Conditional Activation of Bmi1 Expression Regulates Self-renewal, Apoptosis, and Differentiation of Neural Stem/Progenitor Cells In Vitro and In Vivo§


  • Gokhan Yadirgi,

    1. Blizard Institute of Cell and Molecular Science, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Search for more papers by this author
  • Veronica Leinster,

    1. Blizard Institute of Cell and Molecular Science, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Search for more papers by this author
  • Serena Acquati,

    1. Blizard Institute of Cell and Molecular Science, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Search for more papers by this author
  • Heeta Bhagat,

    1. Blizard Institute of Cell and Molecular Science, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Search for more papers by this author
  • Olga Shakhova,

    1. Blizard Institute of Cell and Molecular Science, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Search for more papers by this author
  • Silvia Marino

    Corresponding author
    1. Blizard Institute of Cell and Molecular Science, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    • Blizard Institute of Cell and Molecular Science, 4 Newark Street, London E1 2AT, United Kingdom
    Search for more papers by this author
    • Ph: 44-207-882-2585; Fax: 44-207-882-2180

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • Author contributions: G.Y., V.L., and S.A.: collection and/or assembly of data, data analysis and interpretation, manuscript writing; H.B.: collection and/or assembly of data; O.S.: conception and design, collection and/or assembly of data, data analysis and interpretation; S.M.: conception and design, data analysis and interpretation, financial support, manuscript writing. G.Y. and V.L. contributed equally to this work.

  • §

    First published online in STEM CELLSEXPRESS February 8, 2011.


The Polycomb group protein Bmi1 is a key regulator of self-renewal of embryonic and adult central nervous system stem cells, and its overexpression has been shown to occur in several types of brain tumors. In a Cre/LoxP-based conditional transgenic mouse model, we show that fine-tuning of Bmi1 expression in embryonic neural stem cell (NSC) is sufficient to increase their proliferation and self-renewal potential both in vitro and in vivo. This is linked to downregulation of both the ink4a/ARF and the p21/Foxg1 axes. However, increased and ectopic proliferation induced by overexpression of Bmi1 in progenitors committed toward a neuronal lineage during embryonic cortical development, triggers apoptosis through a survivin-mediated mechanism and leads to reduced brain size. Postnatally, however, increased self-renewal capacity of neural stem/progenitor cells (NSPC) is independent of Foxg1 and resistance to apoptosis is observed in neural progenitors derived from NSC-overexpressing Bmi1. Neoplastic transformation is absent in mice-overexpressing Bmi1 aged up to 20 months. These studies provide strong evidence that fine tuning of Bmi1 expression is a viable tool to increase self-renewal capacity of NSCs both in vitro and in vivo without eliciting neoplastic transformation of these cells. STEM Cells 2011;29:700–712


Stem cells are self-renewing multipotent cells that give rise to a differentiated progeny. The balanced coordination of these two stem cell fates, that is, self-renewal and differentiation, is essential for embryonic development and tissue homeostasis in the adult.

During embryonic development of the mammalian central nervous system (CNS), proliferative radial glial cells that divide at the ventricular surface of the developing neocortex are considered to be neural stem cells (NSC). This is because they self-renew and give rise to cortical neurons, astrocytes, and oligodendrocytes. In the postnatal mammalian telencephalon, self-renewing and multipotent NSC exist only in specific brain areas, namely the subventricular zone (SVZ) and the dentate gyrus of the hippocampus, where they persist throughout life [1]. The relationship between embryonic and postnatal NSC is currently ill-defined, although the latter are believed to originate from embryonic NSC [1].

Bmi1 is a member of the polycomb group (PcG) gene family of chromatin modifiers and transcriptional repressors, the essential role of which in development and homeostasis of the CNS was first highlighted by studies on knockout mice. Bmi1−/− mice displayed impaired self-renewal and maintenance of SVZ NSC leading to their postnatal depletion [2, 3] as well as impaired proliferation of postnatal granule cell progenitors leading to depletion of granule neurons and consequently to ataxia [4, 5]. The molecular mechanisms of Bmi1 function in postnatal NSC and granule cell progenitors are mediated in part by transcriptional repression of the ink4a locus [6]. This locus encodes two cell cycle inhibitors, p16Ink4a and p19ARF [7], the activities of which increase with postnatal age and have been linked to cellular senescence [2].

