Embryonic Stem Cell-Derived Neural Progenitors Display Temporal Restriction to Neural Patterning


  • Isabelle A. Bouhon,

    1. Neurobiology Programme, The Babraham Institute, Babraham, Cambridge, United Kingdom
    Search for more papers by this author
  • Alexis Joannides,

    1. Cambridge Centre for Brain Repair, University of Cambridge, Cambridge, United Kingdom
    Search for more papers by this author
  • Hidemasa Kato,

    1. Neurobiology Programme, The Babraham Institute, Babraham, Cambridge, United Kingdom
    Search for more papers by this author
  • Siddharthan Chandran MRCP Ph.D.,

    Corresponding author
    1. Cambridge Centre for Brain Repair, University of Cambridge, Cambridge, United Kingdom
    • Cambridge Centre for Brain Repair, University of Cambridge, Cambridge CB2 2PY, U.K., Telephone: +44 1223331160; Fax: +44 1223331174
    Search for more papers by this author
  • Nicholas D. Allen Ph.D.

    Corresponding author
    1. Neurobiology Programme, The Babraham Institute, Babraham, Cambridge, United Kingdom
    2. School of Biosciences, Cardiff University, Cardiff, United Kingdom
    • School of Biosciences, Cardiff University, Museum Avenue, Cardiff, CF10 3US, U.K. Telephone: +44 29 2087 6196; Fax: +44 29 2087 6328
    Search for more papers by this author


Neural stem cells have considerable therapeutic potential because of their ability to generate defined neuronal cell types for use in drug screening studies or cell-based therapies for neurodegenerative diseases. In this study, we differentiate mouse embryonic stem cells to neural progenitors with an initial forebrain identity in a defined system that enables systematic manipulation to generate more caudal fates, including motoneurons. We demonstrate that the ability to pattern embryonic stem cell-derived neural progenitors is temporally restricted and show that the loss of responsiveness to morphogenetic cues correlates with constitutive expression of the basic helix-loop-helix transcription factors Olig2 and Mash1, epidermal growth factor receptor, and vimentin and parallels the onset of gliogenesis. We provide evidence for two temporal classes of embryonic stem cell-derived putative radial glia that coincide with a transition from neurogenesis to gliogenesis and a concomitant loss of regional identity.


The ability to generate defined neural cell types from embryonic stem cells (ESCs) offers a powerful resource to study mechanisms of neural differentiation and fate specification. Application of such insights will, furthermore, be valuable for drug screening and cell-based repair strategies for a wide range of neurological disorders. Several studies have demonstrated the in vitro capacity of mouse ESCs to generate neurons [1, [2], [3], [4]–5], including regionally specified subtypes such as dopaminergic neurons and motoneurons (MNs) [6, [7]–8].

Generation of neural complexity is achieved through a process of morphogenetic development and tissue patterning [9, [10]–11]. This involves the induction and integration of signals from graded morphogens. Such signals function to establish regionally determined profiles of transcription factor gene expression that in turn specify distinct neural cell fates. A major principle of neural patterning is that cells exhibit spatially and temporally restricted competence to respond to patterning signals. Thus, forebrain neural progenitors may be progressively caudalized to acquire midbrain and then hindbrain and spinal characteristics.

In this study we sought to predictably manipulate defined ESC-derived neural progenitor cells (NPCs) and examine whether responsiveness to developmental signals is temporally determined. We report the initial derivation of NPCs with a forebrain identity following culture of ESCs in a previously described chemically defined medium (CDM) [12, [13]–14]. Significantly, such precursors can subsequently be patterned with respect to rostral-caudal and dorsal-ventral fates. However, the ability to impose regional identity is temporally restricted, with a late population of NPCs being refractory to developmental signals. This is associated with a loss of regionalization. Furthermore, we provide evidence that the loss of responsiveness to morphogenetic cues correlates with constitutive expression of basic helix-loop-helix (bHLH) transcription factors Olig2 and Mash1 (Ascl1), epidermal growth factor receptor (EGFR), and vimentin and coincides with the onset of gliogenesis.

