Author contributions: M.R.-T.: collection and/or assembly of data, data analysis and interpretation; H.M.-G.: collection and/or assembly of data, data analysis; C.P.: collection and assembly of data; A.H.: data analysis; Y.F.: collection and assembly of data; T.E.: data analysis; H.K.: provision of study material or patients; C.V.-A.: conception and design, data analysis and interpretation, manuscript revision, financial support; M.V.: conception and design, data analysis and interpretation, financial support, manuscript writing and final approval of manuscript; C.V.-A. and M.V. contributed equally to this work.
First published online in STEM CELLS EXPRESS April 2, 2009.
Cell lineages generated during development and tissue maintenance are derived from self-renewing stem cells by differentiation of their committed progeny. Recent studies suggest that epigenetic mechanisms, and in particular the Polycomb group (PcG) of genes, play important roles in controlling stem cell self-renewal. Here, we address PcG regulation of stem cell self-renewal and differentiation through inactivation of Ring1B, a histone H2A E3 monoubiquitin ligase, in embryonic neural stem cells (NSCs) from the olfactory bulb of a conditional mouse mutant line. We show that neural stem/progenitor cell proliferation in vivo and in neurosphere assays is impaired, lacking Ring1B, and their self-renewal and multipotential abilities, assessed as sphere formation and differentiation from single cells, are severely affected. We also observed unscheduled neuronal, but not glial, differentiation of mutant stem/progenitor cells under proliferating conditions, an alteration enhanced in cells also lacking Ring1A, the Ring1B paralog, some of which turned into morphologically identifiable neurons. mRNA analysis of mutant cells showed upregulation of some neuronal differentiation–related transcription factors and the cell proliferation inhibitor Cdkn1a/p21, as well as downregulation of effectors of the Notch signaling pathway, a known inhibitor of neuronal differentiation of stem/progenitor cells. In addition, differentiation studies of Ring1B-deficient progenitors showed decreased oligodendrocyte formation in vitro and enhanced neurogenesis and reduced gliogenesis in vivo. These data suggest a role for Ring1B in maintenance of the undifferentiated state of embryonic neural stem/progenitor cells. They also suggest that Ring1B may modulate the differentiation potential of NSCs to neurons and glia. STEM CELLS 2009;27:1559–1570
The generation of cell lineages during development and the integrity of tissues in the adult depends on a variety of self-renewing embryonic and adult stem cells . Self-renewal allows the maintenance of pools of stem cells needed for appropriate embryo development and cell replacement during an organism's life span . Stem cell self-renewal combines the ability to proliferate with the preservation of developmental potential . Embryonic stem (ES) and fetal stem cells proliferate extensively, in contrast to adult stem cells, which proliferate rather sparsely and rely on transit-amplifying progenitors for cell expansion [1, 4]. Developmental potential is maximal in ES cells (pluripotent), whereas other embryonic/fetal and adult stem cells show lineage-restricted potential [1, 5].
Studies on ES cells show that self-renewal is determined by a regulatory network of transcription factors acting together with chromatin regulators, thus defining a cell type–specific epigenetic landscape [6–9]. Among epigenetic regulators, the Polycomb group (PcG) of genes encodes subunits of chromatin complexes (Polycomb repressing complexes [PRCs]). Two of these subunits are endowed with chromatin-modifying activities. One, Ezh2, is a methyltransferase that modifies lysine 27 of histone H3 , whereas Ring1B and its paralog Ring1A are E3 monoubiquitin ligase that modify lysine 119 of histone H2A . Both enzymatic activities belong in nonoverlapping core protein complexes unique to a variety of biochemical entities grouped as PRC2 (H3K27 methyltransferase) and PRC1 (H2AK119 monoubiquitin ligase) . These histone modifications are commonly found on transcriptionally silent loci [13–16]. Currently, it is accepted that PRC2 acts as an instructed chromatin-modifying module , whereas PRC1 acts as a reader device that binds to H3K27 methylated nucleosomes [10, 18], although sometimes PRC1 subunits are also recruited to chromatin in a H3K27-independent manner .
The PcG role in stem cell self-renewal is best known for ES cells and hematopoietic stem cells . In ES cells, PcG products act as repressors of transcription factors and of components of signaling modules widely used in the generation of cell lineages that appear during development [13, 15]. Recently, it was shown that PRC2-dependent markers and other epigenetic markers are related to changing transcriptional programs associated with transitions from ES cell–derived neural stem cells (NSCs) to neural progenitors and their differentiated progeny . The impact of PcG regulation on NSC self-renewal is more limited and is derive mostly from the study of loss-of-function models of Pcgf4/Bmi1, a PRC1 subunit that acts as a cofactor in H2A ubiquitylation by Ring1A/Ring1B . Mice lacking Bmi1 have a lower number of adult NSCs whose self-renewal ability is severely impaired [23–25]. Acute inactivation of Bmi1, by short hairpin (sh)RNA-mediated inactivation of embryonic neural stem/progenitor cells also results in defective self-renewal . Both Bmi1-deficient adult and fetal stem/progenitor cells have upregulated inhibitors of cell cycle progression. Although Ring1B and Bmi1 are part of a common regulatory device [22, 27], they also show distinctive activities, as seen, for example, in the adult hematopoietic compartment . Here, we set out to investigate the role of the PcG histone E3 monoubiquitin ligase Ring1B in the self-renewal and differentiation potential of embryonic NSCs. Using a model of conditional inactivation of Ring1B to study self-renewal of NSCs from the fetal olfactory bulb, we show that Ring1B promotes embryonic NSC self-renewal by sustaining their proliferative activity and maintaining their undifferentiated state and developmental potential.