However, concomitant deletion/suppression of p16Ink4a or p19ARF or both does not completely rescue the above-mentioned defects in self-renewal of NSC and proliferation of granule cell progenitors derived from Bmi1−/− mice [8, 9]. Another cell cycle inhibitor, namely p21WAF1/Cip1, contributes to mediating Bmi1 function in both developmental contexts [10, 11]. Other NSC functions, such as migration, have also been shown to be controlled by Bmi1 in an ink4a-independent manner [12].

Elevated Bmi1 expression induces tumorigenesis in the hematopoietic system in transgenic mice [13]. Increased expression of BMI1 has been demonstrated in humans, not only in hematological malignancies, such as high-grade lymphomas [14] but also in neural tumors, for example, medulloblastomas and neuroblastomas [4, 15]. A tight regulation of Bmi1 expression might therefore not only be crucial for maintaining a balance between self-renewal and differentiation or senescence of NSC, but it may also play a role in CNS tumorigenesis, when deregulated.

Overexpression of Bmi1 in nestin+ NSC and neural progenitor cells (NPC) during CNS development has been shown to have little effect in vivo albeit increasing self-renewal, proliferation, and neuronal differentiation of these cells in vitro. Neoplastic transformation of these cells leading to brain tumors was not seen in these mice [16]. However, virus-mediated induction of Bmi1 overexpression in NSC/NPC during embryonic cortical development and in the postnatal brain increased their survival and self-renewal in vivo [17]. Moreover, brain tumors were observed in a percentage of these mice. Although He et al. showed that Bmi1 mediated increased self-renewal in vitro through downregulation of the ink4a locus, Fasano et al. showed a new association between Bmi1 and Foxg1 in mediating increased self-renewal in vivo in their experimental set up.

In the light of these partially contradictory findings, it is currently unclear (a) whether or how Bmi1 overexpression increases the self-renewal potential of NSC in vivo, (b) whether Bmi1 overexpression influences survival and differentiation of non-stem neural progenitors in vivo, (c) what is the relative contribution of the ink4a/ARF and Foxg1/p21 axes to the observed phenotypes, and (d) whether Bmi1 overexpression in NSC and cells derived thereof leads to tumor formation in the mouse.

Here, we describe a Cre/LoxP-based conditional transgenic approach to induce gain of function of Bmi1 in embryonic and postnatal NSC and in all their progeny in vivo.


See Supporting Information.


Generation of a Mouse Line for Conditional Gain of Function Expression of Bmi1 in the CNS

The pccall2 construct [18] was chosen to drive conditional expression of Bmi1. It is composed of a cytomegalovirus (CMV) enhancer/chicken β-actin promoter followed by a βgeo (LacZ/neomycin) fusion gene and three copies of the SV40 polyA signal flanked by LoxP sites. The full-length murine Bmi1 cDNA was inserted after the second LoxP site, in front of an IRES-eGFP (enhanced green fluorescent protein) sequence (Supporting Information Fig.S1A). The CMV enhancer/chicken β-actin promoter drives β-galactosidase expression and confers neomycin resistance, however, on Cre-mediated excision of βgeo, Bmi1 and eGFP will be simultaneously expressed [19]. The pccall2-Bmi1 vector was introduced into TC-1 mouse ESCs. Four ESC clones showing high LacZ expression (Supporting Information Fig. S1B) and single copy integration (Supporting Information Fig.S1C) were selected and electroporated with a Cre recombinase expression vector to test the induction of Bmi1 overexpression. Clones IB5 and IE1 showed negative LacZ expression and Bmi1 protein level comparable with a medulloblastoma sample (Supporting Information Fig. S1D). They were therefore selected for injection into blastocysts to generate chimeric mice. Germline transmission and line establishment was achieved for clone IB5, from now on referred to as STOPFloxBmi1.