Materials and Methods

Cell Culture

ESCs were cultured as previously described and differentiated in CDM [12, 14]. ESCs were maintained by routine feeder-free culture in Iscove's Modified Dulbecco's Medium (IMDM)/GlutaMax (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) supplemented with 15% fetal calf serum (BioSera, Ringmer, East Sussex, U.K., http://www.biosera.com), penicillin/streptomycin, nonessential amino acids (Invitrogen), 2-mercaptoethanol (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), and 10 ng/ml leukemia inhibitory factor (Chemicon, Temecula, CA, http://www.chemicon.com). ESC lines E14, CGR8.8, IMT11, and 46C were used, and they showed similar differentiation properties. For differentiation, enzymatically dissociated ESCs were resuspended in CDM [12] and plated at 5–10 × 104 cells per ml in 10-cm bacteriological grade culture dishes. CDM consisted of IMDM/Ham's F-12 medium 1:1 with GlutaMax I, 1× lipid concentrate, 1× penicillin/streptomycin (all from Invitrogen), 5 mg/ml albumin fraction V (Sigma-Aldrich), 15 μg/ml transferrin (Roche Diagnostics; Basel, Switzerland), 7 μg/ml insulin (Sigma-Aldrich), and 450 μM 1-thioglycerol (Sigma-Aldrich). Cultures were passaged by either dissociating spheres using 0.1% trypsin/EDTA or mechanical trituration with a P200 Gilson pipette tip. Growth factors were used at the following concentrations unless otherwise stated: fibroblast growth factor 2 (FGF2), 20 ng/ml (R&D Systems Inc., Minneapolis, http://www.rndsystems.com); epidermal growth factor, 20 ng/ml (R&D Systems Inc.); Sonic hedgehog N-terminal peptide (SHH-N), 250 ng/ml (gift from Curis Inc., Cambridge, MA, http://www.curis.com); all-trans-retinoic acid (RA), 1 μM (Sigma-Aldrich); and SU5402, 10 μM (Calbiochem, EMD Biosciences, San Diego, http://www.emdbiosciences.com/html/CBC/home.html). For terminal differentiation of specified progenitor populations, trypsinized single-cell suspensions were plated on poly-d-lysine (PDL) coated coverslips at 5 × 104 cells per coverslip and cultured in Dulbecco's modified Eagle's medium/2% B27 (Invitrogen).


Immunocytochemistry was performed on either free-floating whole-sphere (whole-mount) cryostat sections or plated cells as previously described [15]. Primary antibodies used were 39-4D5, 4F2, 81.5C10, PAX6, and l12 anti-Hoxb4, reactive to Islet1, Lim1/2, HB9, Pax6, and Hoxb4 respectively (all from Developmental Studies Hybridoma Bank, Iowa City, IA, http://www.uiowa.edu/∼dshbwww); Rat401 anti-nestin (1:400; BD Pharmingen, San Diego, http://www.bdbiosciences.com/pharmingen); anti-β III tubulin (1:400; Sigma-Aldrich); anti-Mash1 (1:20; Pharmingen); anti-vimentin (1:100; Chemicon); anti-Sox2 (1:500; gift of Robin Lovell-Badge); anti-En1/2 (1:1,000; gift of Alex Joyner); anti-Olig2 (1:40,000; gift of David Rowitch; gift of Tom Jessell); anti-Sox1 (1:1,000; Chemicon); anti-EGFR (1:50; Affinity Bioreagents, Golden, CO, http://www.bioreagents.com); anti-GABA (1:500; Sigma-Aldrich); anti-glutamate (1:500; Sigma-Aldrich); anti-choline acetyltransferase (anti-ChAT) (1:500; Chemicon); and anti-glial fibrillary acidic protein (anti-GFAP) (1:500; DAKO, Glostrup, Denmark, http://www.dako.com). Secondary antibodies (Invitrogen) were used at concentrations between 1:500 and 1:1,000.

Cells were viewed under a Leitz microscope. Using a grid, five consecutive fields were counted for each coverslip. The number of positive cells was expressed as a mean ± SEM from three or four coverslips and from three separate experiments. Statistical significance was assessed by analysis of variance with post hoc Neuman-Keuls and Student's t test using Prism 3 (GraphPad, San Diego, CA, http://www.graphpad.com).

