An epigenetic signature of developmental potential in neural stem cells and early neurons


  • Matthew J. Burney,

    1. Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, London, United Kingdom
    Search for more papers by this author
  • Caroline Johnston,

    1. Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, London, United Kingdom
    Search for more papers by this author
  • Kee-Yew Wong,

    1. Stem Cell and Developmental Biology Group, Genome Institute of Singapore, Singapore
    Search for more papers by this author
  • Siaw-Wei Teng,

    1. Stem Cell and Developmental Biology Group, Genome Institute of Singapore, Singapore
    Search for more papers by this author
  • Vassilios Beglopoulos,

    1. Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, London, United Kingdom
    Search for more papers by this author
  • Lawrence W Stanton,

    1. Stem Cell and Developmental Biology Group, Genome Institute of Singapore, Singapore
    2. Adjunct Professor, Department of Biological Sciences, National University of Singapore, Singapore
    Search for more papers by this author
  • Brenda P Williams,

    1. Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, London, United Kingdom
    Search for more papers by this author
  • Angela Bithell,

    Corresponding author
    1. Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, London, United Kingdom
    • Correspondence: Noel J. Buckley, Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, 125 Cold harbour lane, London SE5 9NU, UK. Telephone: +44-2078480784; Fax: +44–2078485308; e-mail: Angela Bithell, Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, 125 Cold harbour lane, London SE5 9NU, UK. Telephone: +44-2078485606. Fax: +44–2078485308; e-mail:

    Search for more papers by this author
  • Noel J. Buckley

    Corresponding author
    1. Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, London, United Kingdom
    • Correspondence: Noel J. Buckley, Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, 125 Cold harbour lane, London SE5 9NU, UK. Telephone: +44-2078480784; Fax: +44–2078485308; e-mail: Angela Bithell, Centre for the Cellular Basis of Behaviour, Department of Neuroscience, Institute of Psychiatry, King's College London, James Black Centre, 125 Cold harbour lane, London SE5 9NU, UK. Telephone: +44-2078485606. Fax: +44–2078485308; e-mail:

    Search for more papers by this author

  • Author contributions: M.J.B.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; C.J., K-Y.W., and S-W.T.: collection and/or assembly of data; data analysis and interpretation; V.B.: provision of study material or patients; L.W.S.: conception and design, financial support; B.P.W.: conception and design; A.B.: conception and design, data analysis and interpretation, manuscript writing; N.J.B.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript.


A cardinal property of neural stem cells (NSCs) is their ability to adopt multiple fates upon differentiation. The epigenome is widely seen as a read-out of cellular potential and a manifestation of this can be seen in embryonic stem cells (ESCs), where promoters of many lineage-specific regulators are marked by a bivalent epigenetic signature comprising trimethylation of both lysine 4 and lysine 27 of histone H3 (H3K4me3 and H3K27me3, respectively). Bivalency has subsequently emerged as a powerful epigenetic indicator of stem cell potential. Here, we have interrogated the epigenome during differentiation of ESC-derived NSCs to immature GABAergic interneurons. We show that developmental transitions are accompanied by loss of bivalency at many promoters in line with their increasing developmental restriction from pluripotent ESC through multipotent NSC to committed GABAergic interneuron. At the NSC stage, the promoters of genes encoding many transcriptional regulators required for differentiation of multiple neuronal subtypes and neural crest appear to be bivalent, consistent with the broad developmental potential of NSCs. Upon differentiation to GABAergic neurons, all non-GABAergic promoters resolve to H3K27me3 monovalency, whereas GABAergic promoters resolve to H3K4me3 monovalency or retain bivalency. Importantly, many of these epigenetic changes occur before any corresponding changes in gene expression. Intriguingly, another group of gene promoters gain bivalency as NSCs differentiate toward neurons, the majority of which are associated with functions connected with maturation and establishment and maintenance of connectivity. These data show that bivalency provides a dynamic epigenetic signature of developmental potential in both NSCs and in early neurons. Stem Cells 2013;31:1868-1880


A number of recent studies have focused on the epigenome as an indicator of cell fate or potential. The epigenome is an attractive territory to scour for the characteristics of neural stem cell (NSC) plasticity as it occupies a key position in the hierarchy of development, acting as custodian of cellular memory, allowing long lasting changes to be established, maintained, and perpetuated [1]. Epigenetic regulation of development is most keenly recognized in embryonic stem cell (ESC) research, where the concept of “bivalency” was born. This notion is based on the observation that many lineage-specific developmental genes in ESCs are transcriptionally silent, yet their promoters carry both “active” and “repressive” epigenetic signatures, typically trimethylation of histone H3 lysine 4 and 27 (H3K4me3 and H3K27me3, respectively) [2, 3]. Upon differentiation, these signatures typically lose both marks or resolve into monovalent epigenetic signatures, characteristic of repressed or active genes, consistent with the lineage of the progeny. In this way, bivalency is thought to represent a mechanism by which lineage-specific regulators can be held in a transcriptionally silent yet poised state, ready for immediate expression upon receipt of an appropriate differentiation cue. Importantly, changes in the epigenetic signature of developmental genes during ESC differentiation occur in the absence of any concurrent changes in transcription or translation [4]. These observations provide an epigenetic framework for identifying and understanding stem cell potential.

In this study, we explore this concept within the framework of NSC differentiation. The mammalian brain consists of several hundred distinct neuronal types arising from multipotent progenitors that reside in the germinal zones along the developing neural tube. These undergo expansion, migration, and differentiation throughout embryonic development to generate specific neuronal types [5]. Like all developmental processes, neural differentiation is essentially unidirectional and manifests as a concurrent loss of progenitor potential and gain of differentiated phenotype. Each stage of this largely unidirectional pathway must be tightly regulated to ensure that the loss of multipotentiality and concomitant gain of lineage-specific characteristics is properly orchestrated. Transplantation and reprogramming studies have shown that the fate of NSCs, both in vitro and in vivo, is readily manipulatable in line with the generally held developmental axiom of “potential is greater than fate.” It is this plasticity that underpins the usefulness of NSCs in regenerative medicine. Many of the signaling events and transcriptional programs that drive acquisition of specific neuronal phenotypes are known, at least in part. However, in themselves, knowledge of signaling and transcriptional programs that drive neuronal differentiation does not inform us about the neurogenic potential of a NSC.