MATERIALS AND METHODS
Ring1B conditional knockout, Ring1A knockout, and genotyping were described previously [28, 29]. Inducible Cre-expressing mouse lines were Polr2a::CreERT2  and Rosa26:CreERT2 . Animal handling procedures were institutionally approved and were in accordance with national and European regulations.
Neural Stem/Progenitor Cell Culture
Reagents for cell cultures were bought from Invitrogen (Carlsbad, CA, http://www.invitrogen.com), Sigma-Aldrich (St. Louis, MO, http://www.sigmaaldrich.com), and PeproTech (Rocky Hill, NJ, http://www.peprotech.com). Embryonic olfactory bulb stem cells (OBSCs) were prepared from the olfactory bulb of 13.5 days postcoitum (dpc) mice. Cells were plated and expanded in Dulbecco's modified Eagle's medium (DMEM)/nutrient mixture F12/insulin, apotransferrin, putrescine, progesterone, and sodium selenite supplemented with fibroblast growth factor (FGF)-2 and epidermal growth factor (EGF) . Proliferation assays were performed in cells growing in the presence of FGF-2 and EGF. Cultures were pulsed with 5′-bromo-2-deoxyuridine (BrdU) for 1 hour before fixation. Differentiation assays were performed after mitogen removal from the cultures. For clonal analysis, OBSC spheres were dissociated into a single-cell suspension and diluted to eight cells/ml in DMEM/F12/N12:OBSC-conditioned medium (1:1). Two hundred μl of this suspension was plated into each well of 96-well plates. One day after seeding, wells containing a single cell were marked and induced to proliferate for 9 days, when the wells were screened for the presence of clonal neurospheres. Differentiated clonally derived secondary spheres were finally fixed and triple immunostained to detect the presence of neurons, astrocytes, and oligodendrocytes .
Immunostaining of Cultured Cells
Cells were fixed in 4% paraformaldehyde (PFA) and incubated with antibodies to: Ring1B (1:250 ), β-tubulin III (Tuj1, 1:2,000; Covance, Princeton, NJ, http://www.covance.com), glial fibrillary acidic protein (GFAP, 1:800; Millipore, Billerica, MA, http://www.millipore.com), O4 (1:8), microtubule-associated protein 2ab (Map2ab) (1:250; Sigma), neurofilament medium chain (NF-m) (1:400; Encor Biotechnology, Gainesville, FL, http://www.encorbio.com), and BrdU (1:1,000; Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com). Differentiated clones were triple immunostained with antibodies against β-tubulin III (rabbit antibody diluted 1:2,000; Covance), GFAP (mouse antibody, diluted 1:800; Millipore), and O4 (diluted 1:8). The O4 antibody was obtained from the culture medium of O4-producing hybridoma cells, kindly provided by A. Rodríguez-Peña, Consejo Superior de Investigaciones Cientificas. Appropriate secondary antibodies were used to detect specific signals. Cultures were stained with 2 μg/ml 4′,6-diamino-2-phenylindole (DAPI) to stain the nuclei and allow total cell counting.
A terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed following the manufacturer's instructions (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com).
Immunostaining of Tissue Sections
Embryonic day (E)17.5 mouse embryos were perfused with 4% PFA and embryo heads were postfixed in 4% PFA for 2 hours, cryoprotected with Tissue-Tek (Sakura, Tokyo, Japan, http://www.sakura.com), and then frozen at −80°C. Coronal cryostat sections (14 μm) were immunostained with antibodies against: Ring1B (1:400), phospho-H3 (PH3, 1:1,000; Millipore), Tuj1 (1:2,000), GFAP (1:800), and nestin (mouse monoclonal diluted 1:50; Millipore). The TUNEL assay (Roche's kit) was also performed on cryostat sections. Ring1B and Tuj1 immunohistochemistry was done on 7-μm sections of paraffin-embedded fixed embryos. Binding of anti-Ring1B  and anti-Tuj1 antibodies was detected with the avidin–biotin–peroxidase complex method (Peroxidase Elite, Vectastain, and ABC kit; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com) followed by counterstaining with hematoxylin-eosin. Cryostat sections were also prepared from the OBs of E17.5 embryos from pregnant mice that were injected i.p. with BrdU (100 μg/g) 4 hours before embryos were removed. Sections were double immunostained with antibodies against BrdU (1:200; Abcam, Cambridge, U.K., http://www.abcam.com) and TuJ1 or nestin (rabbit polyclonal, diluted 1:1,000, provided by R. McKay, NIH), followed by appropriate fluorescent secondary antibodies. Some sections were stained with DAPI.
Western Blot Analysis
Cells were lysed in RIPA buffer (10 mM Tris-HCl, pH 7.2; 150 mM NaCl; 1% Triton-100; 0.1% sodium dodecyl sulfate; 1% sodium deoxycholate; 5 mM EDTA, 20 mM NaF and Complete [Roche]). Thirty-five μg of total extract was subjected to SDS-PAGE electrophoresis and posterior immunoblotting following standard procedures. Nitrocellulose membranes were incubated with antibodies specific for Ring1B  and α-tubulin (Sigma, clone B-5-1-2).