Flow Cytometry

Cell aggregates were trypsinized to obtain a single-cell suspension, washed in Ca2+/Mg2+-free PBS, 2% serum, 10 mM sodium azide (PFN), fixed in ice-cold 4% paraformaldehyde for 30 minutes, and then washed in ice-cold PFN. Cells were aliquoted in samples of 2–5 × 105 cells per well in a 96-well plate. Cells were pelleted by centrifugation (1,200 rpm). Incubation with primary antibodies (see above) was performed in PFNS (PFN + 0.14% saponin) for 30 minutes at 4°C. Secondary antibodies were applied for 30 minutes at 4°C. Cells were then washed in PFN and used for sorting. FACS analysis was performed using a FACScalibur flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com) equipped with an argon laser with an emission wavelength of 488 nm. Data analysis was performed using CellQuest software (Becton, Dickinson and Company).

Gene Expression Profiling

Reverse transcription-polymerase chain reaction (RT-PCR) analysis was performed using SMART technology (Clontech, Palo Alto, CA, http://www.clontech.com). SMART enabled reproducible, linear amplification of RNA from small growth factor-induced cultures allowing up to 1,000 RT-PCRs per μg of total RNA to be performed. Gene expression comparisons were always made using matched samples, normalized between each other for the level of Gapdh expression quantified by ethidium bromide or SYBR Green staining and band intensity measurements using FLA-3000 Imager System (Fujifilm, Valhalla, NY, http://www.fujifilm.com). Polymerase chain reaction (PCR) primers and conditions are shown in supplemental online Table 1. All PCR primers were amplified for 32 cycles, except for 30 cycles for Olig2 and 27 cycles for Gapdh.


Neural Differentiation of ESCs in Serum-Free Medium

Efficient neural differentiation was obtained by plating ESCs in CDM devoid of exogenous mitogens, retinoids, or other known neural inducers [12, 13]. Serial immunohistochemical analysis confirmed a gradual loss of the pluripotent marker Oct4 (Pou5f1) and a concomitant rise in the neuroectodermal marker Sox1 (Fig. 1A), with the vast majority of cells positive for Sox1 (87.1% ± 3.3%) and largely negative for Oct4 protein (2.43% ± 0.5%) after 8 days of culture. At 7 days following plating of Sox1+ progenitors in terminal differentiation medium, 89.4% ± 4.3% of cells expressed β-III-tubulin. Subtype analysis showed that 40.1% ± 3.4% were glutaminergic and 55.4% ± 1.9% were GABAergic neurons (Fig. 1B). Importantly, no O4+ and GFAP+ glial cells were detected (data not shown). In addition, no cells immunoreactive for non-neural markers such as the muscle marker desmin or the epidermal marker pan-cytokeratin were detected (data not shown).

Figure Figure 1..

Neural differentiation in chemically defined medium (CDM). (A): Schematic of the differentiation system and corresponding immune micrographs for Sox1 and Oct4 of undifferentiated embryonic stem cells and CDM cultures at D4 and D8. (B): Phase and immune micrographs for β-III-tubulin, glutamate, and GABA expression of D8 CDM cultures differentiated for a further 7 days on poly-d-lysine/laminin. (C): Reverse transcription-polymerase chain reaction analysis showing temporal changes in expression of forebrain (Otx1, Pax6, Emx2, Gsh2, Dlx2), midbrain (En2), and hindbrain (Egr2) markers in CDM cultures, with representative immune staining showing Pax6 and Otx2 expression in D8 control spheres. (D): Differential expression of forebrain (Otx1, Otx2, Tbr1, Foxg1, and Pax6), midbrain (En2), and hindbrain (Gbx2, Egr2, Hoxb3) markers in control D8 CDM cultures and in CDM cultures supplemented with fibroblast growth factor 2 (FGF2) and FGF2 + all-trans-retinoic acid (RA) from D4, with representative immune staining showing Pax6 and En1/2 following FGF2 treatment and Hoxb4 expression following FGF2 + RA treatment. Abbreviations: C, control; D, day; F, FGF2; F/R, FGF2 + RA; mESC, mouse embryonic stem cell.

