During spinal cord development, the combination of secreted signaling proteins and transcription factors provides information for each neural type differentiation. Studies using embryonic stem cells show that trimethylation of lysine 27 of histone H3 (H3K27me3) contributes to repression of many genes key for neural development. However, it remains unclear how H3K27me3-mediated mechanisms control neurogenesis in developing spinal cord. Here, we demonstrate that H3K27me3 controls dorsal interneuron generation by regulation of BMP activity. Our study indicates that expression of Noggin, a BMP extracellular inhibitor, is repressed by H3K27me3. Moreover, we show that Noggin expression is induced by BMP pathway signaling, generating a negative-feedback regulatory loop. In response to BMP pathway activation, JMJD3 histone demethylase interacts with the Smad1/Smad4 complex to demethylate and activate the Noggin promoter. Together, our data reveal how the BMP signaling pathway restricts its own activity in developing spinal cord by modulating H3K27me3 levels at the Noggin promoter.
During embryogenesis, multipotent neuroepithelial precursor cells originate specialized neurons and different glial cell types (Gage, 2000; Roegiers and Jan, 2004; Temple, 2001). Whether a precursor cell either self-renews or differentiates is regulated by interactions between transcription factors and secreted signaling proteins that provide positional information (Jessell, 2000; Zhu and Scott, 2004). Bone morphogenetic proteins (BMP) are some of these extracellular proteins. In developing spinal cord, BMPs regulate several processes such as differentiation to dorsal interneurons (Liu and Niswander, 2005; Timmer et al., 2002). These differentiation processes require nuclear reorganization and general changes in gene expression, indicating that epigenetic changes may be involved (Buszczak and Spradling, 2006; Hsieh and Gage, 2005; Kondo, 2006). One of the best-illustrated epigenetic effects on regulation of pluripotency and differentiation induction is the effect mediated by Polycomb repressive complexes (PRC) (Ringrose and Paro, 2007). The hallmark for Polycomb-mediated repression is the methylation of lysine 27 of histone H3 (H3K27me3) (Czermin et al., 2002; Muller et al., 2002). Enhancer of Zeste Homolog 2 (EZH2), a subunit of the PRC2 complex, is responsible for histone methyltransferase activity (Cao et al., 2002; Czermin et al., 2002; Kuzmichev et al., 2002). This mark is recognized by the chromodomain of Polycomb protein that forms part of the PRC1 (Cao et al., 2002). The recruitment of PRC1 leads to final transcriptional repression that is reversible by JMJD3 and UTX demethylase activity (Agger et al., 2007; De Santa et al., 2007; Lan et al., 2007; Lee et al., 2007). The balance between methyltransferase and demethylase activity gives a dynamic character that is reflected by the fact that many key developmental promoters are often marked by H3K27me3 (Boyer et al., 2006; Bracken et al., 2006; Lee et al., 2006; Pan et al., 2007). In addition to H3K27me3, an active modification, H3K4me, is often found at these promoters (Bernstein et al., 2006; Mikkelsen et al., 2007; Pan et al., 2007; Zhao et al., 2007). It is believed that the presence of both activating and repressive chromatin marks keeps these developmental regulators poised for rapid resolution after the appropriate stimulus is received (Bernstein et al., 2006).
Many models have been used to examine epigenetic changes that take place during cell differentiation and, in particular, to analyze the role of Polycomb complexes using embryonic stem cells (ESCs) (Boyer et al., 2006; Bracken et al., 2006; Lee et al., 2006; Pietersen and van Lohuizen, 2008). However, these systems generally do not progress toward a specific terminal cell type as it occurs in vivo. In this paper, we take advantage of an in vivo model for neurogenesis, the chick embryo neural tube, to analyze the role of H3K27me3 during nervous system development.
Here, we report an H3K27me3-mediated mechanism to edge the boundaries of BMP activity during spinal cord development. We show that expression of Noggin, an extracellular inhibitor of BMPs, is regulated by H3K27me3. As a consequence, a reduction of H3K27me3 disturbs BMP-regulated dorsal spinal cord development. Moreover, we show that H3K27me3-mediated Noggin repression is sensitive to changes in BMP activity. Hyperactivation of BMP signaling pathway induces JMJD3 interaction with Smad1/Smad4 and their recruitment to the Noggin promoter for H3K27me3 demethylation that triggers Noggin expression. Our results reveal an essential role of H3K27me3 in the negative-feedback regulation of BMP signaling that guarantees proper neurogenesis in developing chick neural tube.
MATERIALS AND METHODS
Plasmids and recombinant proteins
Human EZH2 and its deleted form lacking SET domain (amino acids 622-707) were cloned from pCDNA3 (Caretti et al., 2004) into pCIG vector (Megason and McMahon, 2002), upstream of an internal ribosomal entry site (IRES) and three nuclear localization sequences-tagged EGFP. Human Myc-JMJD3 and Myc-JMJD3DN (H1390A truncated JMJD3) (Xiang et al., 2007) were cloned into pCIG. Mouse BMP4/7, chick Shh (Roberts et al., 1998) and human FlagSmad1 (Liu et al., 1996) were also cloned into pCIG. Mouse Wnt1, chick Noggin and human HaSmad4 are described elsewhere (Alvarez-Medina et al., 2008; Garcia-Campmany and Marti, 2007). DNA sequences for chick Noggin shRNA and a random shRNA (shRNA C-) were cloned into pSHIN vector (Kojima et al., 2004): Noggin shRNA target sequence, 5′-GTGGCTCTGGTCCCAGACC-3′; random sequence, 5′-GCTCCACATCAGTCCTTCC-3′
Chick in ovo electroporation
Eggs from White-Leghorn chickens were incubated at 38.5°C and 70% humidity. Embryos were staged following Hamburger and Hamilton (HH) (Hamburger and Hamilton, 1992). Chick embryos were electroporated with purified plasmid DNA at 0.25-3 μg/μl in H2O with 50 ng/ml of Fast Green. Plasmid DNA was injected into the lumen of HH10 or HH16 neural tubes, electrodes were placed at both sides of the neural tube and embryos were electroporated by an IntracelDual Pulse (TSS-100) electroporator delivering five 50 ms square pulses of 20-25 V.
