depicting strategy used to isolate different neural induction intermediates. Abbreviations: DE, distal enhancer; ES cell, embryonic stem cell; GFP, green fluorescent protein; PE, proximal enhancer; PP, proximal promoter; SSEA1, stage specific embryonic antigen 1.

Download figure to PowerPoint

The second major transition in neural cell induction is the actual formation of the neural plate. Loss of E-cadherin expression has long been recognized as the earliest step in neural cell commitment [8]. This loss of expression, which is a result of direct repression by Sip1, is necessary for neural tube closure [10]. Sip1 plays an essential role in neural cell commitment, as loss of a Sip1 activator results in differentiation into the mesodermal lineage in the neural plate of chick embryos [6]. E-cadherin is expressed in the neural tube of Sip1-deficient embryos, suggesting that E-cadherin can be used as a valid marker for Sip1 activity and neural cell commitment [10]. SSEA1 is specifically expressed in the embryonic ectoderm and visceral endoderm [42]. However, E-cadherin expression is downregulated within the embryonic ectoderm following neural commitment, whereas it is maintained in the visceral endoderm. We briefly corroborated this expression pattern using immunostaining (Fig. 1C, 1D). Therefore, we used the combination of GFP, SSEA1, and E-cadherin to isolate ES cells, epiblast cells, and cells committed to the neural plate (Fig. 1D).

Sequential Differentiation in ES Cells into Neural Commitment

Having established a system to discriminate the earliest events in neural cell commitment, we next set about monitoring ES cell differentiation (Fig. 2A). Although various methods exist to direct the differentiation of ES cells into the neural cell lineage, we specifically selected to use the serum-free monolayer method established by Ying et al. [22]. This is the best characterized system to date and appears to involve signaling pathways that accurately mirror those in vivo. Within 3 days of initiation of differentiation, onset of GFP downregulation was apparent (Fig. 2A). By day 4, a distinct population of cells had formed. These cells were SSEA1+, E-cadherin+, but GFP−, suggesting that the ES cells had formed epiblast cells. By day 5, downregulation of E-cadherin was first observed, marking the beginning of neural induction. By day 9, the expression of all three markers was lost, indicating further differentiation, accompanied by the formation of neurons (Supporting Information Fig. 1). The sequential loss of GFP and E-cadherin expression in SSEA1+ cells was an indication that ES cells were following the same sequence of differentiation steps along the neural cell lineage as those described for the developing embryo. It is also noteworthy that we have never observed loss of E-cadherin expression concomitant with loss of GFP expression, suggesting that ES cells require the formation of epiblast cells prior to undergoing final neural cell commitment.

thumbnail image

Figure 2. Expression profile of neural induction intermediates. (A): Scatter plot shows GFP versus SSEA1 staining by flow cytometry. Histogram shows E-cadherin levels in the GFP+/SSEA1+ quadrant (outlined in black) and GFP−/SSEA1+ quadrant (filled in gray) of the scatter plot directly above. (B): Real-time quantitative reverse transcription polymerase chain reaction results for the indicated genes in ES, EpiInt, and NInt cells. Abbreviations: GFP, green fluorescent protein; ES cell, embryonic stem cell; SSEA1, stage specific embryonic antigen 1; EpiInt, epiblast intermediate cells; NInt, neuroectodermal intermediate cells.

Download figure to PowerPoint

Intermediate Stages Have Characteristics of Epiblast and Neuroectoderm

Expression of genetic markers of the neural induction intermediates confirmed their identity as epiblast and neuroectoderm (Fig. 2B). Epiblast stem cells can be clearly distinguished from ES cells by their abundant expression of Fgf5 and comparatively low expression of Rex1 [1, 2]. Consistent with this expression pattern, GFP−/SSEA1+/E-cadherin+ (hereafter referred to as epiblast intermediate or EpiInt) cells isolated on day 4 exhibited >200-fold higher expression of Fgf5 but >17-fold lower expression of Rex1 compared with ES cells (Fig. 2B). The lack of OG2 activity (as assayed by GFP expression), along with the Fgf5 and Rex1 expression data, suggests that the isolated cells had undergone epiblast cell fate commitment.