Serial gene expression studies between D2 and D8 showed a gradual upregulation of markers, consistent with the acquisition of an anterior neuroectodermal fate (Fig. 1C) [16, [17]–18]. By D8, the majority of cells were Pax6+ (69.0% ± 6.1%) and Otx2+ (61.5% ± 2.9%) (Fig. 1C). Treatment of D4 neural cultures with FGF2 and RA, two morphogens implicated in posteriorization of the neural tube, resulted in downregulation of Pax6 (26.0% ± 7.2%) and upregulation of midbrain and hindbrain markers, including an increase in En1/2 from <1% to 16% ± 5.8% with FGF2 alone, and Hoxb4 expression to 64.0% ± 8.0% with FGF2 and RA treatment (Fig. 1D). These results suggest that CDM-derived NPCs are responsive to developmental cues.

Temporal Restriction of NPCs to Patterning Cues

Developmental insights into neural patterning have been previously applied to generate defined neuronal cell types [6, 8]. Specifically, motoneuron generation is dependent on sequential caudalization and ventralization of cultures and can be achieved by the combinatorial application of FGF, retinoic acid, and sonic hedgehog (Shh) [8, 19]. To determine whether CDM-derived precursors were similarly responsive, cultures were treated with FGF2, RA, and SHH-N between D4 and D8. MN development required the dual activities of SHH-N and RA (Fig. 2A–2C). Immunohistochemistry revealed that Isl1- and Hb9-positive cells were dispersed throughout the spheres, and ChAT expression upon differentiation of patterned progenitors confirmed MN identity. The proportion of cells fated to form MNs was analyzed by FACS for immunoreactivity to Isl1 and Hb9 (Fig. 2C). Control cultures contained only a small population of Isl1+ cells and no Hb9. After treatment of cultures with FGF2/RA/SHH-N, a significant increase in the populations of Isl1+ (58%) and Hb9+ (36%) cells was seen.

Figure Figure 2..

Motoneuron directed differentiation of early and late chemically defined medium (CDM) cultures. (A): Reverse transcription-polymerase chain reaction (RT-PCR) analysis showing induction of transcription factors involved in motoneuron (MN) fate determination in CDM cultures treated with fibroblast growth factor 2 (FGF2) + SHH-N + all-trans-retinoic acid (RA) between D4 and D8. (B): Immune micrographs showing Isl1 and Hb9 expression in early CDM cultures treated with FGF2 + SHH-N + RA between D4 and D8 (top panel), and phase and immune micrograph of differentiated MNs positive for choline acetyltransferase (bottom panel). (C): FACS analysis of dissociated D8 spheres for Isl1 and Hb9 FL1-H indicates fluorescence intensity, with cell percentages given for gates M1. (D): RT-PCR analysis showing absence of induction of transcription factors involved in MN fate determination in CDM cultures treated with FGF2 or FGF2 + SHH-N + RA between D16 and D20. (E): Immune micrographs showing the absence of Isl1 and Hb9 induction in late CDM cultures treated with FGF2 + SHH-N + RA between D16 and D20. Abbreviations: D, day; F, FGF2; FL1-H, fluorescence intensity; R, RA; S, SHH-N.

An important feature of stem cells is the ability to generate long-term-expanded cultures. Addition of FGF2 to CDM cultures enabled long-term propagation of NPCs. Serial passaging every 4 days resulted in greater than 800-fold expansion in cell numbers by D20. However, in vivo, neural patterning is temporally regulated [20] such that similar signals can have different, time-dependent effects on the generation of dorsoventral and rostrocaudal patterns [21, [22], [23]–24]. To examine whether the developmental plasticity of ESC-derived neural precursors would also be temporally restricted, we therefore examined whether long-term or late NPCs could also be patterned to motoneurons.

Critically, treatment of late ESC-derived NPCs (D16–D20) with SHH-N and RA failed to induce expression of MN markers Nkx2.9, Nkx6.1, or Hb9 (Fig. 2D). Immunohistochemistry confirmed the absence of Isl1 upregulation and Hb9 expression (Fig. 2E). In addition, control and patterned late NPCs expressed comparable amounts of Olig2 and Nkx2.2. This finding is in marked contrast to the differential profile of similarly treated early NPCs (D4–D8). These results show that the ability to perform systematic manipulations of neural fate, with respect to patterning determinants, is restricted to early ESC-derived NPCs.