Antibodies used were: anti-BrdU, anti-Pax7, anti-ISL1/2 (DSHB), anti-neural β-Tubulin III (Tuj1), anti-Pax6 (Covance), anti-LHX2/9 (from TM Jessell), anti-trimethylH3K27, anti-acetylH3 (K9 and K14), anti-acetylH4 (K12), anti-di/trimethylH3K4, anti-trimethylH4K20, anti-phosphoH3S10 (Upstate Biotechnology), anti-EZH1 (kindly provided by Dr Reinberg), anti-trimethylH3K9 (Abcam, UK), anti-HA (Sigma H6908), anti-Flag (Sigma M2), anti-phosphoSmad1/5/8 (Cell Signaling), anti-myc ChIP Grade (Abcam) and anti-myc (from Dr S. Pons).
The collected brachial regions from embryos were fixed for 2 hours at 4°C in 4% paraformaldehyde, rinsed, sunk in PBS 30% sucrose solution and embedded in OCT for sectioning in Leica cryostat (CM 1900). Sections were blocked at room temperature for 1 hour in 1% bovine serum albumin (in PBS with 0.1% Triton X-100) before overnight incubation at 4°C with primary antibodies. Finally, sections were incubated for 2 hours at room temperature with Alexa-conjugated goat secondary IgG antibodies (Jackson ImmunoResearch) and 0.1 ng/μl DAPI (Sigma). Images were captured by Leica SP5 confocal microscope using LAS-AF software. Fluorescence intensity was quantified using Leica LAS-AF software.
Histone extraction and immunoblotting
HH14 and HH25 wild-type (WT) embryos were collected and neural tubes were dissected out. For histone acid extraction, neural tubes were incubated for 30 minutes at 4°C in lysis buffer (HEPES/KOH 10 nM, MgCl2 1.5 mM, KCl 10 mM, DTT 0.5 mM, PMSF 1.5 mM, TSA 0.33 μM, HCl 0.2 M). After centrifugation for 10 minutes at 13,400 rpm (16,863 g), supernatant was collected and dialysis was carried out with acetic acid 0.1 M and miliQ H2O, using slide-A-lyzer mini dialysis units (Pierce). Histone concentration was measured with Bradford protein assay reagent and 10 μg of proteins were separated in SDS-PAGE gel. Immunoblotting was performed with standard procedures and visualized by ECL Kit (Amersham).
Bromodeoxyuridine (BrdU, 0.5 μg/ml) was injected into the chick embryo neural tube lumen 30 minutes before fixation. BrdU was detected on sections by treatment with HCl2 N for 30 minutes, NaBorate 0.1 M (pH 8.5) and incubation with anti-BrdU antibody.
In situ hybridization
Embryos were fixed overnight at 4°C in 4% paraformaldehyde, rinsed and processed for whole-mount RNA in situ hybridization, following standard procedures using ESTbank probes for chick EZH2, JMJD3, Hes5, NeuroD, NeuroM, Noggin, ID1 and ID3. After hybridization, embryos were post-fixed in paraformaldehyde 4% for 2 hours, embedded in sucrose 10%, agarose 5% solution and sectioned in a Leica vibratome (VT 1000S).
Fluorescent associated cell sorting (FACS)
Electroporated embryos were dissected out and trypsinized for 5-10 minutes in Trypsin-EDTA 0.5% (Sigma). Trypsinization was stopped with 20% horse serum in PBS/0.1% glucose solution. GFP+ cells from cell suspension were sorted by flow cytometry using a MoFlo flow cytometer (DakoCytomation, Fort Collins, CO, USA).
Cell cycle analysis
Trypsinized cell suspension from electroporated neural tubes was treated for 2 hours at room temperature with 10 μg/ml Hoescht 33342. Hoescht and GFP fluorescence were determined by flow cytometry using a MoFlo flow cytometer (DakoCytomation). DNA content analysis (Ploidy analysis) was carried out using Multicycle software (Phoenix Flow Systems, San Diego, CA, USA).