Here, we have examined the dynamic profile of bivalency (H3K4me3/H3K27me3) during the differentiation of NSCs to early GABAergic neurons using ESC-derived NSCs engineered to permit sorting of pure neuronal populations. By focusing on newly differentiated neurons, we were able to assess changes in the epigenome that precede many of the subsequent gene expression changes that occur later during neuronal maturation. This provides a clear perspective of the capacity of the epigenome to predict subsequent acquisition of a particular differentiated neuronal phenotype. We find that bivalent promoters in NSCs reflect a broad neuronal potential, whereas in early neurons, we find both retention and an unexpected gain of bivalency at the promoters of genes governing later acquisition of mature neuronal phenotype. Thus, bivalency provides a dynamic epigenetic read-out of fate restriction and phenotypic plasticity throughout the entire developmental axis, from pluripotency through multipotency to committed neuron.

Materials and Methods

Mouse Cell Culture and Differentiation

Feeder-free murine 46C ESCs [6] were electroporated with a RedStar and Blasticidin selection cassette, containing the first two exons of the Tau locus (supporting information Fig. S1A). ESCs were screened for bona fide homologous recombination events by long-range polymerase chain reaction (PCR) and confirmed by Southern blot (supporting information Fig. S1B). NSCs were derived by adherent monolayer culture in N2B27 [7], selected using 0.5 µg/ml puromycin (Sigma) and a clonal line was derived by plating in NSC expansion medium (Euromed-N, EuroClone, supplemented with 1× penicillin/streptomycin (Sigma), 2 mM l-glutamine (Sigma,, N2 (insulin [25 µg/ml], sodium selenite [30 nM], apo-transferrin [100 µg/ml, SCIPAC], putrescine [16 µg/ml], progesterone [19.8 ng/ml], all Sigma except where shown), bovine serum albumin (50.25 µg/ml, Invitrogen), and 10 ng/ml of epidermal growth factor (EGF) and fibroblast growth factor (FGF2) (both Peprotech,, at limiting dilution in Terasaki plates (Nunc). NSCs were routinely grown in NSC expansion medium and passaged with 0.2× trypsin-EDTA (Sigma; diluted in 0.5 mM EDTA). NSCs were differentiated into neurons by plating at 1 × 105 cells per square centimeter in Euromed-N medium supplemented with 0.5 × N2, 5 ng/ml FGF2 and 0.5 × B27 (Invitrogen, After 4 days, cells were passaged using Accutase (PAA Laboratories, onto polyornithine (Sigma) and laminin-coated (Sigma) plastic in Neurobasal medium (Invitrogen) supplemented with 0.5 × B27. Tripotential differentiation was performed as reported previously [8].

FACS Purification of NS-TRB-Derived Neurons

NS-TRB-derived neurons were harvested using Accutase (PAA), resuspended in cold PBS and 0.1 mM EDTA (Sigma), and passed through a 70-µm cell sieve (BD Biosciences). NS5 cells [7] that underwent an identical differentiation paradigm were used as a negative control for RedStar fluorescence. Dead cells were labeled with 1 µM SYTOX Blue dye (Invitrogen) for 5 minutes. Cells were sorted on a BD Bioscience fluorescence-activated cell sorting (FACS) Aria I SORP running FACSDiva6.3 software. RedStar+/SYTOX cells were collected directly into LoBind tubes (Eppendorf) and processed immediately for total RNA extraction using the RNEasy column purification system (Qiagen,, or frozen as 1% formaldehyde-fixed cell pellets for chromatin preparation (see below). A small sample of sorted cells was also used to re-sort (to determine the efficiency of sorting) and for subsequent immunocytochemistry and fluorescence microscopy to determine the purity of neuronal populations obtained (see below).

Chromatin Immunoprecipitation and Sequencing

For chromatin preparation, cells were fixed in 1% formaldehyde in PBS at room temperature (RT) for 8 minutes, quenched in 0.125 M glycine (Sigma) for 5 minutes and washed in cold PBS with 1× protease inhibitor cocktail (Roche) before pelleting. Fixed cell pellets were either frozen or used directly. FACS-purified cell pellets were pooled and lysed in lysis buffer and chromatin was sheared using a BioRuptor Sonicator (Diagenode,, to an average size of 300–500 bp, determined using High Sensitivity Bioanalyser DNA Chips (Agilent). All steps included 1× protease inhibitor cocktail, (Roche,

Chromatin immunoprecipitation (ChIP) was performed as described previously [9]. Briefly, 10 µg precleared chromatin was immunopreciptated at 4°C overnight with 1–2 µg of antibodies specific to H3K4me3 (Active Motif,, H3K27me3 (Upstate), H3 (Abcam), or nonspecific rabbit IgG (Santa Cruz). Antibody-protein complexes were captured using preblocked magnetic protein-G beads (Active Motif). Following washes, eluted chromatin was de-crosslinked, proteinase K-treated, and purified using a QIAquick PCR purification kit (Qiagen). For sequencing, ChIP DNA was end-repaired, ligated to sequencing adaptors, and subjected to 18 rounds of amplification, according to standard Illumina Solexa protocols. Size-selected fragments were hybridized to the flow cell surface and parallel sequenced to yield approximately 30 million 35 bp reads (GEO accession number GSE46792 and GSE46793). For pre- and postsequencing validation (n = 3), real-time PCR was performed using ChIP DNA (ChIP-qPCR) and locus-specific primers (see supporting information Table S1). Data were normalized using nonspecific IgG, total H3, and enrichment at nonspecific loci.

For sequential ChIP experiments, the ReChIP kit (Active Motif) was used according to manufacturers instructions. Briefly, 20 µg of precleared sheared chromatin was sequentially immunoprecipitated, first with 1 µg of anti-H3K27me3 followed by 1 µg of anti-H3K4me3 or vice versa. Both ChIPs were carried out overnight at 4°C in a total volume of 200 µl. Subsequent ChIP-qPCR analysis was performed as described above.

Epigenome and Transcriptome Analyses

RNA samples from NSCs (n = 5) and neurons (n = 3) were used for microarray analysis using Illumina Expression BeadChip arrays (MouseRef-8v2.0) with standard protocols (GEO accession number GSE46791). Raw microarray data were processed using the BeadArray R package [10] provided as part of the Bioconductor libraries [11] and analyzed for differential gene expression using LIMMA [12]. Clustering was performed using the Heatmap.2 package, with gene ontology (GO) annotation on representative clusters computed using DAVID [13].