Microarray Methods and Gene Ontology Analysis
Total RNA was extracted using the Trizol reagent (Invitrogen) and purified with Qiagen RNeasy separation columns (Qiagen, Hilden, Germany, http://www1.qiagen.com). First-strand cDNA was synthesized and hybridized to Affymetrix GeneChip Mouse Genome 430 2.0 arrays (Affymetrix, Santa Clara, CA, http://www.affymetrix.com) to assess and compare the overall gene expression profiles. The signal intensities of probe sets in each chip were normalized with the gcRMA algorithm  implemented in the Bioconductor package (Bioconductor, http://bioconductor.org) to estimate the expression value of each probe set in the chip. The annotation information for the probe sets was obtained from RefDIC (RIKEN Research Center for Allergy and Immunology, http://refdic.rcai.riken.jp) , and probe sets having no gene information or matching two or more genes were excluded from the analysis. We considered only probe sets whose expression values varied by twofold or more between Ring1B-deficient cells and control cultures. For 972 up/downregulated genes, gene ontology (GO) term enrichment analysis was performed according to a method described by Draghici et al. , with a correction for multiple testing using the false discovery rate (FDR) . Eventually, GO terms with a FDR < 0.05 were selected as functionally enriched terms.
Total RNA was isolated using Trizol (Invitrogen) and the reverse transcription (RT) reaction was carried out using the First Strand Synthesis Kit according to the manufacturer's guidelines (Invitrogen). A quantitative polymerase chain reaction (QPCR) analysis was performed in triplicate using 150 ng cDNA per reaction in an iQ5 with Bio-Rad SYBR-Green (Bio-Rad, Hercules, CA, http://www.bio-rad.com). Gene expression data were analyzed with iQ5 software (Bio-Rad). Sequences of primer pairs for the QPCR were obtained from the Primer Bank : Hes1, 6680205a2; Hes5, 6754182a1; Hey1, 6754188a1; NeuroD1, 33563268a1; Tubb3, 12963615a1; Cdkn1a, 6671726a3; Ina, 34328368a1; Id4, 13624309a1; Eomes/Tbr2, 5738950a1; Prdm16, 33186878a2; Ascl1, 6678806a1; Zfhx3/Atbf1, 6680736a1; Nr2f2/COUP-TFII, 6753104a1. Other primers were: Cdkn2a, 5′-CGAA CTCTTTCGGTCGTACCC-3′(forward) and 5′-CGAATCTGCAC CGTAGTTGAGC-3′ (reverse) and Cdknd, 5′-CTGAACCGCT TTGGCAAGAC-3′ (forward) and 5′-GCCCTCTCTTATCGCCA GAT-3′ (reverse).
Cell Counts and Statistical Analysis
To determine the number of cultured cells expressing a specific antigen, 10 random fields were counted per coverslip using a 40× objective and fluorescence filters . PH3+ cells and GFAP+ cells were counted in 24 serially cut cryostat sections (n = 3 wild-type mice and n = 3 Ring1BΔ/Δ mice). To count BrdU+ and TuJ1+ cells, confocal images from cryostat sections were taken every micron. The entire image stacks were then analyzed and the cells were counted using ImageJ software (Wayne Rasband, National Institutes of Health, USA). Statistical analysis was performed using Student's t-test and Sigma Stat Software (Sigma).
Ring1B Is Required for Proliferation of NSCs
Multipotent self-renewing NSCs can be propagated in cell suspension cultures, where they form cell aggregates known as neurospheres . Under these conditions, survival and expansion of stem and progenitor cells result in populations of undifferentiated cells that retain their ability to differentiate in vitro to neuronal and glial derivatives. Typically, neurosphere cultures are established from embryonic brain regions or from the adult subventricular zone and the OB [31, 41]. To investigate the role of Ring1B in NSC self-renewal, we isolated cells from the embryonic OB that are known to show multipotential and self-renewing capacities under clonal analysis conditions. These cells reside in the proliferating layers (neuroepithelial zone) of the OB and express Ring1B (Fig. 1A). We dissected the OB from 13.5 dpc fetuses homozygous for a floxed Ring1B (Ring1Bf) allele that can be inactivated by Cre-mediated recombination, a strategy needed to circumvent the embryonic lethality of Ring1B inactivation [28, 42]. This Ring1B mutant allele was obtained by homologous recombination of a targeting construct in which loxP sequences flank the first coding exon of the Ring1B gene and a selectable neo cassette (Fig. 1B) . These fetuses also ubiquitously expressed a cytoplasmic Cre recombinase–estrogen receptor fusion protein (Cre-ERT2) whose activity is controlled by translocation to the cell nucleus induced by 4-hydroxy-tamoxifen (4HT)  (Fig. 1B). NSCs were expanded in the presence of growth factors, EGF and FGF-2, and cultured in neurosphere-forming conditions, as previously described . To obtain Ring1B-deficient progenitors, we treated parallel cultures seeded at 5,000 cells/cm2 with 800 