Fifteen chick embryo neural tubes were electroporated with empty vector or EZH2DSET for each replicates. RNA-s from 100,000 FACS-purified GFP+ cells of each replicate were supplied to the Affymetrix Facility at the IRB for quality control, quantification, reverse transcription, labeling and hybridization onto Affymetrix Chicken GeneChip. Results in CEL files were provided to the Bioinformatics and Data Analysis service of UB-SCT, for analysis of data quality, normalization with the RMA algorithm and selection of differentially expressed genes (absolute fold change >1.5 in both replicates).
mRNA extraction and quantitative PCR (qPCR)
mRNA from FACS-separated cells or from dissected neural tubes was extracted by TRIZOL (Invitrogen) protocol with 2 μl of pellet paint co-precipitant (Novagene). Reverse transcription was performed with Transcriptor Kit (Roche), following the manufacturer's procedure. qPCR was performed with Sybergreen (Roche) in LC480 Lightcycler (Roche). GAPDH was used for normalization. Primer sequences: Noggin FW, 5′-GCTACAGTAAAAGGTCTTGCTC-3′; RW, 5′-CCTCAGGATCGTTAAATGCAC-3′; Hes5 FW, 5′-TAAACCATAACTCAAGCTGTGTAG-3′; RW, 5′-GCATACATATCCTGAACCTCAC-3′; NeuD FW, 5′-ACTACTTGTTACCTTTCCCG-3′; RW, 5′-ATGGTTTAAAATAGGAAATCCACG-3′; Id1 FW, 5′-TGAAGGGCTGCTACTCG-3′; RW, 5′-AGGTCCCAGATGTAGTCG-3′; Id3 FW, 5′-CGACTGCTACTCCAAATTGC-3′; RW, 5′-GAAGATATAATCGATGACGTGCTG-3′.
Chromatin immunoprecipitation (ChIP)
For histone ChIPs, 300,000-400,000 cells separated by FACS, or two or three dissected neural tubes, were used. ChIPs were performed essentially as described elsewhere (Attema et al., 2007). Cells were treated with formaldehyde 1% at room temperature for 10 minutes. The reaction was stopped with 125 mM glycine. Cells were washed once in ice-cold PBS with protease inhibitors and lysed with 50 μl lysis buffer [SDS 1%, EDTA 10 mM, Tris 50 mM (pH 8), protease inhibitors 1 μg/ml, PMSF 1 mM], incubated 5 minutes on ice and diluted with 150 μl PBS before 10 minutes sonication in a Bioruptor (Diagenode, Liège, Belgium) (high power, 30 seconds ON, 1 minute OFF). Chromatin fragments (200-500 bp) containing supernatant were precleared with 8 μl of proteinA agarose/ssDNA beads (Upstate 16-157) for 30 minutes at 4°C under rotation. Supernatant was recovered by centrifugation and input was separated. The rest of the chromatin was diluted with IP buffer [Triton X-100 1%, EDTA 2 mM, NaCl 150 mM, Tris 20 mM (pH 8) and protease inhibitors] and divided into three eppendorfs. Antibodies or IgG (2 μg) were added and incubated overnight at 4°C under rotation. Antibody:protein:DNA complexes were then collected with 10 μl of proteinA agarose/ssDNA beads for 1 hour. The beads were washed with buffers I [SDS 0.1%, Triton X-100 1%, EDTA 2 mM, Tris-HCl 20 mM (pH 8), NaCl 150 mM], II [SDS 0.1%, Triton X-100 1%, EDTA 2 mM, Tris-HCl 20 mM (pH 8), NaCl 500 mM] and III [LiCl 0.25 M, NP40 1%, NaDOC 1%, EDTA 1 mM, Tris-HCl 10 mM (pH 8)], and three times with Tris-EDTA. Washed pellets were eluted with 400 μl of SDS 1%, NaHCO3 0.1 M solution and de-crosslinked overnight at 65°C. DNA was recovered by using phenol chloroform extraction and ethanol precipitation.
For non histone proteins ChIPs were performed as described above with some modifications: 400,000-500,000 cells separated by FACS or 20-25 dissected electroporated side of neural tubes, were fixed at room temperature 45 minutes with DSG 2 mM (Sigma) and 20 minutes with formaldehyde 1%. Cells were lysed with 1.1 ml lysis buffer [SDS 0.1%, EDTA 1 mM, Tris 20 mM (pH 8), Triton X-100 1%, NaCl 150 mM, protease inhibitors 1 μg/ml, PMSF 1 mM]. After immunoprecipitation, beads were washed with buffers I (SDS 0.1%, NaDOC 0.1%, Triton X-100 1%, EDTA 1 mM, HEPES 20 mM, NaCl 150 mM), II (SDS 0.1%, NaDOC 0.1%, Triton X-100 1%, EDTA 1 mM, HEPES 20 mM, NaCl 500 mM), III (LiCl 250 mM, NP40 0.5%, NaDOC 0.5%, EDTA 1 mM, HEPES 20 mM) and IV (EDTA 10 mM, HEPES 200 mM).
DNA was analyzed by qPCR with Sybergreen (Roche) in LC480 (Roche). Primer sequences are: Noggin(–2000) FW, 5′-CTTGCGATGCTTTTTGTGAC-3′; RW, 5′-CGTGGAGCAGTTTTACAGAC-3′; Noggin(–1000) FW, 5′-GGGTAGGGTGGGTTAGAAGA-3′; RW, 5′-CCAAAGCCTTTAAATTTCCTGC-3′; Noggin(+300) FW, 5′-GCTACAGTAAAAGGTCTTGCTC-3′; RW, 5′-CCTCAGGATCGTTAAATGCAC-3′; Hes5(–2000) FW, 5′-TGAAAGATTGGCAGAGGAAAC-3′; RW, 5′-GTACCCATTTCTCACTACAGC-3′; NeuD(–2000) FW, 5′-TTCCGTAATCGTGTGTACTCG-3′; RW, 5′-ATTTAGTTAATGGAAAAGACATCGC-3′.