ChIP and sequencing (ChIP-Seq) raw data were uploaded to the Amazon EC2 platform, files converted to fastq format, and quality checked using FastQC, while poor quality sequences and duplicate reading were removed. Of the remaining reads, approximately 60% aligned to the mouse reference genome (NCBIv37 and mm9) using Bowtie, with a seed sequence length of 15 bp [14]. Bowtie output sam files were converted to binary bam files then sorted and duplicates removed before indexing. Enriched regions of histone modification were detected using SICER [15] as peaks of H3K27me3 often have a wider more diffuse signature throughout gene bodies [16] that SICER is able to more accurately identify with high statistical confidence compared with other peak-calling algorithms. Peaks were annotated to the nearest Ensembl gene transcription start site (TSS) using the IRanges R package (available and categorized according to their histone modification and expression profile using R.

Immunocytochemistry and Fluorescence Imaging

Cells were fixed in 4% cold paraformaldehyde (BDH) for 10 minutes at RT and washed twice in PBS before and after permeabilization in 0.1% TritonX-100 (Sigma) in PBS for 5 minutes. Cells were then incubated overnight at 4°C or for 1 hour at RT with primary antibodies in 10% normal serum in PBS (primary antibodies used were: Blbp [rabbit IgG, 1:200, Abcam, www.abcam]; Dcx [rabbit IgG, 1:2000, Abcam]; GABA [rabbit IgG, 1:2000, Chemicon,]; GFAP [mouse IgG1, 1:200, Millipore]; Ki67 [rabbit IgG, 1:1000, Abcam]; MAP2 [mouse IgG1 1:200, Abcam]; Nestin [mouse IgG1, 1:100 [Chemicon]; NeuN [mouse IgG1, 1:100, Chemicon]; O4 [mouse IgM, 1:200, Chemicon]; Olig2 [rabbit IgG, 1:1000, Millipore]; RC2 [mouse IgM, 1:100, Developmental Studies Hybridoma Bank, DSHB]; Sox2 [goat IgG, 1:100, Santa Cruz]; Tau [rabbit IgG, 1:10,000, DAKO]; TTF1 [Nkx2.1, mouse IgG1, 1:100, Progen]; TuJ1 [mouse IgG2a, 1:1000, Covance]; Vimentin [mouse IgM, 1:100, DSHB]). Cells were then washed twice in PBS and incubated for 30 minutes at RT with an appropriate secondary antibody (secondary antibodies used were mouse, rabbit, or goat AlexaFluor 488, 594, or 647 as appropriate [all 1:1000, Molecular Probes,]). Cell nuclei were counterstained with DAPI (0.5 µg/ml; Sigma) and mounted in ProLong Gold (Molecular Probes). Cover slips were analyzed using a Zeiss AxioImager Z.1 fluorescence microscope and AxioVision 4.6 software. Where cells from pre- and post-FACS were immunolabeled, cells were first plated onto PDL-coated cover slips before being processed as described above.


A Selectable Population of Pure Differentiated Neurons from NSCs

As neuronal differentiation of NSCs is often inefficient due to the production of heterogeneous progeny, we first designed a NSC culture system capable of generating pure neuronal populations, to obviate any confounds introduced by cell heterogeneity. 46C Sox1-GFP ESCs [6] were electroporated with a selection cassette comprising the Tau promoter driving RedStar fluorescence and Blasticidin resistance (TRB) (supporting information Fig. S1A) and screened for clones that had underwent homologous recombination (supporting information Fig. S1B). We then used the procedure of Conti et al. [7] to generate a clonal line of NSCs (NS-TRB NSCs) that faithfully recapitulates the phenotype and behavior of embryonic cortical neurogenic radial glia (supporting information Fig. S2) and allowed sorting of NS-TRB-derived neurons by flow cytometry or selection with blasticidin (Fig. 1A–1C). NS-TRB NSCs expressed characteristic NSC markers (supporting information Fig. S3) and were tripotent (supporting information Fig. S4). Stepwise removal of growth factors and use of N2B27 medium supported differentiation of NS-TRB NSCs into GABAergic neurons (Fig. 1D and supporting information Figs. S5 and S6). Neurons were routinely harvested after 11 days by FACS based on RedStar fluorescence (>96% Redstar positive) generating essentially pure neuronal populations (>99% βIII-tubulin positive, not shown), with <1% of glia retained post-sort (supporting information Fig. 1A–1C). Of the βIII-tubulin-positive neurons, 80.7% (±6.1 SD) were GABAergic as adjudged by immunolabeling with anti-GABA antibodies.

Figure 1.

Generation of pure neuronal populations by fluorescence-activated cell sorting (FACS). Purity of NS-TRB-derived neuronal populations before (A) and after (B) FACS purification, assessed by β3-tubulin (neuronal) and GFAP (astrocyte) immunoreactivity. (C) Cell were sorted based on RedStar fluorescence, analyzed using the PE Phycoerythrin detector (red peak), with wild-type (Redstar-negative) neurons (from a Redstar-negative neural stem cell line) as a negative control (blue peak). (D) Neurons were also 80.7% (±6.1SD) gamma-aminobutyric acid-immunoreactive (C). Scale bars = 20 µm (A, B, D). Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; FACS, fluorescence-activated cell sorting; GABA, gamma-aminobutyric acid; GFAP, glial fibrillary acidic protein.

Correlation of Epigenome and Transcriptome Data

We performed ChIP-seq for H3K4me3 and H3K27me3 on chromatin from NS-TRB NSCs and pure NS-TRB neurons, followed by bioinformatic analyses to identify significantly enriched peaks. All histone modification peaks that overlapped a TSS were annotated with Ensembl gene IDs, leading to identification of 8,588 and 6,416 H3K4me3 peaks and 1,784 and 1,507 H3K27me3 peaks in NSCs and neurons, respectively (supporting information Data). Peaks displayed characteristic profiles similar to those reported previously [16] (supporting information Fig. S7) and a subset were validated by real-time PCR (ChIP-qPCR) (supporting information Fig. S8A). In addition, 624 and 206 “bivalent” promoters were identified in NSCs and neurons, respectively. To ensure that the observed bivalency was due to each modification being present on the same nucleosome, rather than as a result of a mixed cell population, we analyzed the co-occurrence of H3K4me3 and H3K27me3 by sequential ChIP at selected bivalent NSC loci. Enrichment for both modifications was observed compared to IgG controls, regardless of which antibody was used to immunoprecipitate first (supporting information Fig. S9).