nM 4HT in ethanol, or with only vehicle, for 18 hours and scored the resulting neurospheres 72 hours later. At this stage, 4HT treatment resulted in Ring1B-depleted cell cultures, as shown by Western blot and Ring1B immunofluorescence analysis (Fig. 1C, supporting information Fig. 1A). We termed 4HT-treated and ethanol-treated cultures as Ring1BΔ/Δ and Ring1Bf/f, respectively. Ring1BΔ/Δ cultures consisted of neurospheres whose size was smaller than that of wild-type neurospheres (Fig. 2A, left, and also the clonal analysis in Fig. 3). This observation was consistent with a poorer expansion of Ring1B-deficient cells, as indicated by the total number of cells in mutant versus wild-type cultures under proliferating conditions (Fig. 2A, right). To eliminate the possibility that transient nuclear activity of the recombinase Cre had a toxic effect on cell growth, we characterized the proliferative properties of neurospheres derived from Cre-ERT2::Polr2a mice and found that the proportion of BrdU-incorporating cells in tamoxifen-treated cultures was similar to that in control, ethanol-treated cultures (supporting information Fig. 1B). We also investigated the possible role of 4HT treatment characterizing neurosphere cultures from OB cells isolated from wild-type mice and found them to be indistinguishable from vehicle-treated neurospheres in cell proliferation assays (supporting information Fig. 1C). Thus, these experimental conditions had no effect on the observed results and, therefore, we conclude that the appropriate proliferation of neural stem/progenitors depends, in a cell-autonomous manner, on the activity of Ring1B. We then studied whether the decrease in cell numbers in mutant cultures was due to differences in apoptosis or in proliferative rates. The TUNEL analysis showed almost no apoptotic cells in either culture (0.3% and 0.2% of total cells in Ring1BΔ/Δ cultures and wild-type cultures, respectively; data not shown). In contrast, quantitation cultures that were BrdU pulsed for 1 hour demonstrated that 4HT-treated cultures contained fewer cells that had incorporated BrdU (19%) than cultures treated with ethanol (12%; p < .005) (Fig. 2B). To see whether this observation on OB-derived NSCs could be generalized to NSCs from other regions of the central nervous system, we investigated the proliferative properties of neurospheres derived from the cortex and ganglionic eminences of 13.5 dpc embryos. The results (supporting information Fig. 2) showed that, in the absence of Ring1B, neurosphere formation and proliferation, assessed as BrdU incorporation, were lower than in Ring1Bf/f cultures. Collectively, these results show that defective expansion of mutant neural stem/progenitors results from a decrease in the proliferative rate of Ring1B-deficient cells.
Defective Self-Renewal of Ring1B-Deficient NSCs
The nonclonal neurosphere assay assesses the presence of NSCs but it does not inform on their self-renewal capacity. Instead, self-renewal is usually analyzed by the ability of single cells to reform spheres while retaining the production of neurons and glia. To determine whether Ring1B plays a role in NSC self-renewal, we quantitated the number of cells able to reform neurospheres when cultured as single cells. First, we treated single cell cultures with 4HT and cultured them for 7-9 days to obtain primary Ring1BΔ/Δ neurospheres (Fig. 3A, top). Efficient inactivation of Ring1B in these cultures was assessed by PCR analysis of single spheres (supporting information Fig. 3A). Second, we set up single-cell cultures from neurospheres derived from Ring1B-deficient cells to derive mutant secondary neurospheres (Fig. 3A, bottom). We also tried to establish neurospheres from two stages of single-cell cultures, but under these conditions no neurospheres were obtained from mutant OBSCs (supporting information 3B, 3C). The assays revealed that the capacity of Ring1BΔ/Δ cells to give rise to primary neurospheres was one third lower (p < .001) than that of wild-type cells, and that such a capacity was further reduced in secondary neurospheres, to only one fifth of that of wild-type cells (p < .001) (Fig. 3B). Sizes of primary and secondary mutant neurospheres from the clonal analysis were, as already seen in nonclonal neurosphere assays, smaller than those of wild-type cells (46% and 63% of that of wild-type spheres) (Fig. 3D, 3E). The comparison of neurosphere sizes obtained in these clonal assays showed a distinctive distribution in which wild-type cells generated neurospheres in the range of 200-1,800 microns in diameter, whereas mutant cells never reached diameters >800 microns (Fig. 3C). Altogether, the data show that Ring1B is required for self-renewal and proliferation of OB stem cells.