Cell culture, transfection and co-immunoprecipitation (CoIP) assay
Hek293T cells were grown in DMEM with 10% fetal calf serum and 1% penicillin/streptomycin at 37°C, 5% CO2. Cells were transfected by a standard calcium phosphate co-precipitation protocol and harvested 48 hours after transfection. Immunoprecipitations and immunoblot analysis were performed essentially as described elsewhere (Valls et al., 2003). Immunoblotting was performed using standard procedures and visualized using ECL Kit (Amersham).
Identification of Smad binding sequences in the Noggin promoter region
Sequence comparison of the Noggin locus between human and chicken was performed using the global alignment program Shuffle-LAGAN (Brudno et al., 2003) and visualized with VISTA visualization tool (Mayor et al., 2000). Smad conserved binding sites were found using rVISTA 2.0 searches for Smad matrix from the TRANSFAC library.
Quantitative data were expressed as mean and standard deviation (s.d.). Significant differences between groups were tested by Student's t-test.
H3K27me3 global levels increase during neural differentiation
To understand the contribution of histone marks in neural development, we examined their distribution throughout early development. We studied the pattern of histone acetylation and methylation in HH14 and HH25 chick embryo neural tubes. Although in HH14 embryos the neural tube is mainly formed by proliferating neuroblasts, in HH25 embryos two zones are distinguished: the ventricular zone (VZ) formed by proliferating neuroblasts and the mantle zone (MZ) where differentiated neurons reside (Fig. 1A). First, we analyzed histone marks along neurogenesis by immunoblotting of histones purified from neural tubes. Although global histone acetylation and H3K4me2/3 levels were similar in HH14 and HH25 embryo neural tubes (see Fig. S1A in the supplementary material), a clear increase in global H3K27me3 was detected from stage HH14 to stage HH25 (Fig. 1B). This correlates not only with neurogenesis progression but also with enrichment in differentiated neurons. Then, we tested whether H3K27me3 levels were higher in differentiated neurons than in proliferating neuroblasts. By HH25 embryo neural tubes immunostaining we observed that global H3K27me3 levels were three times higher in differentiated neurons (MZ) than in neuroblasts (VZ) (Fig. 1C). These data indicate that global H3K27me3 levels increase along neurogenesis, whereas differentiated neurons accumulate in the neural tube. Similar results were observed for H3K9me3 and for H4K20me3 (see Fig. S1B in the supplementary material).
The observed global H3K27me3 level increase led us to analyze the expression of the two mayor enzymes responsible for H3K27me3: EZH2 and JMJD3. Transversal sections of HH10 and HH14 embryo in situ hybridization show that EZH2 and JMJD3 are ubiquitously expressed in the neural tube, although their expression level is higher in the dorsal region (Fig. 1D,E). At HH25 neural tubes, EZH2 and JMJD3 are highly expressed in VZ, whereas the mRNA levels of both enzymes in MZ are lower. Moreover, the VZ expression of EZH2 and JMJD3 is higher at dorsal neuroblasts, resulting in a well-defined dorsal JMJD3 domain and a dorsoventral EZH2 gradient (Fig. 1D,E). The EZH2 and JMJD3 expression pattern suggests that H3K27me3 could be more dynamically controlled in VZ neuroblasts, especially at dorsal region. However, it does not completely correlate with global H3K27me3 increase during neural differentiation. Thus, we analyzed the expression of EZH1, the histone methyltransferase (HMT) activity of which has been recently described (Margueron et al., 2008; Shen et al., 2008). Fig. S1C (see supplementary material) shows that EZH1 expression is higher in differentiated neurons than in proliferating neuroblasts at HH30 embryo neural tubes. This result suggests that, in addition to EZH2 and JMJD3, EZH1 contributes to the maintenance of H3K27me3 levels at differentiated neurons.
Maintenance of global H3K27me3 is not required for neural differentiation or for progenitor proliferation
Consistent with the observed global H3K27me3 levels increase during neural differentiation (Fig. 1B,C) recent evidences demonstrates that PcG proteins and H3K27me3 are present at promoters of many genes specific for neurogenesis (Boyer et al., 2006; Lee et al., 2006; Pietersen and van Lohuizen, 2008). Therefore, we investigated whether a global H3K27me3 increase is required for neurogenesis (although this does not necessarily reflect a requirement of H3K27me3 at a particular chromatin locus). To address this possibility, we reduced global H3K27me3 levels in the chick neural tube to analyze the effects on neural differentiation. In order to reduce H3K27me3 levels, chick embryo neural tubes were in ovo electroporated with a dominant-negative form of EZH2 (EZH2DSET), which lacks the SET domain responsible for HMT activity. First, we tested the ability of EZH2DSET to block endogenous EZH activity. Fig. 2A shows that EZH2DSET overexpression (GFP+ cells) reduces by 70% endogenous H3K27me3 levels 24 hours post-electroporation (PE). H3K27me3 is maintained at low levels for 48 hours, but it starts to recover 72 hours PE (see Fig. S2A in the supplementary material). EZH2 overexpression caused no change in global H3K27me3 levels (Fig. 2A). Similarly, no changes in global H3K9me3 or H3K4me3 levels were detected after EZH2DSET overexpression (see Fig. S2B,C in the supplementary material). Once global H3K27me3 levels were reduced in chick neural tube, we examined its role in neural differentiation by analyzing the expression of NeuroD and NeuroM, proneural genes expressed in differentiating neurons and previously identified as H3K27me3 targets in ESC (Boyer et al., 2006; Lee et al., 2006). To this end, EZH2DSET, EZH2 or the empty vector were electroporated in HH10 embryo neural tubes, when the neural tube is mainly formed by proliferating neuroblasts. No changes in the expression levels of NeuroD and NeuroM differentiation markers were detected after H3K27me3 reduction (see Fig. S3A in the supplementary material). To confirm these results, we checked the pan-neural differentiation marker Tuj. Embryos transfected with EZH2DSET and stained for Tuj does not show changes in the number of differentiated cells either at 48 hours or at 72 hours PE (Fig. 2B and see Fig. S4A in the supplementary material). These results indicate that global H3K27me3 maintenance is not essential to neural differentiation; the observed global H3K27me3 increase during neurogenesis might be a consequence of the differentiation process itself, in which extensive structural changes in chromatin are known to take place over all the genome (Keenen and de la Serna, 2009). In line with this hypothesis, other heterochromatin marks, such as H3K9me3 and H4K20me3, also increase during neurogenesis (see Fig. S1B in the supplementary material).