To investigate the relationship between epigenetic signatures and gene expression, we performed parallel transcriptome analyses using microarrays. Expression of 946 genes changed significantly upon neuronal differentiation (fold change greater than ±2, adjusted p < 0.05), with 404 upregulated and 542 downregulated genes (supporting information Data). We chose a number of genes representing the dynamic range of the microarray to validate, which displayed similar differential expression whether assessed by microarray or qualitative real time PCR (supporting information Fig. S8B). The top 500 most significantly changed genes were used to construct a heatmap (Fig. 2), with genes clustered according to their relative raw expression levels in NSCs and neurons. Four overall clusters were identified corresponding to: (i) genes whose expression was maintained or upregulated, where GO analysis revealed an overrepresentation of functions related to chromatin dynamics and terminal neuronal differentiation, including nucleosome or chromatin assembly, regulation of synaptic transmission, and cell morphogenesis involved in differentiation; (ii) genes whose expression was downregulated or (iii) silenced, both dominated by GO categories linked with regulation of cell cycle including DNA replication, DNA metabolic processes, and mitosis; and (iv) silent genes that became activated and were overrepresented by GO terms connected to neuronal differentiation, including neurogenesis, neuron projection development, and axonogenesis (adjusted p < 0.05).

Figure 2.

Heat map of gene expression levels in NS-TRB neural stem cells (NSCs) and neurons. The heat map shows all genes that significantly changed expression during neuronal differentiation (adjusted p < 0.05). Genes are clustered according to their relative expression levels from low (blue) to high (red). Four overall clusters were identified corresponding to (A) genes whose expression was maintained or upregulated (B) genes whose expression was down regulated or (C) silenced; and (D) repressed genes that became activated (vertical red bars). Each cluster of similarly changing genes is annotated with the corresponding gene ontology terms that are significantly over-represented (adjusted p < 0.05). Abbreviation: NSC, neural stem cell.

Gene promoters with peaks of H3K4me3 were significantly enriched at expressed genes, both in NSCs (Fisher's exact test = 1.79, p < 2.2 × 10−16) and neurons (Fisher's exact test = 1.93, p < 2.2 × 10−16) (Fig. 3A). Furthermore, most expressed gene TSSs were marked by H3K4me3 peaks in NSCs (92.8%) and neurons (78.1%). In contrast, genes with H3K27me3 peaks were significantly enriched for silent genes in both NSCs (Fisher's exact test = 1.26, p = 1 × 10−7) and neurons (Fisher's exact test = 1.25, p = 4.61 × 10−8) (Fig. 3A). The majority of bivalent promoters were also associated with silent genes in both NSCs and neurons (97.6% and 95.1%, respectively). We examined these data further by comparing histone modification peak size with the raw gene expression level of each gene (Fig. 3B). We found that higher enrichment for H3K4me3 is associated with higher levels of gene expression in both NSC and neurons. Genes that were monovalent for H3K27me3 were largely silent, regardless of the size of the H3K27me3 peak. In addition, we classified all genes enriched for only H3K4me3 or H3K27me3 by how they changed upon differentiation and correlated this with changes in gene expression and GO term membership (Fig. 4). We identified a large number of monovalent H3K4me3 genes in NSCs that remained so in neurons. Despite this, 1,833 H3K4me3 enriched genes switched to H3K27me3 or to no modification during differentiation, associated with GO terms including “Gastrulation” and “Mesoderm formation” or “M phase of the mitotic cycle”, respectively. Intriguingly, a small number of monovalent H3K4me3 genes in NSCs also became enriched for H3K27me3 in neurons, with such de novo bivalency found to be significantly associated with genes involved in “Axonogenesis” and “Neuron projection development”.

Figure 3.

Correlation of gene expression with histone modification. (A) Box and whisker plot comparison of raw gene expression level with the histone modification status of the transcriptional start site (TSS) with neural stem cells (NSCs) (NS-TRB) in yellow and neurons in blue. The line shows the median, the box the 25/75th quartiles, and the whiskers the min/max values of the data. The majority of genes with promoters enriched for H3K4me3 alone were expressed in both cell types. Conversely, gene promoters marked by H3K27me3 were largely silent. Bivalent promoters (enriched for both H3K4me3 and H3K27me3) were silent, as were those promoters marked by neither modification. (B) Correlation between histone modifications and gene expression. Plots show the relative enrichment levels for H3K4me3 and H3K27me3 at TSSs in NS-TRB NSCs and neurons versus the relative level of gene expression for each gene.

Figure 4.

Analysis of NS-TRB neural stem cell (NSC) monovalent genes following differentiation. The left panel shows the proportion of gene promoters in NSCs that are monovalent for either H3K4me3 or H3K27me3 (purple and brown, respectively) and their corresponding gene expression status – expressed (green) or silent (red). The right panel shows the same promoters following neuronal differentiation. Promoters were found to: (1) remain monovalent for H3K4me3 (purple) or H3K27me3 (brown); (2) switch from H3K4me3 to H3K27me3 or vice versa; (3) become bivalent (blue), or (4) become unmodified (yellow). Categories of promoter status following differentiation (outlined in 1–4) with memberships statistically enriched for specific gene ontology (GO) terms are shown in red (adjusted p < 0.05). No significant GO term membership was detected for gene promoters that were monovalent for H3K27me3 in NSCs and became either monovalent for H3K4me3 or bivalent in neurons. Abbreviation: NSC, neural stem cell.

Promoter Bivalency Defines the Neurogenic Potential of Uncommitted NSCs

The original concept of bivalency was born from observations in ESCs and was proposed to act as a mechanism by which lineage-specific developmental genes could be held in a silent yet poised state, ready for possible later expression in response to differentiation cues [2, 17]. Therefore, bivalency represents an epigenetic read-out of cellular potential and likely functions not just in the pluripotent state, but also to maintain multipotency in more lineage-restricted stem cells [3, 18, 19]. With this in mind, we reasoned that bivalent domains in NSCs would mark genes that are necessary for regulating subsequent neuronal differentiation, providing an epigenetic read-out of the neurogenic potential of a NSC. We found that bivalent genes in NSCs were significantly associated with “neurogenesis,” “tissue morphogenesis,” and “CNS development” GO terms (supporting information Table S2). Genes important for the generation of glial lineages, such as the transcription factors Tcf4, Sox9, Rxrb, and Sox10 [20], and the astrocyte-specific markers Gfap (Gfap is not expressed by embryonic NSCs or progenitors) and Aqp4 had neither modification or were monovalent for H3K4me3 in both NSCs (consistent with published findings in neural progenitors [NPCs] [3]) and neurons. This may be partially explained by the low CpG content of several astrocyte promoters including Gfap and S100β, which are known to be silenced by DNA methylation during embryonic neurogenesis [21] and is also consistent with findings in ESCs that bivalent genes associated with gliogenesis are more likely to resolve to H3K4me3 in NPCs than those associated with neurogenesis [3].