Premature Differentiation of Ring1B-Deficient NSCs
To address whether the impaired proliferation and self-renewal abilities of mutant NSCs is related to Ring1B regulation of differentiation, we evaluated the potential of E13.5-derived OB neurospheres to generate the three major neural cell lineages normally found when they are cultured under differentiation conditions, without mitogens. We found that both Ring1B-deficient and wild-type cultures formed similar numbers of neurons (Tuj1 immunoreactive cells) and astrocytes (GFAP immunoreactive cells), as shown in Figure 4. In contrast, we found a consistently lower number of oligodendrocytes (O4 immunoreactive cells) in cultures that lacked Ring1B. We tested the possibility that mutant cells were counterselected during their time in culture by identifying Ring1B-expressing cells with anti-Ring1B antibodies in cultures undergoing differentiation. The results (supporting information Fig. 4A) showed that, whereas all cells in control cultures were stained for Ring1B, 4HT-treated cultures lacked Ring1B immunoreactivity. Because the differentiation potential of NSCs can change with developmental time, we also addressed the phenotype of neurospheres derived from 18.5 dpc OBs. Similarly to observations for Ring1B-deficient 13.5 dpc NSCs, mutant 18.5 dpc OBSCs showed impaired neurosphere formation and decreased proliferation (supporting information Fig. 5A), compared with that of untreated cultures. In the absence of growth factors, both 4HT-treated or untreated cultures generated similar proportions of the three cell lineages, although a (statistically not significant) trend toward a lower number of oligodendrocytes was observed for Ring1BΔ/Δ cultures (supporting information Fig. 5B). Considering that most differentiated derivatives from wild-type cells are neurons and astrocytes, the overall differentiating potential of Ring1B-deficient cells growing under nonclonal conditions was similar to that of Ring1Bf/f cells after induction of differentiation. To better ascertain the differentiation potential of mutant stem cells, single cell–derived neurospheres were induced to differentiate, revealing a lower (30% of that of Ring1Bf/f cells) number of tripotent clonal neurospheres (Fig. 4C, 4D). Also, the overall differentiation potential of mutant clonally derived neurospheres was lower (50% of that of Ring1Bf/f cells).These findings unveil a defective differentiation potential of Ring1BΔ/Δ OBSCs that was masked under nonclonally derived neurospheres in nonclonal neurosphere cultures.
We next asked whether Ring1B might play a role in maintaining neural stem/progenitor cells undifferentiated, as has been observed in ES cells [15, 44]. To this end, we analyzed 4HT-treated and untreated OB stem cell cultures growing in the presence of mitogens for the expression of the neuronal differentiation marker β-tubulin III (Tuj1). We found that Ring1BΔ/Δ cultures contained a larger fraction (13%) of immunoreactive cells than wild-type cultures (1.4%) (Fig. 5A), suggesting that Ring1B deficiency leads to a weakened stability of the undifferentiated state of mutant cells. If, indeed, Ring1B is playing a role in the maintenance of the undifferentiated state, we reasoned that a differentiated phenotype would be better observed in stem/progenitor cells derived from compound mutant embryos in which both Ring1B and its paralog Ring1A are inactivated. Figure 5B shows that more than one third of Ring1A−/−Ring1BΔ/Δ OB stem/progenitor cells growing in the presence of FGF and EGF-2 were immunoreactive for Tuj1, compared with <1% among those in single Ring1A mutant cell cultures. In addition, examples of cells possessing clear neuronal morphologies could be seen among Ring1A- and Ring1B-lacking cultures (Fig. 5B, inset). This was corroborated by double immunostaining of TuJ1+ cells with Map2ab or NF-m, neuronal markers that label dendrites and axons, respectively (Fig. 5C, 5D). Neuronal differentiation of Ring1BΔ/Δ stem/progenitor cells, however, was not accompanied by glial differentiation, assessed as GFAP immunoreactive cells (data not shown). Differentiation, in the absence of mitogens, of Ring1A−/−Ring1BΔ/Δ OB stem/progenitor cells, showed a pronounced phenotype with neurons possessing long processes, fewer astrocytes and the absence of oligodendrocytes (supporting information Fig. 6). Together, the results suggest that Ring1B (and Ring1A) contribute to the stability of the undifferentiated state of neural stem/progenitor cells.
Gene Expression Analysis of Ring1B-Deficient OBSCs
Genome-wide mRNA profiling of OB-derived neurospheres under proliferating conditions by microarray analysis showed 972 probes derepressed or repressed more than twofold in Ring1BΔ/Δ cultures compared with control spheres. GO analysis showed an enrichment of categories related to neuronal differentiation and to cell proliferation (Fig. 6A, supporting information Table 1). For instance, Neurod1, a transcription factor specific for neurogenesis , Tubb3, the Tuj1 antigen, and Ina (internexin neuronal intermediate filament protein, alpha), well-known markers of neuronal differentiation, were among the genes derepressed in Ring1B-deficient cells. Also, effectors of the Notch signaling pathway, such as Hes5, were repressed in mutant cells, which is consistent with previous observations showing the large impact that their inactivation has on the self-renewal of NSCs [46, 47]. Among cell cycle regulators, Cdkn1a, an inhibitor of cyclin-dependent kinases (CDKs), which regulate progression through the G1 phase of the cell cycle , was the one more clearly derepressed. To validate the microarray analysis data, we carried out QPCR on a number of select targets (Fig. 6B). Among transcriptional regulators with relevance in neural progenitor differentiation, we studied Neurod1, Eomes/Tbr2, Id4, Prdm16, Ascl1/Mash1, Zfhx3/Atbf1, and Nr2f2, all of which were upregulated in mutant neurosphere cultures, except Ascl1/Mash1. Genes encoding neuron-specific filaments, Tubb3 and Ina, were also derepressed. Of the CDK-encoding genes analyzed (data not shown for Cdkn1c/p57, Cdkn2b/p15, and Cdkn2c/p18), only Cdkn1a/p21 was confirmed to be upregulated. Finally, transcripts for the Notch effectors, Hes5, Hes1, and Hey1, were found to be repressed in an RT-QPCR analysis of Ring1B-deficient neurosphere cultures (Fig. 6B). These data show that, in the absence of Ring1B, neural progenitors upregulate an inhibitor of cell cycle progression, attenuate Notch signaling, and cannot maintain their undifferentiated state, all of which can contribute to the impaired self-renewal and reduced proliferation abilities observed in Ring1B-deficient OB stem/progenitor cells.