Cells of the neuroectoderm express a vastly different set of genes compared with epiblast and ES cells. Whereas concomitant expression of Oct4 and Nanog is observed in epiblast stem cells and ES cells, and appears to be required for self-renewal, significantly lower expression is observed in neuroectodermal cells. Instead, expression of Sip1, along with that of other markers of neural commitment, such as Pax6 and Sox1, and loss of E-cadherin expression are observed. Expression profiling of GFP−/SSEA1+/E-cadherin- (hereafter referred to as neuroectodermal intermediate [NInt]) cells revealed an expression pattern characterized by the downregulation of the pluripotency-related markers Oct4 and Nanog and the upregulation of the neural cell commitment genes Pax6, Sox1, and Sip1 (Fig. 2B). Consistent with cell sorting for lower levels of E-cadherin, expression of E-cadherin was approximately 10-fold lower in NInt cells compared with ES cells (Supporting Information Fig. 2). Although NInt cells express markers for the neural cell lineage, they are closely related to EpiInt cells, as shown by microarray profiling (Fig. 3A, 3B). In contrast, NInt cells are very different from neural stem cells (Fig. 3A, 3B). Therefore, NInt cells represent the earliest known cell type to have committed to the neural cell lineage following in vitro ES cell differentiation.

thumbnail image

Figure 3. Properties of EpiInt and NInt cells. (A): Principal component analysis and (B) sample clustering results. ES and ES2 cells are two different cultures of OG2 ES cells. (C): Embryo contribution of OG2-ROSA26 ES cell-derived epiblast cells after blastocyst injection. Abbreviations: EpiInt, epiblast intermediate cells; ES cell, embryonic stem cell; NInt, neuroectodermal intermediate cells; NSC, neural stem cell.

Download figure to PowerPoint

Epiblast stem cells and ES cells differ in their ability to form chimeras when injected into blastocysts [1, 2]. ES cells readily incorporate into the ICM and subsequently differentiate into all definitive embryonic cell lineages; however, epiblast stem cells accomplish these two events inefficiently. In addition, although chimeras have been isolated from epiblast-injected blastocysts, their formation is a rare event [2]. The underlying cause is an apparent failure of the cells to incorporate into the ICM. To test whether our epiblast intermediate cells functionally resemble epiblast stem cells, we introduced either ES cells or epiblast intermediate cells, derived from ROSA26 mice, into diploid blastocysts. Whereas ES cells readily contributed to the development of embryos, there was no evidence that epiblast intermediate cells contributed as well (Fig. 3C). This failure to form chimeras demonstrates that EpiInt cells derived from day 4-differentiated ES cells are functionally similar to embryo-derived epiblast stem cells.

Although ES and epiblast cells differ in their ability to form chimeras, they both exhibit pluripotent differentiation potential. Unlike neuroectodermal cells, epiblast stem cells are readily capable of differentiating into both endoderm and mesoderm. To test whether EpiInt cells and NInt cells differ in their ability to form endoderm, we exposed both cell types to Activin A for 7 days. We found that EpiInt cells had formed many Sox17-positive endodermal cells, but few TuJ-positive neurons (Fig. 4A). In contrast, NInt cells had undergone extensive neurogenesis, with little endoderm formation. The few Sox17-positive cells found in the NInt sample are likely due to contamination from the fluorescence-activated cell sorting (FACS) isolation. We also treated the different cell populations with BMP4 for 4 hours to induce mesodermal gene expression and found increased T, also known Brachyury, expression in both ES and EpiInt cells but decreased expression in NInt cells (Supporting Information Fig. 3). These data demonstrate that EpiInt and NInt cells have a differentiation potential similar to that of embryonic epiblast stem cells and neuroectodermal cells, respectively.

thumbnail image

Figure 4. Characterization of NInt cells. (A): Cellular differentiation into endodermal, which is Sox17 (green) positive, versus neuronal, which is TuJ1 (red) positive, lineages after 7 days of Activin A stimulation of either EpiInt or NInt cells. (B): Quantitative reverse transcription polymerase chain reaction results as indicated. Abbreviations: EpiInt, epiblast intermediate cells; ES cell, embryonic stem cell; NInt, neuroectodermal intermediate cells.