We next focused on whether the existence of bivalency at specific promoters was indicative of subsequent fate upon differentiation and whether bivalent marks were restricted to key transcriptional regulators of neuronal developmental, or were also present at downstream “effector” genes of cellular phenotype. Furthermore, as our in vitro NSC differentiation paradigm generates exclusively GABAergic neurons, we looked for correlation between bivalency at genes required for GABAergic neuron development compared with other neuronal subtypes. Surprisingly, bivalent promoters were present at many genes encoding transcription factors required for the generation of both non-GABAergic and GABAergic neuronal subtypes (supporting information Data). The former included Neurog1, required for the generation of glutamatergic cortical projection neurons [22], Lmx1a, required for mesencephalic dopaminergic neuronal differentiation [23], Isl1, required for cholinergic neuronal differentiation [24], and Pitx2, implicated in glutamatergic neuronal development [25]. Furthermore, several neuronal subtype-specific synaptic components were identified in this group, including the glutamate receptors Grin1 and Grin2b, and the 5HT serotonin receptors Htr7 and Htr6 characteristic of glutamatergic and serotonergic neurons, respectively. Together, these data show that, despite the apparent restriction of these NSCs to a GABAergic fate, the promoters of developmental regulators and effectors of both GABAergic and non-GABAergic genes bear bivalent epigenetic signatures, consistent with the broad developmental potential of NSCs.

Resolution of Bivalency During Neuronal Differentiation

As promoter bivalency is associated with cell potential then we next assessed how promoter bivalency resolved upon neuronal differentiation. Of the 624 bivalent promoters in NSCs, 491 (79%) resolved to monovalency or became unmodified upon differentiation (supporting information Data). As anticipated, resolution to monovalent H3K4me3 correlated with an upregulation of gene expression, while resolution to monovalent H3K27me3 was accompanied by loss of expression or maintenance of repression (Fig. 5). To more rigorously correlate bivalency resolution to monovalent H3K4me3 with changes in gene expression, we used the DeSeq package [26] to calculate the change in H3K27me3 peak size between NSCs and neurons. The majority of those genes that showed a decrease in H3K27me3 showed either no change or an increase in expression (supporting information Fig. S10).

Figure 5.

Changes in promoter bivalency and gene expression during neuronal differentiation. The heat map shows all 697 gene promoters that are bivalent in either neural stem cells (NSCs) (NS-TRB) or neurons, clustered according to their relative changes in enrichment levels for each histone modification between NSCs and neurons (light to dark blue bars). Corresponding gene expression level changes for each gene following differentiation are also shown in the far-right panel (“Expression”) with red showing decreased expression and green showing increased expression. Black indicates no change in expression between NSC and neuron. Genes are clustered into four main groups according to their ‘epigenetic status' after neuronal differentiation: (1) monovalent (H3k4me3 or H3k27me3) gaining bivalency (dark blue); (2) bivalency retained (green); (3) bivalent resolving to H3K4me3 monovalent (light brown), or (4) bivalent resolving to monovalent H3K27me3 (dark brown).

Resolution of bivalency to H3K27me3 alone or loss of both H3K4me3 and H3K27me3 during differentiation predominantly occurred at promoters of genes related to “organ morphogenesis” and “tissue development” (supporting information Table S3), including the developmental regulators Gli1 and Wnt1. Several of the 306 bivalent genes that resolve to H3K27me3 during neuronal differentiation were NSC specific transcription factors, including Six1, Foxd1, and Gsx1. Genes required for the generation of other lineages were also significantly enriched, including regulators of the mesenchymal lineage, such as Ovol2 and Crb3 [27, 28]. In contrast, bivalent genes that resolved to monovalent H3K4me3 upon differentiation were largely members of GO categories associated with neuronal development, neuronal morphogenesis, and axonogenesis (supporting information Table S4). The latter included the GABA receptor β3 subunit Gabrb3, the neuronal-specific microtubule protein βIII-tubulin (Tubb3) (supporting information Fig. S11), the Eph B1 receptor (Ephb1), implicated in synaptogenesis [29], and transcription factors Brn3a (Pou4f1), Ctip2 (Bcl11b), and Lhx4, all involved in regulation of neuronal differentiation [30-33]. Most of these changes in epigenetic signature occurred in the absence of any change in gene expression. These findings suggest that bivalent promoters that resolve to H3K27me3 upon differentiation mark genes that regulate NSC establishment and maintenance, whereas those that resolve to H3K4me3 are required for neuronal differentiation. Interestingly, upon neuronal differentiation, bivalent genes encoding non-GABAergic neuronal subtype-specific transcription factors (e.g., Neurog1 and Lmx1a) and neuronal synaptic machinery (e.g., Grin1 and Htr6), resolved to monovalent H3K27me3. Also consistent with the commitment of the cells to the GABAergic lineage, we did not detect expression of genes important for other neuronal subtypes, including many of those bivalent in NSCs (supporting information Table S5). These data show that in newly differentiated neurons, changes in bivalency reflect the restriction in developmental potential as NSCs differentiate into GABAergic neurons.