Considering that Ring1B is a transcriptional repressor, it appears likely that repressed genes in mutant cells are indirect targets of Ring1B. To distinguish between direct (Ring1B bound) and indirect (not Ring1B bound) targets, we used a chromatin immunoprecipitation (ChIP) on chip approach. Supporting information Figure 7A shows the results on a subset of chosen genes. Whereas the promoter regions of a number of genes derepressed in Ring1BΔ/Δ cultures were bound by Ring1B (Neurod1, Id4, Eomes, Prdm16, Nr2f2), other upregulated loci (Tubb3, Ina, Atbf1, or Cdkn1a) showed very little bound Ring1B, suggesting that their derepression is secondary to Ring1B inactivation. Of interest is the fact that despite Ring1B occupancy of promoter regions of Neurog1 and Neurog2, two important regulators of neural development, their expression was found unaltered in Ring1B-deficient cells. On the other hand, no Ring1B was found recruited to the chromatin of genes encoding Notch pathway members, such as Hes1, Hes5, Hey1, and Dll3 (supporting information Fig. 7A), all of which were repressed in Ring1B-deficient neurospheres.
Analysis of chromatin modifications by ChIP analysis on select targets (supporting information Fig. 7B) revealed increases in the promoter proximal levels of H3K4me3, a histone marker typically associated with transcriptional activation in mutant cells. Levels of H3K27me3, a histone marker usually associated with transcriptional repression, however, showed little variation between mutant and control cultures. On the other hand, Ring1B inactivation lead to only a partial reduction in global levels of monoubiquitylated histone H2A (supporting information Fig. 7C), and complete depletion was only observed when both Ring1A and Ring1B were deleted (supporting information Fig. 7C). The data suggest that gene expression changes in directly regulated Ring1B targets are accompanied by chromatin modifications and that Ring1A partially compensates for the absence of Ring1B in mutant neurosphere cultures. Overall, the data suggest that Ring1B contributes to the maintenance of self-renewal and modulation of developmental potential of OBSCs throughout an intricate network of regulatory events that may be under both direct and indirect control by Ring1B.
In Vivo Altered Proliferation/Differentiation of Neural Progenitors in Ring1B Mutant Embryos
To assess the impact of Ring1B inactivation in an in vivo setting, we analyzed fetuses 4 days after tamoxifen treatment of gestating females, of which some received a dose of BrdU 4 hours before perfusion. For these experiments, we used Ring1Bf/f females that were mated to Ring1Bf/f, Rosa26::CreERT2 males, so that mutant (Cre-expressing) and wild-type (without the Cre-expressing transgene) embryos were obtained in the same litter. Efficient Ring1B inactivation assessed by immunohistochemistry is shown in supporting information Figure 4B. Proliferation was evaluated by staining with an antibody that recognizes the phosphorylated form (at serine 10) of histone H3 (PH3), a mitotic marker. We found a lower number of labeled cells (54.5%; p < .001) (Fig. 7A), which agrees with the reduced proliferation seen in neurosphere assays. These PH3+ cells expressed nestin (supporting information Fig. 8), indicating their neural stem/progenitor nature. The numbers of PH3+ cells were also lower (25% lower than in controls; p < .05) in the cortex of Ring1BΔ/Δ 17.5 dpc fetuses (sections, n = 16 from two animals of each genotype). This reduction in mitotically active cells was still more pronounced in 17.5 dpc Ring1A−/−Ring1BΔ/Δ embryos than in controls (supporting information Fig. 9).
We also analyzed differentiation in Ring1BΔ/Δ embryos. Gliogenic differentiation was analyzed in 17.5 dpc embryos, at the beginning of astrocyte formation in the developing embryo. The sections, stained with an anti-GFAP antibody, showed a lower number of stained cells in mutant embryos than in wild-type embryos (41% of those in wild-type embryos; p < .001) (Fig. 7B). In addition, using an alternative assay that allows the identification of differentiating, unexpanded, progenitors (32), we found a 50% lower (p < .05) number of GFAP+ cells in 17.5 dpc mutant OBs, compared with controls (n = 9 control cultures and n = 7 Ring1BΔ/Δ cultures). In contrast, neurogenesis, assessed by immunostaining with anti-Tuj1 antibodies, was found to be augmented in mutant OBs, which showed greater numbers of Tuj1+ cells (Fig. 7C, 7D). To test if the increase in TuJ1+ cells and the reduction in GFAP+ cells were due to changes in the proliferation of OB progenitor cells in vivo and/or to an increased tendency of progenitors towards neurons, dividing cells were labeled with BrdU and their differentiation was examined 4 hours later. As seen in Figure 7D and 7E, the number of BrdU+ cells was 50% lower in sections from mutant embryos, whereas the fraction of BrdU+ cells that expressed TuJ1 was significantly greater than that of control embryos, suggesting that neuronally biased differentiation may be responsible for the lower numbers of GFAP+ cells. A similar analysis to identify doubly BrdU+ and GFAP+ cells could not be done because the acid treatment of sections needed to detect BrdU interferes with the weak GFAP immunostaining of sections at this developmental stage. The impaired proliferation of neural progenitors in mutant embryos, the augmented neuronal generation, and the decreased glial differentiation establish Ring1B as a promoter of proliferation of embryonic neural progenitors and as a modulator of their differentiation potential in the developing OB.