Conditional Activation of Bmi1 Expression in the Developing Neocortex Increases Proliferation of NSC in the Ventricular Zone and Induces Apoptosis of Ectopically Proliferating NPCs

Initially, we analyzed the expression pattern of the newly generated transgenic mouse line in vivo prior to conditional activation of Bmi1 expression. X-gal staining revealed robust LacZ expression in all layers of the developing neocortex at E12.5 (Fig. 1B–1D), including NSC located in the apical region of the ventricular zone (VZ; Fig. 1D, arrow). We, therefore, conclude that the transgene is active in embryonic NSCs and in their progeny during embryonic forebrain development in vivo.

Figure 1.

Analysis of forebrain development in Nestin-Cre;STOPFloxBmi1 at E12.5. (A): Schematic representation of the conditional Bmi1 transgenic construct prior and upon Cre-mediated recombination. Expression of the transgene in the developing neocortex as assessed by X-gal staining of sagittal section, overview (B) and higher magnifications (C, D) of the area boxed in (B). X-gal staining is retained only in the vessels on Nestin-Cre recombination (E, F). Widespread expression of GFP in double transgenic mice ([I, G], detail of GFP expression in ventricular zone (VZ) NSC is shown in the inset) and increased number of NSC expressing higher levels of Bmi1 most prominently in the apical VZ (J, H). Increased and ectopic proliferation of neuronal progenitor cell (NPC) in Nestin-Cre;STOPFloxBmi1 as assessed by pH3 staining (K, N, Q). Increased TUNEL+ apoptotic NPC (L, O) and cCSP3+ cells (M, P) but no increase in the number of apoptotic cells in VZ NSC (R). More prominent RC2 staining in the VZ of double-mutant mice (T, Y). Reduced number of Tbr2+ intermediate progenitors in double-mutant mice (U, Z, V, A′). Quantification of the findings is shown in (S). *, p < .05; **, p < .01; ***, p < .001 with Student's t test, Error bars represents SD, n ≥ 3. Expression of neuronal differentiation markers, DCX and βIII-tubulin (W, B′, X, C′), in NPC which escaped apoptotic death. Scale bar = 2 mm (B), 250 μm (C, E, H, N, L, O), 125 μm (D, F, G, I, H, J, M, P, Q, R, T–C′). Abbreviations: DCX, doublecortin; DAPI, 4′,6-diamidino-2-phenylindole; eGFP, enhanced green fluorescent protein; GFP, green fluorescent protein; IRES, internal ribosome entry site; LV, lateral ventricle; pA, polyA; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labeling.

To assess the effect of Bmi1 upregulation in NSC and NPC, we used a Nestin-Cre transgenic line, which has been shown to successfully recombine loxP-flanked sequences in specific tissues and to effectively express Cre recombinase in these cells on paternal inheritance of the transgene [20] (Fig. 1A). Loss of LacZ expression was indicative of recombination in all cells of the VZ and in a significant proportion of cells throughout all layers of the developing neocortex (Fig. 1E and 1F). To confirm activation of the construct, GFP immunolabeling was performed on adjacent sections. Widespread cytoplasmic staining was seen in double mutant mice with an expression pattern similar to that seen in the X-gal staining; importantly, expression of the construct was noted in NSC (Fig. 1I, inset). Endogenous Bmi1 expression was seen in NSC of the apical region of the VZ and the staining was less intense in progenitor cells throughout the developing neocortex (Fig. 1H). Double mutant mice showed higher expression of Bmi1 in the VZ, which was thicker and showed clustering of cells strongly expressing Bmi1 (Fig. 1J). Stronger expression of Bmi1 was noted also in progenitor cells located in the intermediate zone (IZ) of double mutant mice (Fig. 1J) as compared with control (Fig. 1H), however, these cells seem to overexpress Bmi1 less strongly than NSC. It is currently unclear why the expression patterns of β-galactosidase, GFP, and Bmi1 differ, although it is possible that the protein turnover varies among different proteins in different cell types.