The high EZH2 and JMJD3 expression at the VZ suggests that a dynamic control of H3K27me3 might be important in neuroblast proliferation. Moreover, EZH2 and H3K27me3 regulate several proliferating processes (Agger et al., 2009; Bracken et al., 2003; Ezhkova et al., 2009; Varambally et al., 2002). Thus, we analyzed the function of H3K27me3 in the maintenance of the proliferating neuroblast population. To achieve this, EZH2DSET, EZH2 or the empty vector were electroporated in HH10 embryo neural tubes and the effect on neuroblast proliferation and cell cycle progression was analyzed. Immunostaining using H3S10p antibody shows that electroporated and control sides of the neural tubes have the same number of mitotic cells (see Fig. S3B in the supplementary material). We also evaluated neural tube cells entry into S-phase of the cell cycle: when electroporated embryos were pulse-labeled with BrdU, no differences were observed between EZH2DSET, EZH2 or empty vector electroporations at either 24 hours or at 72 hours PE (Fig. 2C and see Fig. S4B in the supplementary material). Finally, by GFP+ DNA content analysis, no changes in cell cycle phase distribution were observed after global H3K27me3 decrease (see Fig. S3C and Fig. S4C in the supplementary material). Taken together, these findings suggest that the maintenance of global H3K27me3 levels is not essential to neuroblast proliferation.
H3K27me3 regulates BMP activity
Next, we examined whether the observed dorsoventral expression gradient of EZH2 and JMJD3 (Fig. 1D,E) has any physiological significance in the neural tube dorsoventral pattern formation. We analyzed the development of dorsal neural populations by immunostaining of Lhx2/9 and Isl1/2 dorsal interneuron markers, after H3K27me3 reduction (by EZH2DSET overexpression). Lhx2/9 and Isl1/2 positive dorsal interneuron populations decrease in EZH2DSET electroporated side of the neural tubes (Fig. 3A). It is well established that BMP pathway is the main factor responsible for dorsal patterning of the neural tube (Liu and Niswander, 2005; Timmer et al., 2002). Then, we examined whether low levels of H3K27me3 affect BMP activity. To this end, we activated the BMP pathway by in ovo electroporation of BMP expression vector in the presence or absence of EZH2DSET. The expression pattern of Pax6/7 genes, which are known to be regulated by BMP activity (Liu and Niswander, 2005), was analyzed by immunostaining. Fig. 3B shows that BMP activity leads to a ventral expansion of Pax6/7 domains that is counteracted by H3K27me3 reduction after EZH2DSET overexpression. This result indicates that H3K27me3 is required to maintain proper BMP activity in the developing spinal cord.
BMP activity is regulated by Noggin induction via H3K27 demethylation
To understand how H3K27me3 regulates BMP activity, we performed a microarray analysis comparing GFP+ cells purified by FACS from EZH2DSET or empty vector electroporated neural tubes (Fig. 4A left panel). Differentially upregulated genes after H3K27me3 removal (by EZH2DSET electroporation) were analyzed by Gene Ontology (GO; http://www.geneontology.org/). Twenty-two upregulated genes were associated with seven GO biological process terms related to nervous system development (Fig. 4A, right panel). Interestingly, we found among them Noggin (NOG), a known BMP inhibitor (Zimmerman et al., 1996). As confirmed by qPCR, Noggin is upregulated (1.9±0.6-fold) in EZH2DSET-transfected neural tube cells (Fig. 4B), suggesting that Noggin upregulation could be responsible for the phenotype described above.
In order to examine whether changes in Noggin expression are directly associated with H3K27me3 levels, HH16-18 embryo neural tubes were dissected out and H3K27me3 ChIP analysis were performed. In addition to the Noggin, NeuroD and Hes5 promoters, which are not transcriptionally affected by H3K27me3 reduction (Fig. 4B), were also analyzed. Hes5 was used as a negative control of ChIP analysis, as it is highly expressed in neural tubes (in situ hybridization in Fig. 4C) and, thus, it is expected not to be enriched in H3K27me3. NeuroD, which is a known H3K27me3 target in ESC (Boyer et al., 2006; Lee et al., 2006) and repressed in analyzed neural tubes (in situ hybridization in Fig. 4C) was used as positive control of H3K27me3 ChIP. Fig. 4C shows that the Noggin promoter, which is transcriptionally inactive in most of the neural tube cells, is enriched in H3K27me3 nucleosomes (7.2±2.5% of input), although this enrichment is smaller than in the NeuroD promoter (22.1±7.5% of input). As expected, H3K27me3 levels at the Hes5 promoter are in the same range of mock ChIP (Fig. 4C).