Self-renewal of stem cells involves the coordination of signaling pathways that ensure cell replication while maintaining their differentiating potential. We report here that inactivation of Polycomb subunit Ring1B impairs proliferation and self-renewal of embryonic OB neural stem/progenitor cells, concurrently with the loss of developmental potential and accelerated neuronal differentiation. Our observations suggest that the self-renewing function of embryonic neural stem/progenitor cells uses Polycomb epigenetic regulation to balance cell cycle control and differentiation commitment.
Self-Renewal and Proliferation
Major differences between the self-renewal dynamics of embryonic/fetal and adult stem cells lie in their programs of proliferation control . Thus, fetal stem cells proliferate actively and expand their populations throughout development, whereas proliferation of adult stem cells is tightly regulated, dividing rarely, as demanded by tissue homeostasis. The Polycomb role in self-renewal of NSCs is mostly known through the analysis of Bmi1-mutant adult mice, which show a lower number of neural stem/progenitor cells whose self-renewal ability is also impaired. These defects are thought to result from premature senescence induced by the accumulation of products of the Ink4a locus, which is normally repressed by Bmi1 [23–25], that are responsible for decreased neurogenesis during ageing .
Our studies on neural embryonic stem/progenitor cells show that their proliferation, both in neurosphere assays and in the OB, depends, at least in part, on Ring1B. Furthermore, self-renewal, as assessed in secondary neurospheres under clonogenic conditions and in two stages of single cells, is impaired in the absence of Ring1B. These results are in line with those resulting from acute inactivation of Bmi1 in embryonic stem/progenitor cells expressing Bmi1 shRNAs . In both cases, loss of function of either Bmi1 or Ring1B (this report) leads to little or no upregulation of the products of the Ink4a locus. Instead, Cdkn1a/p21, an inhibitor of the CDKs that determine progression of the cell cycle through the G1 phase, is upregulated. A difference, however, between Bmi1 and Ring1B activities promoting fetal NSC self-renewal appears to be suppressing apoptotic programs, which seems to depend more on Bmi1  than on Ring1B, given the indistinguishable apoptotic rates of wild-type Ring1B mutant progenitors. In this regard, it seems peculiar that a set of deregulated genes in Ring1B-deficient OB-derived neurospheres belongs in the GO term of positive regulation of apoptosis. On the other hand, negative regulation of Cdkn1a/p21 by Polycomb products may occur through indirect mechanisms because, at least for Ring1B, we found no evidence of association with gene promoter regions. Cell cycle control of adult neural stem/progenitor cells is controlled, at least in part, by CDKIs encoded by Ink4a and Cdkn1a/p21 [24, 50, 51]. However, our data, together with previous observations such as the growing impact of Bmi1 inactivation on self-renewal of developing neural progenitors while Cdkn1a/p21 upregulation decreases over time  and the lack of p16Ink4a upregulation in the fetal brain from Bmi1−/− mice  and fetal Ring1BΔ/Δ OB-derived neurospheres (Fig. 7B), suggest that negative regulation of Cdkn1a/p21 is a prevalent mechanism during expansion of embryonic neural stem/progenitor cells.
Maintenance of Differentiation Potential
Our striking observation that Ring1B-deficient embryonic neurospheres growing under proliferative conditions express differentiation markers (β-tubulin III) suggests defective maintenance of the undifferentiated state central to stem/progenitor cells. Unscheduled differentiation in the absence of Ring1B was, however, only partial, because when its paralog, Ring1A, was also inactivated, signs of differentiation affected a larger fraction of cells in the culture, some of which expressed morphological features of neurons and expressed not only β-tubulin III but also neuronal markers characteristic of differentiated neurons, such as Map2ab and neurofilament. Nevertheless, Ring1B-deficient neurospheres derepress a number of transcription factors relevant to the differentiation of neural progenitors and the specification of neuronal fates, in particular, Neurod1, which appear as Ring1B direct targets. Acute inactivation of Bmi1, in contrast, does not lead to similar differentiative defects, although downregulation of progenitor markers such as Nestin and Neurog2 were observed . It is worth noting that the fewer GFAP+ cells and the lack of O4+ cells together with the presence of large process-bearing Tuj1+ cells in Ring1A- and Ring1B-deficient cells growing without mitogens suggest that neuronal differentiation is much favored in these mutant progenitors. Most likely this relates to the observation that the neural cell fate appears to be the default specification during neural differentiation [52, 53]. The accelerated neuronal differentiation seen here in Ring1A- and Ring1B-deficient neural progenitors parallels the differentiative defects seen in Ring1B-deficient ES cells [44, 54], except for the fact that the failure to prevent differentiation leads toward a defined, rather than a more disordered, cell fate, possibly due to restrictions on the developmental potential of tissue-specific stem cells.