Download figure to PowerPoint

To determine the identity of NInt cells in greater detail, we performed quantitative RT-PCR with markers specific for early neural plate and neural crest. Neuroectoderm is initially specified from pre-existing epiblast cells through the inhibition of transforming growth factor-β signals. At this time, several characteristic early neural markers are known to be expressed, including Pax3, Pax7, Zic1, and Zic3 [43, 44]. NInt cells were found to significantly upregulate each of these markers with the exception of Zic3, which is already highly expressed in ES cells (Fig. 4B) [45]. Several of these markers are also expressed in neural crest cells. However, NInt cells do not upregulate the neural crest markers FoxD3 and Sox10 (Fig. 4B). Therefore, we conclude that NInt cells most closely resemble neural plate cells at approximately the time of initial specification.

Epiblast Stem Cell Lines Can Be Derived from Epiblast Intermediate Cells

Epiblast stem cells have successfully been derived from early mammalian embryos [1, 2]. To provide further evidence that our EpiInt cells are functionally equivalent to embryonic epiblast cells, we isolated epiblast stem cell lines from sorted EpiInt cells. Clonal epiblast stem cell lines could be produced at a rate of approximately 10%. As EpiInt cells were isolated from differentiated ES cells, the EpiInt-derived epiblast stem cell lines were labeled as ES-derived in the figures. The epiblast stem cell lines were stable and capable of being propagated for more than 100 population doublings. These cells exhibited an epiblast expression profile, stained positive for both Oct4 and Nanog, and were unable to form chimeras (Fig. 5 and Supporting Information Fig. 4). Microarray profiling demonstrated that the expression profile of EpiInt-derived epiblast stem cells is very similar to that of embryonic epiblast-derived epiblast stem cells (Supporting Information Fig. 5). On the basis of these data, we conclude that EpiInt cells are functionally equivalent to epiblast cells, that is, are indeed epiblast cells.

thumbnail image

Figure 5. Epiblast stem cells (EpiSCs) derived from EpiInt cells. EpiInt-derived epiblast stem cell lines are labeled as ES-derived (ESd EpiS cells). Epiblast stem cells derived from postimplantation embryonic epiblast cells are labeled as EPId EpiS cells. (A): Real-time quantitative reverse transcription polymerase chain reaction results for the indicated genes. (B): Phase and GFP fluorescence images of ES cell-derived epiblast stem cells. (C): Oct4 and Nanog immunostaining of ES cell-derived epiblast stem cells. Abbreviations: EpiS cell, epiblast stem cell; ES cell, embryonic stem cell; GFP, green fluorescent protein.

Download figure to PowerPoint

Role of Fgf Signaling in Epiblast Formation and Neural Induction

The role of Fgf signaling in mammalian neural induction is controversial. We sought to use our new method of dissecting neural induction to elucidate the possible function of Fgfs during ES cell differentiation into the neural lineage. Quantitative RT-PCR revealed that ES cells and epiblast stem cells express all four Fgf receptors (Fig. 6A). Although Fgfr4 is expressed at lower levels in epiblast stem cells, it is still detectable. A dose-dependent response on the proportion of EpiInt cells was observed when differentiating ES cells were cultured with Fgf2 (Fig. 6B). The effect of Fgf8 was directly comparable with the lowest concentration of Fgf2 tested (Fig. 6B), which is consistent with the weaker stimulation of the Fgf receptors by Fgf8 compared with Fgf2 [46]. In contrast, when Fgf signaling was removed, formation of epiblast cells from ES cells was inhibited (Fig. 6C). Therefore, we conclude that inhibition of Fgf signaling promotes ES cell self-renewal, a finding consistent with those of previously published studies [5, 29].

thumbnail image

Figure 6. Fgf signaling effects on differentiating ES cells. (A): Quantitative reverse transcription polymerase chain reaction results for Fgfr expression in ES and epiblast stem cells. (B): E-cadherin fluorescence-activated cell sorting (FACS) histogram overlay for SSEA1+/green fluorescent protein negative (GFP−) cells on day 7 of differentiation of ES cells under the indicated conditions. (C): FACS scatter plots for day 3 of ES cell differentiation under the indicated conditions. Quadrant gating shows GFP negative and SSEA1 negative populations. The percent EpiInt is taken from the first quadrant (GFP-negative, SSEA1-positive). The oval gate is for pure ES cells (GFP-high). Abbreviations: ES cell, embryonic stem cell; FGF, fibroblast growth factor; PD, PD0325901; SU, SU5402.