Retention and Gain of Bivalency During Neuronal Differentiation

Although the majority of bivalent promoters in NSCs resolved to monovalency during differentiation, 133 remained bivalent, while 73 de novo loci gained bivalency (Fig. 5; supporting information Data). These genes largely remained silent and were members of GO categories associated with neuronal maturation and development, including “metal ion transport,” “axonogenesis,” and “cation transport” (supporting information Table S6). These included the voltage-gated potassium channels Kv1.1/1.3, the sodium/calcium exchanger Nckx4, and the neuronal specific calcium sensors synaptotagmin II and CaBP7, together with transcriptional regulators of neuronal differentiation (Pou4f2, Rgs6, Lrp2 [supporting information Fig. S11], and Necdin), and regulators of neuronal morphology, migration, and neurite outgrowth (including the Nogo receptor Rtn4r, the ephrin receptor Efna4, and the netrin receptor Unc5d). Similarly, genes whose promoters gained bivalency de novo encode proteins whose functions are commonly associated with neuronal migration, pathfinding, and synaptic function (supporting information Table S7); these included members of the cadherin family (Cdh8, Cdh20, and Pcdhb2), ephrin, and reelin receptors (Epha4 and Vldr), semaphorin 6d (Sema6d), cell adhesion molecules (Cadm3 and Icam5), neuronal growth proteins (Dpysl5 and Csmd1), and the chemokine receptor gene, Cxcr4, all known to be crucial for axonal development and interneuron migration in the cortex [34]. The group of genes with de novo bivalency was also enriched for those involved in synaptogenesis and regulation of synaptic function, including neuregulin (Nrg1), the G-protein regulator Rgs9bp, and the synaptic modulator Tap1 (and for several synaptic components including the sorting protein SorCS2, potassium channels/subunits [Dpp10 and Kcna2], and the SNARE component Cadps). In fact, 40 of the 71 promoters that gained bivalency were present at genes associated with migrating neurons or mature differentiated neuronal function (Table 1). Importantly, none of these genes are expressed in the early neurons captured in this study, yet consistent with their function, all appear to be expressed in mature neurons (Allen Brain Atlas,, suggesting epigenetic changes in immature neurons predict later GABAergic specification and maturation. Intriguingly, many of these genes have yet to be directly linked to neuronal development or activity but are nevertheless associated with cellular functions common to neuronal maturation, such as migration and morphogenesis. For instance, protocadherin β1 (Pcdhb2) is a cell adhesion molecule involved in mammary tumorigenesis [35], but has yet to be linked to neural cell adhesion despite its expression in adult brain (Allen Brain Atlas). These data show that acquisition of bivalency acts as a possible predictor of the involvement of hitherto unknown proteins in the development of the mature neuronal phenotype.

Table 1. Genes that gain bivalency during neuronal differentiation are associated with functions required for a mature neuronal phenotype, including neuronal excitability, axonogenesis, neuronal migration, synaptogenesis and neural development. Several such genes have also been implicated in psychiatric disorders of the nervous system
 GeneGene nameReported function
  1. Abbreviations: cGMP, cyclic guanosine monophosphate; CNS, central nervous system; CUB, complement C1r/C1s, Uegf, Bmp1; GLI, glioma-associated oncogene family members; MDR, multidrug resistance protein; MMTV, mouse mammary tumor virus; PDZ, post synaptic density protein (PSD95), Drosophila disc large tumor suppressor (Dlg1), and zonula occludens-1 protein (zo-1); RING, really interesting new gene; TAP, transporter associated with antigen processing; TEA, transcriptional enhancer activator; WNK, with no lysine (K).

Ion channels and recfdsdvcsdeptorsDpp10Dipeptidylpeptidase 10Interacts with and traffics Kv4 channels
 Ano4Anoctamin 4Ca2+-activated Cl- channels.
 Kcna2Potassium voltage-gated channel, shaker-related member 2 
 Wnk2WNK lysine deficient protein kinase 2Regulates neuronal cation/Cl channel activity
 Gli3GLI-Kruppel family member GLI3Cortical interneuron identity
 Rgs7bpRegulator of G-protein signalling 7 binding proteinRegulates GPCR signalling
 Atp1b1ATPase, Na+/K+ transporting, beta 1 polypeptideNeuronal ion channel
 Kctd8Potassium channel tetramerization domain containing 8Auxillary GABA-R subunit
 Slc9a3r1Solute carrier family 9 (Na/H+ exchanger), member 3 regulator 1 
 Rgs9bpRegulator of G-protein signalling nine binding proteinCalcium-dependent inactivation of NMDA receptors
Adhesion, migration, axonogenesis, neurite outgrowthCadm3Cell adhesion molecule 3Cell adhesion in developing nervous system
 Wnt3Wingless-related MMTV integration site 3Neuronal differentiation and neurite outgrowth
 Nrg1Neuregulin 1Linked to SZ and bipolar. Cell–cell interactions, Synaptic plasticity
 Cdh8Cadherin 8CNS synaptic transmission and cortical migration
 Icam5Intercellular adhesion molecule 5, telencephalinDendrite development
 Foxb1Forkhead box B1Migration from caudal diencephalon to telencephalon. Memory
 Six5Sine oculis-related homeobox 5 homolog (Drosophila)Homolog of unc39 – neuronal migration and axon pathfinding
 Kank4KN motif and ankyrin repeat domains 4Cell motility
 Acer2Alkaline ceramidase 2Regulates integrin beta1 expression and cell adhesion
 Sema6dSema domain semaphorin 6DNeuronal migration and synaptogenesis
 ItpkaInositol 1,4,5-trisphosphate 3-kinase ADendrite morphology
 Cxcr4Chemokine (C-X-C motif) receptor 4Neuronal apoptosis and cortical interneuron migration
 Cdh20Cadherin 20(Motor) neuron migration. Also expressed in cortex.
 VldlrVery low density lipoprotein receptorReelin signalling
 Epha4Eph receptor A4Segregation of MGE- and preoptic interneurons in migratory streams
 St8sia5ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 5NCAM function – cell adhesion
 Bmpr1bBone morphogenetic protein receptor, type 1B(Inter)neuronal development. Axon cell guidance (in retina)
Synapse formation and plasticityPdzrn3PDZ domain containing RING finger 3Inhibition of wnt signaling. Regulates synapse formation at nmj
 Tap1Transporter 1, ATP-binding cassette, subfamily B (MDR/TAP)Synaptic plasticity
 Dpysl5Dihydropyrimidinase-like 5Development, maintenance and synaptic plasticity of Purkinje cells
 Sema6dSema domain semaphorin 6DNeuronal migration and synaptogenesis
 Nrg1Neuregulin 1Linked to SZ and bipolar. Cell–cell interactions, Synaptic plasticity
Neural developmentGli3GLI-Kruppel family member GLI3Cortical interneuron identity
 Tead4TEA domain family member 4Part of hippo pathway regulating proliferation, fate choice, and cell survival.
 Sntg1Syntrophin, gamma 1Neurotrophism
 Fstl1Follistatin-like 1Neural development
 MybMyeloblastosis oncogeneAdult neural stem cell maintenance
 Wnt3Wingless-related MMTV integration site 3Neuronal differentiation and neurite outgrowth
Memory, psychiatric and mood disorderCsmd1CUB and Sushi multiple domains 1Associated with memory and SZ
 Nrg1Neuregulin 1Linked to SZ and bipolar. Cell–cell interactions, Synaptic plasticity
 Il4raInterleukin 4 receptor, alphaMemory
 VldlrVery low density lipoprotein receptorReelin signaling
 Prkg2Protein kinase, cGMP-dependent, type IIControlling anxiety-like behavior
 Sel1l3Sel-1 suppressor of lin-12-like 3 (C. elegans)4p15--p16 candidate region for bipolar disorder and schizophrenia
 St3gal1ST3 beta-galactoside alpha-2,3-sialyltransferase 1SNP in bipolar
 Tgfbr2Transforming growth factor, beta receptor IIAntagonises wnt signaling. SZ linkage (blood levels)
 Foxb1Forkhead box B1Migration from caudal diencephalon to telencephalon. Memory