Next, we assessed the impact of Bmi1 overexpression on proliferation and differentiation of NSC, and on the progenitor cells derived thereof in E12.5 Nestin-Cre;STOPFloxBmi1 embryos compared with controls. Proliferation was assessed by immunostaining with phospho-histone H3 (pH3). We found considerably more proliferating cells in the VZ (Fig. 1N and 1Q as compared with Fig. 1K), which were not only localized at the apical region, as in the control, but also clustering into small aggregates close to the basal region of the VZ and within the IZ. More prominent staining for the radial glia marker RC2 was also noted in mutant mice (Fig. 1T, 1Y). While analyzing these slides, we observed, in the propidium iodide staining, frequent condensed nuclei in the double mutant embryos, suggestive of apoptotic cells. To test this hypothesis, we performed TUNEL staining, which showed massive apoptosis in the area where ectopic proliferation was noted, mainly in the IZ and occasionally in the outermost layers of the differentiating neocortex (Fig. 1O and 1R compared with 1L). Interestingly, no increase in apoptosis was noted in the VZ, implying that NSC are not shunted into apoptosis upon Bmi1-mediated induction of proliferation (Fig. 1R). Similar findings were seen in immunostaining for cleaved Caspase3 (cCSP3), a more specific marker of apoptosis (Fig. 1M, 1P). Immunostaining for Tbr2, a marker of differentiated preplate neurons at E12.5, showed a reduction in the overall number of progenitor cells reaching this stage of differentiation as a consequence of the cell death occurred (Fig. 1U, 1Z, 1V, 1A′). However, staining for doublecortin (DCX) and βIII-tubulin revealed that the surviving cells achieved appropriate differentiation (Fig. 1W, 1B′, 1X, 1C′). Quantification of positive cells for markers pH3, cCSP3, and Tbr2 is shown (Fig. 1S).

H&E staining of sagittal sections of the developing cortex at E16.5 suggested overall reduction of the size of the forebrain (Fig. 2A, 2C); however, not all layers appeared equally affected—that is, the IZ, the subcortical plate, and the cortical plate showed the most pronounced reduction in thickness (Fig. 2B, 2D, 2E [quantification]). Remarkably, the overall thickness of the VZ/SVZ layers was not affected (Fig. 2B, 2D, 2E [quantification]). X-gal staining on adjacent sections confirmed widespread recombination in all cortical layers (Fig. 2A, 2C, inset). Immunostaining for Sox2, a marker of VZ progenitors, allowed us to discriminate between SVZ and VZ and although measurements of both layers did not reveal significant difference in their thickness, a trend toward a thinner SVZ and a thicker VZ was observed (Fig. 2F, 2G, 2H [quantification]). Immunostaining for pH3 and quantification of the positive cells confirmed that also at E16.5 there was an increased number of proliferating apical VZ progenitors (Fig. 2L, 2M, 2N [quantification]), whereas the number of proliferating progenitors in the SVZ and the IZ were not affected (data not shown). Increased proliferation of cells located in VZ/SVZ was confirmed also with an independent marker, BrdU (Fig. 2J, 2K, 2I [quantification]). The number of Tbr2+ intermediate neuronal progenitors was unaffected at this later developmental time point and neuronal differentiation was retained as assessed by expression of βIII-tubulin and DCX (data not shown).

Figure 2.

Analysis of forebrain development in Nestin-Cre;STOPFloxBmi1 at E16.5. Reduced brain size in double mutant, H&E staining (A, C), and complete loss of X-gal staining confirming full recombination (inset). Higher magnification showing reduced cortical thickness (B, D) and layer measurements shows that CP, SP, and IZ are most affected (E). Sox2 immunostaining reveals no significant difference in the thickness of the VZ (F, G, H). Increased proliferation, as assessed by 2-hour BrdU pulse and detection, is observed in VZ/SVZ ([J, K, I], details of the labeling in VZ/SVZ is shown in the inset) and confirmed with an independent proliferation marker, pH3, in the apical VZ (L, M, N). *, p < .05; ***, p < .001 with Student's t test, Error bars represents SD, n ≥ 3. Scale bar = 2 mm (A, C); 1 mm (J, K); 250 μm in inset (A, C), (B, D, F, G); 125 μm in inset (J, K), (L, M). Abbreviations: BrdU, bromodeoxyuridine; CP, cortical plate; IZ, intermediate zone; SP, subcortical plate; VZ, ventricular zone; SVZ, subventricular zone.