Download figure to PowerPoint

We also sought to determine the effects of Fgf signaling directly in epiblast stem cell cultures. Fgf2 is a common supplement to epiblast stem (EpiS) cell medium, in combination with other factors such as Activin A, to promote self-renewal. When EpiS (EPId) cells were cultured in defined medium, a significantly higher proportion of Fgf2-stimulated cells were found to maintain both SSEA1 and E-cadherin expression, indicative of transient self-renewal (Fig. 7A). In contrast, Fgf8-treated EpiS (EPId) cells had homogeneously downregulated E-cadherin, indicating neural commitment. However, unlike the control cells, Fgf8-treated cells maintained a much higher level of SSEA1, signifying that Fgf8 stimulated the transient self-renewal of NInt cells (Fig. 7A). Furthermore, the population of SSEA1−/E-cadherin+ cells was reduced in the Fgf8-treated cultures, indicating less non-neural differentiation and, consequently, a more homogeneous neural commitment.

thumbnail image

Figure 7. Fgf signaling effects on differentiating embryo-derived epiblast stem (EpiS) cells. (A): fluorescence-activated cell sorting (FACS) plots of embryo-derived EpiS cells differentiated as a monolayer under the conditions indicated. (B): Quantitative reverse transcription polymerase chain reaction results of embryo-derived EpiS cells as indicated. Expression of the pluripotency markers (Oct4, Nanog), a neural progenitor marker (Sox1), neurogenesis markers (Tubb3, Dcx), and a non-neural marker (Cdx2) is assayed. Abbreviations: SSEA1, stage specific embryonic antigen 1; PD, PD0325901; SU, SU5402.

Download figure to PowerPoint

To corroborate the FACS data, quantitative RT-PCR was performed on EpiS (EPId) cells after culturing with either Fgfs or with inhibitors of Fgf signaling. Consistent with the FACS results, Fgf2 was found to maintain the expression of pluripotent markers and inhibit the expression of differentiation markers (Fig. 7B). Inhibition of Fgfr signaling caused rapid neurogenesis and non-neural differentiation, as indicated by upregulation of Cdx2, which is expressed in the trophoblast and midgut endodermal lineages (Fig. 7B) [47]. Consistent with the FACS results, Fgf8-treated cells were found to differentiate into the neural lineage, yet failed to undergo neurogenesis (Fig. 7B). Cdx2 expression was considerably reduced by Fgf8 exposure, indicative of more homogeneous neural commitment compared with inhibition of Fgfr signaling. Inhibition of mitogen-activated protein kinase/extracellular signal-regulated kinase (MEK), which is a component downstream of the Fgf signaling cascade, resulted in very rapid neurogenesis, similar to Fgfr inhibition, albeit with more neural progenitors and less non-neural differentiation (Fig. 7B). Addition of Fgf2 or Fgf8 during MEK inhibition increased cellular differentiation into non-neural lineages, indicating that the effects of Fgf signaling are complex, and not simply due to MEK stimulation (Fig. 7B). These results demonstrate that Fgf2 and Fgf8 stimulate the transient self-renewal of EpiS (EPId) cells and NInt cells, respectively. We therefore conclude that Fgf signaling is seemingly dispensable for neural commitment in vitro, consistent with a recently published study [30].

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

The differentiation of pluripotent stem cells into cells of the neural lineage comprises a multistep process in vivo. Here, we demonstrate that mouse ES cells follow this same step-wise progression in vitro. Furthermore, we have developed a system to isolate the different cellular stages during the neural induction process, and thereby clarifying the role played by Fgf signaling.

Fgf signaling has divergent effects on the self-renewal of mouse and human ES cells. Mouse ES cells can be maintained using a combination of Fgf/Erk inhibition and WNT/β-catenin signaling [29]. In contrast, Fgf/Erk signaling promotes the self-renewal of human ES cells, and Fgf2 is widely used in human ES cell cultivation [48]. Although Fgf stimulation has pleiotropic effects, one significant observation is that Fgf2 stimulates Activin A production by mouse embryonic fibroblasts, which are generally used as feeder cells [49]. Fgf2 and Activin A under defined conditions are sufficient for human ES cell self-renewal [50].