We reasoned that the functional and epigenetic connection among this group of genes argued for a common transcriptional control mechanism. Accordingly, we used PScan [36] to search the promoter regions of genes that lost or gained bivalency during differentiation. We uncovered several highly significant motifs, several of which have previously been linked to neuronal differentiation (Table 2). Intriguingly, PScan identified several transcription factors crucial in regulating the balance between proliferation and differentiation of NSCs (including E2F1 and the zinc finger proteins Zfp423 and Zfx) [37-39] and several required for neuronal differentiation (Hif1a, Ctcf, and NF-KB1) [40-42]. PScan also revealed over-representation of motifs of several transcription factors known to recruit histone modifying enzymes, such as Mizf, which recruits the methyl-CpG binding MBD2 protein, and Zeb1, which recruits histone demethylases and deacetylases [43-45]. Crucially, these transcription factor motifs were unique to gene promoters that lost and/or gained bivalency during neuronal differentiation, suggesting such genes may be co-ordinately transcriptionally regulated.

Table 2. Transcription factor motifs significantly enriched around promoters of genes that either lose or gain bivalency during neuronal differentiation. PScan was used to interrogate a region 1kb upstream of the transcription start site of genes, where a change in bivalency was observed. The number of occurrences and reported function of each transcription factor in relation to neuronal differentiation and its conserved motif are shown
Transcription factorGain/lose bivalencyOccurrencesAdjusted p valueMotifReported function
  1. Abbreviations: ESC, embryonic stem cell; HDAC, histone deacetylase; hESC, human ESC; EMT, epithelial–mesenchymal transition; NPC, neural progenitor.

E2F1Lose348/6691.59E-17imageCell cycle suppressor in mature neurons
Zfp423Lose302/6691.56E-12imageControls proliferation and differentiation of NPCs
Hif1aLose389/6693.01E-11imageRequired for neuronal differentiation of midbrain NPCs
MizfLose287/6695.18E-05imageMediates transcriptional repression by recruiting the methyl-CpG-binding MBD2
CtcfBoth278/669 lost34/73 gain7.82E-05 lost1.88E-03 gainimageRegulates functional neural development by controlling Pcdh expression
Tcfcp2l1Both301/669 lost44/73 gain1.04E-05 lost4.32E-03 gainimagePluripotency factor in ESCs, expression regulated by Jmjd1a
ZfxBoth290/669 lost35/73 gain5.07E-13 lost8.21E-03imageRegulates balance between self renewal and differentiation in hESCs
NFKB1Gain34/733.37E-02imageInitiates early differentiation of NSCs
Zeb1Lose527/6693.89E-02imageRecruits LSD1 histone demethylase and HDACs as part of CoREST-CtBP repressor and Brg1 chromatin remodeling complex during EMT transition

Taken together, these data show that bivalency provides a dynamic read-out of fate restriction and phenotypic plasticity throughout the entire developmental axis spanning pluripotency, multipotency, and neuronal commitment.


One of the cardinal features of tissue-specific stem cells is their multipotentiality. This requires genes that are instrumental for the differentiated function of potential progeny to be silenced, but maintained in a poised configuration that is amenable to specific activation or stable repression upon differentiation. Transcriptional profiles can identify patterns of gene expression to allow characterization of a stem cell, but in themselves they do not define the cellular potential or the change in potential that occurs as cells become progressively fate restricted during differentiation. In contrast, epigenetic signatures offer a precocious read-out of multipotentiality, as they appear to mark genes that have the potential to be expressed and ipso facto precede changes in gene expression that occur upon differentiation. One epigenetic feature that has risen to recent prominence is “bivalency”. The concurrent presence of H3K4me3 and H3K27me3 at the promoters of many silenced lineage-specific genes [2, 3] has been interpreted as an epigenetic signature of pluripotency in ESCs. It has also been suggested that some bivalency may be a product of stochastic low levels of gene expression emerging from lineage priming [46] in a subpopulation of cells [47]. Either interpretation lends weight to the notion that bivalent histone signatures offer a read-out of lineage potential. Our work described here suggests that the concept of bivalency extends beyond ESCs and offers an epigenetic signature of NSC neuronal potential that is dynamically regulated during their differentiation.