To explore this observation further, we examined in greater detail the differential responsiveness of early and late precursors to patterning signals (Fig. 3A). Immunohistochemical analysis of early and late FGF2-treated NPCs confirmed maintenance of the neuroepithelial markers nestin and Sox2 (Fig. 3B, 3D). Furthermore, RC2 was also found at comparable levels in the two populations (Fig. 3C, 3D). However, expression analysis of late cultures revealed constitutive expression of Olig1/2 and Mash1 irrespective of the combination of patterning signals and associated failure of induction of Neurod1, Neurog1, and Neurog2. These findings contrast with the differential responsiveness of early cultures to combinations of FGF2, SHH-N, and RA. Furthermore, loss of responsiveness to patterning was also associated with an upregulation of Egfr and the radial glial marker Blbp and with acquisition of the glial genes Gfap and Pdgfr1. In addition, early cultures had very low numbers of Olig2+, EGFR+, and vimentin+ cells, whereas, in contrast, late cultures were strongly positive for Olig2 (61.0% ± 7.3%), EGFR (76.5% ± 4.2%), and vimentin (71.9% ± 9.4%) (Fig. 3C, 3D). A significant upregulation of Mash1 was also observed (19.2% ± 3.6%).

Figure Figure 3..

Responsiveness of chemically defined medium (CDM) cultures to patterning cues is temporally restricted and associated with the onset of gliogenesis. (A): Reverse transcription-polymerase chain reaction comparison of D8 and D20 cultures supplemented with FGF2 for their responsiveness to patterning by SHH-N, all-trans-retinoic acid (RA), or SHH-N + RA. (B): Immune micrographs showing maintenance of nestin and Sox2 expression between D8 and D20 cultures, maintained in fibroblast growth factor 2 from D4 onwards. (C): Immune micrographs showing comparable expression of RC2 and differential expression of Olig2, EGFr, vimentin, and Mash1 between D8 and D20 cultures. (D): Quantification of numbers of cells positive for the markers analyzed (∗, p < .05; ∗∗, p < .01). (E): Immune micrograph showing β-III-tubulin, GFAP, and O4 expression of D20 CDM cultures differentiated on poly-d-lysine/laminin for a further 7 days. Abbreviations: D, day; EGFR (EGFr), epidermal growth factor receptor; F, FGF2; GFAP, glial fibrillary acidic protein; R, RA; S, SHH-N.

Consistent with the acquisition of EGFR expression, 5-bromo-2′-deoxyuridine analysis showed late cultures were responsive to EGF independent of FGF2 signaling (Table 1). Differentiation of late cultures, in addition to showing large numbers of neurons (β-III-tubulin, 63.7% ± 2.7%), also revealed GFAP+ and O4+ cells at 7 days postplating, thus confirming the gliogenic potential of late NPCs (Fig. 3E). Taken together, these observations suggest that the temporal loss of responsiveness of ESC-derived NPCs to neural patterning coincides with loss of regionalization and the acquisition of glial potential.

Table Table 1.. BrdU incorporation in D20 CDM cultures maintained in the absence or presence of the mitogens FGF2 and EGF
original image


The ability of ESC-derived NPCs to respond in a predictable manner to developmental patterning cues allows the generation of defined neuronal cell types for use in drug screening studies or cell-based therapies for neurodegenerative diseases. In this study, we demonstrate that the ability to pattern ESC-derived NPCs is temporally restricted. The loss of responsiveness to morphogenetic cues correlates with constitutive expression of bHLH transcription factors Olig2 and Mash1, EGFR, and vimentin and the onset of gliogenesis.

Long-Term-Propagated NPCs Lose Responsiveness to Neural Patterning Cues

Cells differentiated in CDM share characteristics with early embryonic neural stem cells, are responsive to developmental patterning mechanisms, and can be maintained in FGF2. Early NPCs were sensitive to rostrocaudal and dorsoventral patterning. As proof of principle, we show that fate determination could be manipulated by the selective differentiation of motoneuron progenitors through combined caudalization and ventralization by RA and Shh signaling, consistent with previous reports [8].