This significant difference between mouse and human ES cells prompted the hypothesis that they represented fundamentally different cell types. Two different groups simultaneously reported that epiblast stem cells could be derived from postimplantation mouse embryos when cultivated under human ES cell-like conditions [1, 2]. These cells were functionally pluripotent, but formed chimeras with a significantly lower frequency. Recent work has indicated that mouse ES cell-like human pluripotent cells can be derived under conditions similar to those of mouse ES cells [51]. These observations have prompted a new understanding of pluripotency. Mouse ES cell-like pluripotency is generally referred to as “ground state” or “naïve” pluripotency, whereas epiblast stem cell-like pluripotency is referred to as “primed” pluripotency. Therefore, Fgf signaling has opposing effects on ground state and primed pluripotent stem cells.

Dissecting neural differentiation has enabled the study of Fgf signaling during ES cell neural commitment. Of significant interest are the effects of Fgf2, a factor widely used in the cultivation of pluripotent cells, and Fgf8, considered to be the endogenous neural inducer. To date, different publications have reported divergent results. Here, we demonstrate that Fgf2 is necessary for the formation of epiblast cells from ES cells but that it inhibits the neural commitment of epiblast cells. Furthermore, this inhibition occurs in a dose-dependent manner. In contrast to Fgf2, Fgf8 is unable to stimulate epiblast self-renewal, but results in more homogenous neural commitment and limited self-renewal of neuroectodermal progenitors. Previous work has demonstrated that Fgf8 has a significantly reduced ability, compared with Fgf2, to stimulate all tested Fgf receptors [46]. Our results are consistent with these data and suggest that Fgf8 functions as a neural inducer in vivo, in part, by failing to stimulate epiblast self-renewal and enabling cellular differentiation.

However, in contrast to the demonstrated requirement for Fgf signaling during in vivo neural induction [14], we have consistently observed neural induction and even neurogenesis in the absence of Fgf signaling in vitro. This can be explained by the effects of Fgf stimulation on Activin and Bmp signaling. In vivo, Fgf8 exposure induces Churchill, which cooperates with Sip1 to repress Activin-induced Smad signaling necessary for cellular differentiation into the mesendoderm lineage. Fgf signaling in vivo is also thought to promote neural commitment through the inhibition of Bmp signaling via phosphorylation of the Smad linker region [52]. However, in vitro, exogenous Activin and Bmp signals are missing. This is because pluripotent cells are cultivated in chemically defined basal medium and autocrine Activin and BMP signals are reduced through the cultivation of cells at a very low density. We argue that the difference in the requirement for Fgf signaling in vitro and in vivo can be attributed to the culture conditions used for the differentiation of cells into the neural lineage.

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

We successfully isolated epiblast cells and neural plate-like cells from cultures of differentiating ES cells. We demonstrate that Fgf2 promoted self-renewal of epiblast cells, but Fgf8 promoted more homogenous neural induction and transient self-renewal of early neural cells. Inhibition of Fgf signaling promoted ES cell self-renewal.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

This work was supported by Grants SCHO340/5-1 and ZE 432/5-1 from the Deutsche Forschungsgemeinschaft (DFG) Priority Program: Pluripotency and Cellular Reprogramming (SPP 1356).

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
STEM_498_sm_SuppFig1.tif22791KSupporting Information Figure 1. Differentiation of ES cells into neuronal lineage. Neurons (TuJ1-positive) formed from ES cells after 9 days in N2B27 medium.
STEM_498_sm_SuppFig2.tif325KSupporting Information Figure 2. Relative E-cadherin expression. E-cadherin expression levels in ES cells and their derivatives.
STEM_498_sm_SuppFig3.tif319KSupporting Information Figure 3. Relative competence for Brachyury induction by BMP. Relative expression levels of Brachyury in the presence or absence of BMP stimulation for 4 hours for the indicated cell types.
STEM_498_sm_SuppFig4.tif19382KSupporting Information Figure 4. Chimera formation potential of ES cell–derived EpiSCs. Blastocysts formed from ES cells or ES cell–derived EpiSCs aggregated with morulas.
STEM_498_sm_SuppFig5.tif2444KSupporting Information Figure 5. Microarray comparison of EpiSCs derived from embryonic epiblast or EpiInt cells. Scatter plot depicting the expression profile of epiblast stem cells derived from embryonic epiblast (EPId) versus that of differentiated ES cell–derived (ESd). Number of similar expressed genes: 44,873 (99.5℅); Correlation coefficient R; 0.985; Number of different expressed genes: 228 (0.506℅) (using a 2 fold change threshold in log2 scale).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.