Next, GFP+ cells from neural tubes electroporated in ovo with empty vector or with EZH2DSET were sorted by FACS and H3K27me3 levels at the Noggin, NeuroD and Hes5 promoters were analyzed by ChIP assays. Results in Fig. 4D show that, after EZH2DSET expression, H3K27me3 levels decrease at both the Noggin promoter (2.8±1.2 fold) and the NeuroD promoter (2.1±0.01 fold), although only Noggin is activated (Fig. 4B). Many promoters of key development regulators bear, in addition to H3K27me3, the active H3K4me mark (Bernstein et al., 2006; Mikkelsen et al., 2007; Pan et al., 2007; Zhao et al., 2007). It has been hypothesized that the combination of these marks creates a poised state that is suitable for rapid induction (Bernstein et al., 2006). To determine whether the presence of H3K4me together with H3K27me3 contributes to gene expression regulation, we tested the presence of this histone modification at the three gene promoters analyzed previously. Fig. 4C shows that in HH16-18 embryo neural tube cells, Hes5, Noggin and NeuroD promoters are enriched in H3K4me2 (Fig. 4C; 14.8±1.3, 6.±2.7, 3.1±1.5% of input, respectively). H3K4me2 levels at the Noggin promoter are similar to H3K27me3 levels, whereas at the NeuroD promoter H3K4me2 levels are clearly lower than H3K27me3 levels (Fig. 4C, right). ChIP assays analyzed by qPCR show that H3K4me2 is not affected by EZH2DSET overexpression (Fig. 4D). However, loss of H3K27me3 levels at the Noggin promoter, leads to a predominance of H3K4me2 over H3K27me3. The contrary happens at the NeuroD promoter, where H3K27me3 is maintained above H3K4me2 levels even after the H3K27me3 reduction by EZH2DSET overexpression. These ratios correlate well with Noggin and NeuroD expression levels observed after H3K27me3 reduction (Fig. 4D, right).
Noggin transcription is regulated by the BMP pathway
Next, we asked whether Noggin upregulation upon a signal induction requires H3K27me3 removal. To establish the activity responsible for Noggin regulation, BMP, Wnt and Shh signaling pathways were activated and Noggin mRNA levels were determined by qPCR and in situ hybridization. Results in Fig. 5A show a clear induction of Noggin expression 24 hours after BMP electroporation (20.19±7.6-fold BMP7; 44.0±7.6-fold BMP4). A 7.1±1.2 fold increase in Noggin expression is already observed 6 hours after BMP electroporation (see Fig. S5A in the supplementary material). In addition, electroporation of constitutively active Smad1 (one of the BMP pathway effectors) also induces Noggin expression (see Fig. S5B in the supplementary material). Next, we analyzed whether this activation is associated with changes in H3K27me3 at the Noggin promoter. To do this, HH10 neural tubes electroporated in ovo with BMP were dissected out 24 hours PE and analyzed by ChIP assays. Fig. 5B shows that BMP-induced Noggin activation correlates with 3.3±2-fold decrease in H3K27me3 and a 1.6±0.003-fold increase in H3K4me2. Consequently, the resulting H3K4me2/H3K27me3 relative levels at the Noggin promoter reach those observed at the transcriptionally active Hes5 promoter (Fig. 5B, right). Furthermore, levels of H3K4me3, a histone modification associated with transcriptionally activated genes (Barski et al., 2007), increase 2.11±0.33-fold close to transcriptional start site of Noggin after BMP induction (see Fig. S6 in the supplementary material). Together, these data suggest that active H3K27me3 demethylation of the Noggin promoter takes place upon BMP signaling activation. JMJD3 has been described as an H3K27me3-specific demethylase. Moreover, it has been suggested that JMJD3 activity could be associated with an H3K4 methyltransferase activity (Lim et al., 2009). Therefore, we wondered whether JMJD3 histone demethylase (HDM) activity is involved in BMP-dependent Noggin induction. To address this, neural tubes were electroporated with BMP together with an empty vector or a dominant-negative form of JMJD3 (JMJD3DN) that lacks HDM activity. Electroporated neural tube cells were separated by FACS in order to analyze Noggin mRNA and promoter H3K27me3 levels. Fig. 5C shows that Noggin mRNA levels are 1.7±0.2 fold lower when JMJD3DN is overexpressed, correlating with 1.8±0.2 higher H3K27me3 level at the Noggin promoter. These results suggest that JMJD3 contributes to BMP-induced Noggin expression. Then we tested whether JMJD3 interacts with the Noggin promoter upon BMP pathway activation. To achieve this, neural tubes were electroporated with an empty vector, with myc-JMJD3 or with myc-JMJD3 plus BMP. Electroporated neural tube cells were separated by FACS, and the presence of JMJD3 at the Noggin promoter was analyzed by ChIP assays using myc antibody. Results show 1.7±0.035-fold higher JMJD3 recruitment to the Noggin promoter upon BMP signaling activation (Fig. 5C). However, immunoprecipitated chromatin levels are similar for myc-JMJD3 alone and for the empty vector electroporated neural tube cells. This indicates that, in the absence of BMP hyperactivation (when Noggin is repressed in most of the neural tube cells), JMJD3 is not recruited to the Noggin promoter. In agreement with this result, JMJD3 overexpression, which effectively reduces global H3K27me3 levels (see Fig. S7A in the supplementary material), has no significant effect on Noggin expression (see Fig. S7B in the supplementary material), or on H3K27me3 levels at the Noggin promoter (see Fig. S7C in the supplementary material). Altogether, these data suggest that BMP pathway activates the Noggin promoter through JMJD3-mediated H3K27 demethylation.