In NSCs, we find that most active gene promoters are marked by H3K4me3, while only a minority of silent genes are marked by H3K27me3, indicating that maintenance of gene silencing in NSCs is largely carried out by polycomb-independent mechanisms. Indeed, earlier studies have shown that differentiation of ESCs to NSCs is predominantly marked by widespread changes in DNA methylation, whereas little global change in DNA methylation occurs during subsequent differentiation of NSCs to neurons [48]. Under normal physiological conditions, NSCs only ever give rise to neural cells and in line with this developmental restriction many neural developmental genes have bivalent promoters in NSCs. This stands in marked contrast to promoters of non-neural, regulatory factors such as those of the myogenic regulatory factors MyoD, Myf5, Myf6, and the cardiac regulatory factors Nkx2–5, Mef2, Gata6, all of which are bivalent in ESCs [3, 48] but in NSCs are mainly marked by H3K27me3 only or are devoid of both H3K27me3 and H3K4me3. Although in vitro NSCs only give rise to neural cells, in vivo, one group of neural progenitors at the boundary of the neural plate and epidermis undergo an epithelial–mesenchymal transition to produce mesenchymal cells of the neural crest. Interestingly, a group of genes, all involved in development and differentiation of mesenchyme (comprising Aldh1a2, Ovol2, Sema3c, Sema3f, Fgf15, Foxc2, Alx1, Nrtn, Edn1, Eomes, Pax3, Isl1, Gdnf, Tgfb1, Ret, and Bmp2), nearly all of which orchestrate neural crest development [49], are bivalent in NSCs, and mostly resolve to H3K27me3 upon neuronal differentiation. Thus, bivalency marks developmental regulatory genes required for differentiation of both neuronal and neural crest derivatives thereby mirroring the developmental potential of NSCs in vivo. Our finding that bivalent genes in NSCs are particularly linked with neurogenesis rather than gliogenesis is also consistent with previous data showing that genes bivalent in ESCs and linked with gliogenesis are more likely to resolve to H3K4me3 than those linked to neurogenesis [3] and that promoters that gain H3K27me3 upon differentiation from ESCs to NPCs are enriched for genes associated with neurogenesis [48].

Further evidence that bivalency provides a map of neuronal developmental potential is provided by examination of those genes linked to regulation of neuronal subtype identity. We find that bivalency is not restricted to the promoters of genes regulating exclusively GABAergic specification or phenotype, despite the apparent restriction of these NSCs to generating GABAergic interneurons in vitro [7], but also marks promoters of genes regulating glutamatergic (Neurog1), cholinergic (Isl1), and dopaminergic (Lmx1a) differentiation. Upon differentiation, all such non-GABAergic promoters resolve to H3K27me3 monovalency, whereas GABAergic promoters resolve to H3K4me3 monovalency or retain bivalency. Comparison of the data produced in our study presents an interesting comparison with an earlier study that also looked at epigenetic changes during neuronal differentiation. In contrast to the investigation reported here, this study looked at changes in H3K4me2/H3K27me3 during glutamatergic neuronal differentiation. Interestingly, they reported little resolution of H3K4me2/H3K27me3 bivalency at such loci [48]. The reason for this contrast is unclear, but while it may reflect distinct differences in neuronal progenitor production, differentiation protocols and cell population purity between the two studies as well as the different epigenetic marks used to monitor bivalency, it may also illustrate differences in the dynamic acquisition, loss, and retention of these modifications in different neuronal lineages. Thus, it would appear that in NSCs, developmental genes of multiple neuronal lineages are bivalently marked, reflecting their broad developmental potential [7], and that upon differentiation this bivalency resolves in accordance with acquisition of a GABAergic fate.

Although the majority of bivalent promoters resolve during differentiation, a distinct group of promoters either retain or gain bivalency. Many of these genes are associated with neuronal maturation, migration, axonogenesis, and synaptic specification. However, whereas those gene promoters that retain bivalency in both NSCs and neurons (e.g., Atoh1, Ndn, Kif5c, Rtn4r, Pou4f2, and Efna5), encode guidance and adhesion molecules involved in morphogenesis of a diverse array of neuronal types, those that become bivalent de novo upon neuronal differentiation are, to the best of our knowledge, specifically associated with migration and axonogenesis of interneurons (e.g., Epha4, Cxcr4, Dpysl5, Bmpr1b, Nrg1, and Gli3). Thus, it would appear that pan-neuronal genes required for mature neuronal function are bivalently marked before the promoters of subtype specific genes, which only become marked upon initiation of GABAergic differentiation. This provides an interesting reflection of in vivo acquisition of GABAergic identity; although GABAergic interneuron fate becomes apparent relatively early during embryonic development [50], GABAergic interneurons undergo substantial maturation after cortical entry and throughout the first few weeks of postnatal life [51, 52]. During this time, interneurons respond to local cues and develop characteristic morphologies, electrophysiological properties and synaptic connections appropriate for their inhibitory function. Taken together, these observations suggest a step-wise acquisition and resolution of bivalency that precedes and predicts progressive fate restriction in NSCs.

Our data support the notion that restriction and refinement of bivalency occurs at multiple stages of development. The first restriction can be seen during the transition from ESCs to NSC where key developmental regulators associated with all three germ layers are bivalent in ESCs [3] but in NSCs, only those associated with neuroectoderm or neural crest retain bivalency or resolve to H3K4me3 (Tal1, Emx1, Neurog1, Otx2, Nkx2-2, Astn, Olig1, Dlx5, Bmp6, Bmp2, Sox9, and Cebpa). Meanwhile, those associated with formation of mesoderm, endoderm, and non-neural ectoderm either lose both H3K4me3 and H3K27me3 or resolve to H3k27me3 in NSCs (Brachyury, Foxd3, Cdx4, Myod1, CD34, Cnfn, Gata6, Pdx1, FoxA2, Hist1h2aa, Sycp2, Sp7, Pnarg, Col2a1, and Runx1). As NSCs differentiate to early neurons another restriction occurs, as bivalency is lost from promoters associated with neural crest and non-GABAergic development and function. The final refinement is the gain of bivalency at promoters of genes associated with neuronal migration, maturation and connectivity. At all stages, these changes in epigenetic signature occur before any subsequent gain of gene expression underlining the ability of the epigenome to provide a read-out of developmental potential. It is likely that further studies may reveal key intermediate and subsequent stages of epigenetic refinement that will indicate further nuances of stage-specific neuronal development and the next goal will be to refine and develop methods to permit parallel studies using in vivo cell populations.


In summary, we propose that bivalency marks promoters of genes concerned, in the broadest sense, with phenotypic plasticity at all stages of neural development. Thus, in multipotent progenitors, bivalency is largely manifest at the promoters of lineage-specific regulators while in early neurons, bivalent promoters are found at genes important for regulating establishment and maintenance of neuronal connectivity.


This work was supported by grants from the Wellcome Trust (to N.J.B.) and a KCL/NUS partner award (to N.J.B.), a MRC UK PhD studentship (to M.B.), and grants from the Agency for Science and Technology Research (Singapore; to L.W.S.). The original Tau-Blasticidin cassette used to generate NS-TRB neural stem cells was the kind gift of Dr. Steve Pollard (University College London) and the Redstar sequence was cloned into this cassette by Dr. Yuh-Man Sun.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.