In conclusion, activation of Bmi1 expression during embryonic forebrain development leads to increased proliferation of VZ stem cells. However, increased and ectopic proliferation of migrating and differentiating neuronal progenitors shunts them into apoptotic death at E12.5, leading to a reduced number of these progenitors and eventually to a reduced size of the developing neocortex in the mutant mice.

Embryonic Basic Fibroblast Growth Factors NSPC-Overexpressing Bmi1 Show Increased Self-Renewal Capacity and Increased Proliferation In Vitro

To further investigate functional effect of increased Bmi1 expression in embryonic NSC and NPC, two in vitro culture systems were used, the neurosphere assay [21] and the adherent NSC culture system [22, 23].

The telencephalon of Nestin-Cre;STOPFloxBmi1 and control mice were isolated at E16.5, dissociated into a single-cell suspension and plated into serum-free medium containing epidermal growth factor and basic fibroblast growth factors. In these conditions, mitogen-responsive cells [NSC and other progenitor cells, here referred to as neural stem/progenitor cells (NSPC)] proliferate and form three-dimensional structures, named neurospheres (NS). Repeated cycles of harvesting and dissociation of NS followed by replating of the single-cell suspension under the same culturing conditions were performed to enrich for NS-forming NSPC.

X-gal staining performed on controls (STOPFloxBmi1) and double transgenic NS confirmed expression of the transgene in all cells of the NS and showed a loss of expression in the double mutant (Fig. 3A–3D). Immunohistochemical staining for Bmi1 and GFP confirmed that recombination and activation of the construct had occurred in the double mutant NS (Fig. 3B–3E, 3C–3F). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) and semiquantitative assessment of protein expression by Western blot analysis was performed on NS of Nestin-Cre;STOPFloxBmi1 and controls and they confirmed overexpression of Bmi1 as observed in situ (Fig. 3G, 3H).

Figure 3.

Increased self-renewal and proliferation of neural stem/progenitor cells (NSPC)-overexpressing Bmi1 in vitro. Loss of β-galactosidase activity (A, D) and gain of GFP and Bmi1 expression in neurospheres (NS) originating from NSPC isolated from Nestin-Cre;STOPFloxBmi1(B, E, C, F). Increased Bmi1 expression in Nestin-Cre;STOPFloxBmi1 NS is confirmed by quantitative reverse transcription polymerase chain reaction (G) and Western blot analysis (H). Increased self-renewal, as assessed by counting the number of NS originating from the same number of plated cells over several passages (I). Increased alamar blue reduction over 72 hours (J) indicates increased proliferation of NSPC-overexpressing Bmi1. NSC cultured on adherent conditions (K) are homogeneously expressing Nestin (L), Musashi1 (M), and Sox2 (N). Growth curves show increased proliferation in Bmi1-overexpressing cultures (O). Overexpression of Bmi1 did not significantly increase apoptotic death in NSC cultured in adherent conditions (Q). However, it significantly increased the number of differentiating progenitors undergoing apoptosis in NS originating from these NSPC (P). *, p < .05; **, p < .01; ***, p < .001 with Student's t test; Error bars represents SD, n ≥ 3. Scale bar = 125 μm. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; DIV, days in vitro; GFP, green fluorescent protein.

To determine whether the in vitro self-renewal capacity of NSPC isolated from Nestin-Cre;STOPFloxBmi1 at E16.5 would be increased in vitro, we dissociated and replated tertiary NS over nine passages as a measure of their self-renewal capacity. A significant increase in NS frequency was consistently noted in the cultures overexpressing Bmi1 (Fig. 3I, p < .001). Similar results were obtained when NS were dissociated, replated in serial dilutions, and the total number of NS arising after 6–7 days counted (p < .001), therefore, implying that the effect was not dependent on the original cell density of the culture (Supporting Information Fig.S2A, blue and purple bars).