Next, we sought to determine whether a direct link exists between BMP signaling pathway effectors and JMJD3. Smad1 is a BMP pathway effector (Liu and Niswander, 2005). Upon BMP signaling activation, Smad1 is phosphorylated and interacts with Smad4 to enter into the nucleus. Once in the nucleus, Smad1/Smad4 heterodimer regulates its target genes by interacting with co-activator or co-repressor proteins (Liu and Niswander, 2005). Fast BMP-induced Noggin expression (see Fig. S5A in the supplementary material) and evidence for the presence of a conserved Smad response-element at the Noggin promoter (see Fig. S5C in the supplementary material) suggest that Noggin could be a direct target of Smad1/Smad4. To test this hypothesis, HH10 neural tubes electroporated with BMP were dissected out 24 hours PE and analyzed by ChIP using pSmad1/5/8 antibody. Fig. 6A shows that BMP-induced Noggin activation correlates with endogenous phospho-Smad1 recruitment (2.3±0.3 fold increase) to the Noggin promoter. The observed effects of JMJD3 on BMP-induced Noggin expression (Fig. 5C), as well as the association of active Smad1 (Fig. 6A) and JMJD3 (Fig. 5C) at the Noggin promoter upon BMP pathway activation led us to test whether Smad1/Smad4 interacts with JMJD3. CoIP experiments indicate that JMJD3 interacts with Smad1/Smad4 complex (Fig. 6B,C,D). Fig. 6B shows that Smad1 interacts with JMJD3 only in the presence of Smad4. Smad4 binding to JMJD3 also requires the presence of Smad1 (Fig. 6C). Moreover, Smad1/Smad4-JMJD3 interaction increases after BMP pathway activation (Fig. 6D). All these data suggest that upon BMP pathway activation, Smad1/Smad4 heterodimer recruits JMJD3 histone demethylase to the Noggin promoter. This targeting results in a decrease of H3K27me3 levels that correlates with gene transcription activation.
Finally, we sought to analyze whether BMP-dependent upregulation of Noggin plays any role controlling BMP activity. To do that, HH10 neural tubes were electroporated with BMP and shRNA for Noggin (that partially reduces BMP-induced Noggin expression, see Fig. S8A in the supplementary material) or BMP and shRNA control. They were dissected out 24 hours PE and electroporated cells (GFP+) were separated by FACS for mRNA extraction. Then we tested whether the reduction of Noggin levels affect BMP activity by analyzing mRNA levels of Id3 [a well known BMP transcriptional target (Hollnagel et al., 1999)] by qPCR and in situ hybridization. Fig. 7A shows that co-electroporation of BMP and Noggin shRNA leads to a 2.2±1.1 fold higher Id3 mRNA levels compared with BMP and shRNA control electroporation. Similar results were observed for Id1 (see Fig. S8B in the supplementary material). Moreover, Noggin overexpression leads to a reduction of Lhx2/9 and Isl1 dorsal interneurons, regulated by endogenous BMP pathway activity (see Fig. S8C in the supplementary material). Altogether, these data strongly suggest that BMP-induced Noggin activation modulates BMP activity in the neural tube.
Our studies have uncovered new insights into the in vivo role of H3K27me3 mark in the context of lineage establishment within a tissue. We have shown that this epigenetic mark regulates dorsal patterning in developing neural tube by repressing the Noggin promoter. BMPs are needed for the formation of dorsal neural cell types (Liu and Niswander, 2005). Thus, regulation of BMP activity by expression of BMP inhibitors plays an important role in this process. Our studies suggest a model in which Noggin, a known BMP antagonist, is regulated by an H3K27me3-dependent mechanism in the developing spinal cord (Fig. 7B). Noggin is repressed in most of the developing neural tube cells. This repression requires H3K27me3 at the Noggin promoter (Fig. 7B, 1). In response to high BMP activity, Smad1/Smad4 and JMJD3 are recruited to the Noggin promoter, which in turn demethylates this promoter (Fig. 7B, 2). A decrease in H3K27me3, together with and increase in H3K4me, leads to Noggin full induction (Fig. 7B, 3), which in turn rapidly moderates high BMP activity (Fig. 7B, 4). This may occur in the most caudal developing spinal cord, in which Noggin is expressed at dorsal cells to finally fine-tune levels of BMP activity along the anterior-posterior axis of the neural tube (Fig. 7B).