The utility of propagated precursors is in part dependent on their ability to respond to extrinsic signals, thus permitting the generation of specific regional neuronal subtypes. Late precursors were refractory to both posterior and ventral patterning signals (RA/SHH-N). Temporal regulation of patterning events is well documented in vivo and may result from extrinsic influences associated with progressive development of the neural tube [23, 25]. In addition, this study suggests that temporal restriction to developmental cues may be an intrinsic property of early NPCs. The uniform molecular marker profile of late NPCs irrespective of external signaling is in contrast to that of early NPCs and suggests constitutive expression that does not directly correlate with a specific regional identity. This is perhaps best illustrated by the profile of Olig2; in early NPCs it is markedly upregulated specifically following treatment with RA/SHH-N and is consistent with the known requirement of Olig2 for motoneuron specification. In contrast, late cultures express a uniform upregulation of Olig2 irrespective of extrinsic signaling, consistent with evidence implicating Olig2 in self-renewal [26]. It is uncertain whether such an explanation also accounts for the observed upregulation of another bHLH transcription factor, Mash1, upon expansion. The marked failure of induction of other neurogenic and fate-determining transcription factors, including Ngn2, Nkx2.9, and Nkx6.1, following patterning suggests a general loss of developmental plasticity upon expansion. Together these results are consistent with the idea that FGF2 alters the developmental competence of NPCs [26, 27].

ESC-derived NPCs share some characteristics with the radial glial lineage. Accumulating evidence suggests that radial glial cells can behave as neural stem cells and contribute to developmental neurogenesis [28, [29]–30] and that neuronal differentiation of ESCs occurs through radial glial intermediates [31, [32], [33]–34]. It is unclear whether embryonic radial glia are heterogeneous. Fate mapping studies support the idea of regional differences with differential neuronal and glial potential, although more recent studies provide evidence for a more widespread neurogenic potential of radial glia [35, 36].

During development, neurogenesis precedes gliogenesis, a process influenced by both intrinsic and extrinsic signals. Neurogenic radial glia are distinguished by the presence of Pax6 and Ngn1/2 and the absence of GFAP and EGFR [32, 35, 37]. In contrast acquisition of EGFR and GFAP is characteristic of gliogenic radial glia observed in midgestation and is associated with astrocyte differentiation [38, 39]. The early NPCs identified in this study have a phenotypic profile consistent with putative neurogenic radial glia. In contrast, the absence of Ngn and upregulation of Olig2, Mash1, along with acquisition of EGFR found in late NPCs, are similar to the profile of putative gliogenic radial glia. The late NPCs identified in this study share some features with a recent study that selected a homogenous population of Olig2/Egfr/Mash1-positive cells that could be maintained long-term and retained the ability to differentiate into neurons and astrocytes [40].

One interpretation of our findings is that propagation of ESC-derived NPCs results in a transition from early (pattern competent, neurogenic) to late (pattern refractory, gliogenic) radial glia, compatible with studies suggesting that FGF determines hierarchical ordered production of cellular diversity [41, 42]. The two distinct populations of NPCs demonstrate a neurogenic to gliogenic transition comparable to that observed during central nervous system development. The acquisition of glial potential is associated with the loss of competence to neuronal patterning and coincides with constitutive Olig2 and EGFR expression. It would be of interest to determine whether these findings are relevant for human embryonic stem cells (hESCs), since this would have considerable implications for the use of hESC-derived NPCs in generating defined cell types for transplantation and drug screening.


Together, these findings are consistent with the idea that regional identity and neurogenesis are temporally linked and that propagation of ESC-derived NPCs results in loss of regionalization and responsiveness to neural patterning.


This work was supported by grants from the Biotechnology and Biological Sciences Research Council and Parkinson's Disease Society, United Kingdom (to N.D.A.) and the Medical Research Council (to S.C.). We thank Lee Rubin for SHH-N protein and Robin Lovell-Badge, Alex Joyner, David Rowitch, Tom Jessell, and the Developmental Studies Hybridoma Bank for antisera. I.A.B., A.J., and H.K. contributed equally to this work. I.A.B. passed away on October 6, 2005. We dedicate this manuscript to Isabelle Bouhon for the inspiration and joy she gave to all who knew her.