Next, we analyzed whether acute upregulation of Bmi1 expression would elicit a similar effect. NS isolated from STOPFloxBmi1 E16.5 embryos were dissociated and single-cell cultures infected with Adeno-Cre [8 multiplicity of infection (MOI)] or Adeno-GFP (8 MOI) viruses. A total of 90%–100% of the NS originating thereof showed lack of LacZ expression upon Adeno-Cre infection, whereas LacZ expression was retained in NS infected with Adeno-GFP (data not shown). Here, GFP positivity was seen in almost 100% of the NS and confirmed the high infection rate and therefore the efficient gene delivery of the adenoviral system used (data not shown). To control nonspecific effects of infection and overexpression of any exogenous proteins, NS were isolated from control (nontransgenic) littermates and infected in parallel with Adeno-Cre and Adeno-GFP. All assays were performed on these cultures under the same experimental conditions as described before and no significant difference was seen between the two viruses (data not shown). Infection with Adeno-GFP will be used throughout the manuscript as control for all experiments where infection with Adeno-Cre was performed. We showed that self-renewal of NSPC was increased upon acute induction of Bmi1 overexpression in a similar fashion as upon Nestin-Cre mediated induction during embryonic development (Supporting Information Fig.S2A, green and red bars).

In accordance with our in vivo observation, increased proliferation of NSPC on activation of Bmi1 expression was confirmed in culture by means of EdU labeling (Supporting Information Fig. S2B; p < .01) and a colorimetric assay, alamar blue, measuring the metabolic activity of the cells (Fig. 3J; p < .05 and p < .01).

To address the question whether the observed increase in proliferation on Bmi1 overexpression was due to an effect on NSC rather than on more committed precursors, we used an adherent culture system, in which, acutely dissociated NS are seeded at clonal concentration (10,000 cells per centimeter square) on Matrigel-coated flasks in expansion medium. Under these conditions, cultures are composed of a homogeneous population of NSC [23], which is characterized by uniform expression of NSC markers such as Nestin, Sox2, Musashi (Fig. 3K–3N), and lack of differentiated cells as assessed by immunostaining for glial fibrillary acidic protein (GFAP), O4, and βIII-tubulin (data not shown). NSC can be induced to efficient trilineage differentiation upon switching to a differentiation inducing medium [22, 23]. Growth curves, obtained by plotting cell counts over five passages, revealed a significant increased proliferation in double transgenic NSC as compared with controls (Fig. 3O; p < .001 and p < .01).

In summary, in vitro assays revealed that chronic and acute enhancement of Bmi1 expression increases (a) self-renewal capacity of NSPC and (b) proliferation of NSC in a similar fashion.

A Cell Autonomous Mechanism Causes Apoptotic Death of Differentiating NPC-Overexpressing Bmi1

In vivo analysis of embryonic cortical development of Nestin-Cre;STOPFloxBmi1 mutants had demonstrated that NSC located in the VZ were not shunted into apoptotic death despite Bmi1 induced increased proliferation (Fig. 1R). Annexin V labeling followed by flow cytometric analysis was performed on NSC isolated from Nestin-Cre;STOPFloxBmi1 and cultured in adherent conditions. In accordance with the in vivo observation, no significant differences were noted between cultures overexpressing Bmi1 and control (Fig. 3Q).

We observed a striking increase of apoptotic cells in the IZ of Nestin-Cre;STOPFloxBmi1 mutants (Fig. 1O, 1R, 1P). The IZ contains mainly progenitors committed to a neuronal fate that originated from NSC upon asymmetric cell division and are migrating away from the ventricle. To determine whether cell death was due to a cell autonomous or to a noncell autonomous mechanism, we assessed the apoptotic rate of NPC originating from NSPC in a nonphysiological environment such as a cultured NS. Annexin V staining revealed a significantly higher percentage of apoptotic cells in Nestin-Cre;STOPFloxBmi1 NS (Fig. 3P; p < .05).

These finding suggest that increased Bmi1 expression is tolerated in VZ NSC leading to an increase of the stem cell pool, both in vitro and in vivo, but its downregulation is essential for survival of committed neuronal progenitors migrating away for the ventricular surface. Upregulation of Bmi1 expression in these progenitors shunts them into apoptosis through a cell autonomous mechanism.