Interestingly, EZH2 and JMJD3 expression is higher at dorsal cells at the analyzed stages of neural tube development. This coincides with both high BMP (Sela-Donenfeld and Kalcheim, 2002) and dorsal Noggin expression (Fig. 7B) in the developing neural tube. These observations support the role of H3K27me3 in the BMP-dependent generation of dorsal neural subtypes described in our study. On the other hand, both EZH2 and JMJD3 are expressed in the VZ of the neural tube, and their mRNA levels are reduced in the MZ occupied by differentiated neurons. The loss of EZH2 expression during differentiation is a common feature, as it has been described in in vitro neural differentiation models (Sher et al., 2008) or during epidermal development (Ezhkova et al., 2009). Nonetheless, it does not explain the observed global H3K27me3 increase during neural differentiation. These data suggest that, in addition to EZH2 and JMJD3, other enzymes might be responsible for global H3K27me3 in the developing neural tube. According to this, high global H3K27me3 levels found in differentiated neurons could be maintained by observed high EZH1 expression at differentiated neurons. This correlates with the data supporting that EZH1 is the main responsible for HMT activity in differentiated tissues (Margueron et al., 2009).
In agreement with the proposed role for H3K27me3 in development, our studies have uncovered a distinct promoter behavior in response to H3K27me3 removal that might be crucial to determining the promoter activity in response to developmental decisions. Our data show that some H3K27me3 target genes are not activated after H3K27me3 removal (e.g. NeuroD). This observation suggests a context-dependent H3K27me3 function that might rely on the targeting of sequence-specific transcription factors in response to different pathway activation. H3K27me3 mark might ensure that further developmental decisions are firmly controlled by robust induction signals. In agreement with this, different Polycomb and H3K27me3 targets have been identified in transformed human cells (Bracken et al., 2006; Squazzo et al., 2006), undifferentiated stem cells (Pasini et al., 2007) and human T cells (Barski et al., 2007; Roh et al., 2006) in which different signaling pathways operate.
Another possibility for the lack of transcription after EZH2DSET overexpression relates to the chromatin context of H3K27me3 target genes. The fact that Noggin, but not NeuroD, is activated after H3K27me3 removal, even in the absence of any induction signal, suggests that the presence of H3K27me3 should be combined with other chromatin features to fine-tune the transcriptional regulation. In ESCs, H3K4me active and H3K27me3 repressive marks co-exist over many of the lineage-regulatory genes that are governed by PRCs (Haudenschild et al., 2004; Liu and Niswander, 2005; Pasini et al., 2007). They are found in transcriptionally silenced genes and they poise them for activation upon a signal induction (Bernstein et al., 2006). Transcriptional activation of these genes requires an increase on H3K4me over H3K27me3 levels. Our results show that both Noggin and NeuroD promoters contain active and repressive marks. However, only Noggin is activated after H3K27me3 reduction. The ratio between H3K4me2 and H3K27me3 at the Noggin promoter is higher than at the NeuroD promoter, suggesting that this ratio (and not only the presence of both marks) might be determinant in the genes becoming active or repressed at some particular development stage. Therefore, our data suggest that H3K27me3/H3K4me2 relative levels might regulate the sensitivity of the promoter to respond to developmental decisions.
We have shown that small changes in Noggin expression, mediated by discreet changes in H3K27me3 levels of its promoter, lead to altered BMP-regulated dorsal patterning in the neural tube. Many processes of neural development, such as neural induction and axon guidance, are also regulated by BMP signaling (Liu and Niswander, 2005). These processes are temporally and spatially separated during the development of the nervous system. Thus, the duration and strength of BMP signaling are essential to correct neural development. One way to modulate BMP activity is by the transcriptional activation of its own extracellular antagonists (Haudenschild et al., 2004; Liu and Niswander, 2005). Here, we also show that Noggin expression is upregulated quickly after BMP overexpression. This upregulation is essential to moderate BMP activity, as the partial blocking of BMP-mediated Noggin induction increases transcriptional activity of Id1 and Id3 BMP target genes (Hollnagel et al., 1999). Moreover, we demonstrate that Noggin expression is directly regulated by BMP responding Smad proteins (Smad1/5/8). BMP-induced Noggin expression is accompanied by a JMJD3-dependent removal of H3K27me3 from the Noggin promoter. The identified JMJD3 interaction with Smad1/Smad4 heterodimer, suggests that they could form a complex responsible for JMJD3 recruitment to the Noggin promoter. One issue that remains to be resolved is whether this mechanism also works for other BMP target genes.
Our study has uncovered a fine-tuning regulatory mechanism of the BMP pathway where chromatin structure is implicated. We propose that the chromatin structure at the Noggin promoter allows a fast response to small BMP activity variations, thus ensuring the proper BMP levels required for nervous system development. A key issue is whether this is a developmental-stage specific mechanism that operates only during embryonic development or whether it can be a more general mechanism. The latter is supported by data showing that both EZH2 and BMPs are implicated in the development of several cancers (Haudenschild et al., 2004; Varambally et al., 2002). However, further evidence will be required to ascertain the contribution of H3K27me3-mediated BMP regulation in tumorigenesis.
We thank Dr X. de la Cruz for critically reviewing the paper. We also thank Dr Pons for reagents; Dr Sartorelli, Dr Xiang, Dr Kojima, Dr Bovolenta and Dr Briscoe for DNAs; and Dr Reinberg for anti-EZH1 antibody. This study was supported by the grants, SAF2005-01285, BFU2006-01493 and CSD2006-00049 to MAMB and BFU2007 to E.M. from the Spanish Ministry of Education and Science. N.A. was recipient of an I3P fellowship (I3P-BPD2005).
The authors declare no competing financial interests.
Competing interests statement