A possible explanation for the premature differentiation of Ring1B-deficient neural progenitors is that some of the signaling pathways involved in stem cell self-renewal are affected. Of these, the Notch pathway appears to be a good candidate, considering the accelerated differentiation of NSCs defective in one or more components of the Notch signaling pathway [46, 55]. Active Notch signaling downregulates proneural gene expression, thus inhibiting neuronal differentiation (reviewed in [56–59]). Indeed, expression analysis of Ring1B-deficient neurosphere cultures, under proliferating conditions, shows that stationary mRNA levels of Notch effectors, such as Hes5, are downregulated. Considering that Ring1B is a transcriptional repressor, it is likely that downregulation of Notch effectors occurs through indirect effects, perhaps through the upregulation of uncharacterized transcriptional repressors or microRNAs. In this regard, it is worth noting that regulation of Notch target genes is preferentially modulated by postranslational modifications in subunits of transcriptional complexes, leading to the dismissal/recruiting of coactivators and corepressors . Although Bmi1 and Ring1B associate with each other for efficient H2A monoubiquitylation [18, 22, 27, 61], the fact that Bmi1 inactivation does not have a similar impact on Notch signaling may be due to functional substitution by Bmi1 paralogs known to act as cofactors in H2A modification [61, 62]. Alternatively, it is possible that Ring1B complexes other than PRC1 are involved, at least in part, in transcriptional control of the regulatory network associated with the program defined by Notch signaling.
Coordinated regulation of cell cycle exit and differentiation of progenitors is essential for the generation of appropriate numbers of cells. Thus, in epithelial cells, Notch1 has been found to activate p21 expression . In contrast, in our Ring1BΔ/Δ neural stem/progenitor cells, Notch signaling downregulation occurs with p21 upregulation. However, rather than an inconsistency, both observations may only be a reflection of the known context dependency of Notch signaling activity, by which it can serve both as an oncogene and also as a tumor suppressor . In neural stem/progenitor cells, it is not known whether slowing proliferation, perhaps due to upregulation of inhibitors of cell proliferation (Cdkn1a/p21), and unscheduled differentiation are related processes or whether they are independent of each other. It has been noted that cell cycle lengthening of neural progenitors may contribute to a switch from proliferative to neuron-generating cell division . If this is the case, it is conceivable that downregulation of the Notch pathway may occur as a consequence of a primary alteration in the cell cycle. This is illustrated by Hes1 repression mediated by p107 , an inhibitor of G1–S transition. Alternatively, premature neurogenesis, induced by weakening of Notch signaling, may lead, even in the presence mitogens, to slower, neurogenic-like cell divisions previous to eventual cell cycle withdrawal.
Our results show that, in the absence of mitogens, loss of function of Ring1B has little effect on cell differentiation under nonclonal conditions (as a fraction of neural and glial derivatives and the total number of differentiated cells). An exception is the significantly lower number of oligodendrocytes formed in Ring1BΔ/Δ than in control neurosphere cultures, a phenotype strongly enhanced in Ring1A−/−Ring1BΔ/Δ mutant cultures. This observation could not be confirmed in vivo, because olidogendrocyte formation peaks postnatally and the administration of tamoxifen to gestating mice leads to defective delivery of pups (also, tamoxifen treatment at very early stages leads to a severe developmental defect in our Ring1B mutant mouse line). However, a similar phenotype was recently reported for neural progenitors depleted of Ezh2 (the Polycomb protein with activity of histone H3K27 methyltransferase), which were found to be unable to differentiate into oligodendrocytes . In vivo, we found alterations in astrocyte and neuron differentiation in the OB. First, fewer GFAP+ cells were detected in 17.5 dpc mutant fetuses that were exposed to tamoxifen at day 13.5 of development, compared with controls, suggesting defective astrocytic differentiation. In contrast, neurogenesis at the same developmental stage, assessed as Tuj1+ cells, appeared enhanced. Although Ring1BΔ/Δ OB progenitor proliferation is lower than in controls, they contain more double Tuj1+BrdU+ cells, suggesting that the tendency toward neuronal differentiation of mutant progenitors may account for the lower number of GFAP+ cells in mutant OBs. Finally, alterations in the differentiating ability of Ring1BΔ/Δ progenitors were also observed in the lower developmental potential of clonally derived mutant neurospheres.
In summary, we have shown that the Polycomb protein Ring1B promotes proliferation and self-renewal of embryonic neural stem/progenitor cells through the repression of cell cycle inhibitors and maintenance of signaling pathways (Notch) that prevent differentiation. The data also suggest that Ring1B is part of the mechanisms that regulate competence of embryonic neural stem/progenitor cells to generate neurons and glia over developmental time.
We thank Y. Mizutani and R. Varillas for excellent technical help. We thank V. Campuzano and M. Barbacid for the RERTert mouse line and M. Endoh for the Eomes primers and help with the ChIP experiments. T.E. acknowledges Dr. Toyoda's support. M.R-T. and H. M-G. were recipients of FPU and FPI fellowships, respectively, from the Ministerio de Educacion y Ciencia and Comunidad de Madrid, respectively. This work was supported by grants SAF2007-65957-C02-01 (M.V.), the OncoCycle program from the Comunidad de Madrid (M.V.), SAF2004-05798, and CIBERNED CB06/05/0065 from Instituto de Salud Carlos III (C.V-A.)
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.