Increased Pool of Undifferentiated NSPC and Delayed Neuronal Differentiation in NS-Overexpressing Bmi1

CNS-derived NS are morphologically and functionally heterogeneous, in fact, they contain not only self-renewing NSPC (∼1%–10% of the total number of cells) but also committed glial and neuronal progenitors as well as differentiated cells.

To confirm that the increased number of apoptotic cells was mainly affecting neuronal committed progenitors and to analyze the impact of Bmi1 overexpression on glial differentiation, we looked at the cellular composition of NS-overexpressing Bmi1. Immunolabeling of OCT-embedded NS for Mash1, an early marker of neuronal differentiation, revealed a significant decrease in the number of positive cells in Bmi1-overexpressing NS (Fig. 4A, 4B, 4G). Concomitant staining for cCSP3 confirmed increased apoptosis occurring upon Bmi1 overexpression, although no increase in the number of double-positive cells was noted, in accordance with progenitors being shunted into apoptotic death at an earlier stage, possibly at the time of commitment (Fig. 4C–4F, 4G). Expression analysis of markers of glial differentiation, NG2, A2B5, GFAP, in NS either by flow cytometry (Fig. 4J, 4M [quantification]) or by immunohistochemical analysis (Fig. 4H, 4I, 4K, 4L), revealed a significantly higher percentage of cells coexpressing GFAP and NG2 (79%), GFAP and A2B5 (73%), NG2 and A2B5 (54%) in Bmi1-overexpressing NS as compared with those of control NS, respectively 41%, 13%, and 17% (p < .001 and p < .01).

Figure 4.

Analysis of differentiation of neural stem/progenitor cells and of the cellular composition of NS-overexpressing Bmi1. Depletion of Mash1+ cells in NS-overexpressing Bmi1 (A, B). Increased number of apoptotic cells as assessed by cleaved caspase3 staining (C, D). Quantification of the findings is shown in (G). Increased number of NG2;A2B5 (histological [H, I] and flow cytometric analysis [J]), GFAP;A2B5 (K, L) in NS upon activation of Bmi1 expression. Quantification of the findings is shown in (M). Upon induction of differentiation, increased number of cells expressing Musashi (N, O) and coexpressing A2B5;NG2 (P, Q) A2B5;GFAP (R, S) and reduction of the number of βIII-tubulin-positive cells (T, U) is seen at DIV5 in Bmi1-overexpressing cells. *, p < .05; **, p < .01; and ***, p < .001 with Student's t test, Error bars represents SD, n ≥ 4. Scale bar = 125 μm (A–F, H, I, K, L), 250 μm (N–U). Abbreviations: cCASP3, cleaved caspase3; DAPI, 4′,6-diamidino-2-phenylindole; DIV, days in vitro; GFAP, glial fibrillary acidic protein; NS, neurospheres.

Upon plating onto laminin-coated surface and withdrawal of growth factors, NS attach to the substrate and cells start migrating radially from the NS body toward the periphery, while they differentiate into neurons, astrocytes, and oligodendrocytes. As NSPC differentiate, nestin expression is downregulated and expression of distinct neuronal and glial markers such as βIII-tubulin, GFAP, and O4 is activated [24]. Differentiation of NS is therefore a useful assay to study differentiation properties of NSPC in a simplified in vitro context.

We analyzed the impact of increased Bmi1 expression on the differentiation capacity of NSPC contained in NS derived from Nestin-Cre;STOPFloxBmi1 compared with controls after 5 days in vitro (DIV5). The percentage of cells expressing nestin was higher (36% vs. 13%, p < .01) in Bmi1-overexpressing cultures (data not shown and Fig. 4V). Moreover, the expression of Musashi, another marker of undifferentiated/uncommitted NSPC [25], was retained in a significantly higher number of cells (50% vs. 28%, p < .01) in the mutant cultures (Fig. 4N, 4O, 4V). Although the percentage of GFAP+ astrocytes and O4+ oligodendrocytes was similar in both cultures (data not shown), a significant reduction in the number of βIII-tubulin+ neurons was noted in Nestin-Cre;STOPFloxBmi1 cultures (Fig. 4T, 4U, 4V; p < .01). The percentage of cells expressing markers of immature glial and oligodendroglial progenitors such as A2B5 (28% vs. 10%, p < .01) and NG2 (19% vs. 10.9%, p < .01) as well as coexpressing GFAP and A2B5 (20% vs. 3.5%, p < .01) and GFAP and NG2 (17% vs. 10.5%, p < .001) was significantly higher in cultures overexpressing Bmi1 (data not shown and Fig. 4P, 4Q, 4R, 4S, 4V). Similar analysis performed at DIV9 showed higher number of O4+ oligodendrocytes (10% vs. 6.5%, p < .01), as expected, but also a higher number of βIII-tubulin+ neurons (8.5% and 5.4% respectively, p < .01; Supporting Information Fig. S3A, S3E and S3B, S3F, S3K).

These results suggest that increased expression of Bmi1 maintains NSPC in an undifferentiated state, both in NS and upon induction of differentiation and that this does not hamper initiation of glial/oligodendroglial differentiation. Downregulation of Bmi1 expression, however, is crucial for initiation of neuronal differentiation, an effect which can be overcome at later stages of differentiation.

Overexpression of Bmi1 in Postnatal SVZ NSC Increases Self-Renewal Capacity In Vitro and Increases the Number of NSC In Vivo

Next, we analyzed the effect of Bmi1 overexpression on postnatal NSPC isolated from the SVZ. We studied the expression pattern of STOPFloxBmi1 transgenic line at two postnatal timepoints, P7 and P70, by X-gal staining. At P7, LacZ expression was seen throughout the cerebral cortex, in the hippocampus, the basal ganglia, and notably, in all cells of the SVZ (Supporting Information Fig.S4A). A similar expression pattern was seen at P70 (Supporting Information Fig.S4B). Double labeling with X-gal and Nestin showed expression of the transgene in the vast majority of cells populating the SVZ at P7 (Fig. 5A, 5B), including scattered cells located within and underneath the ependymal lining. Double labeling of these cells for X-gal and GFAP, a marker of NSC in the postnatal and adult SVZ (Fig. 5C, 5D), confirmed that the transgene was indeed expressed in SVZ NSC. Similar analysis performed on adult brains (P70) showed comparable results (Fig. 5E, 5F), in fact, the lower cellularity of the SVZ allowed us to conclude that SVZ NSC expressed LacZ (Fig. 5F, arrowhead). Moreover, NS originating from NSPC isolated from the SVZ of Bmi1STOPFlox mice at p7 and p70 were strongly and homogeneously LacZ positive (Supporting Information Fig.S4C), indicating that the construct was active in these cells.

Figure 5.

Overexpression of Bmi1 in postnatal neural stem/progenitor cells in vitro and in vivo. In the subventricular zone (SVZ) at P7, several X-gal+nestin+ (A, B) and X-gal+;GFAP+ (C, D) cells are identified. At P70 colocalization is seen between X-gal and GFAP in several subependymal NSC ([E, F], arrowhead).Increased self-renewal (G) and increased proliferation (H) of postnatal NSC/NPC-overexpressing Bmi1. Resilience to apoptosis is seen in differentiating progenitors contained within NS-overexpressing Bmi1 (I). *, p < .05; **, p < .01; ***, p < .001 with Student's t test. Increased number of label retaining BrdU+ cells at the lateral wall of the lateral ventricle of double transgenic P7 mice ([J, K], inset shows BrdU/DAPI). Quantification of the findings is shown in (L). **, p < .01 with ANOVA. Small clusters of GFAP+;GLAST+ NSC are identified in the lateral wall of the SVZ of the mutant mice (M, N). Reduced overall size (O, P), and weight (Q) of the adult brain of NestinCre;STOPFloxBmi1 at 8 months. ***, p < .001 with Student's t test. Error bars represents SD, n ≥ 3. Scale bar = 250 μm (A–D), 125 μm (E, F, J, K), 62.5 μm (M, N). Abbreviations: BrdU, 5-bromo-2-deoxyuridine; DAPI, 4′,6-diamidino-2-phenylindole; GFAP, glial fibrillary acidic protein; GFP, green fluorescent protein; GLAST, l-glutamate/d-aspartate transporter.

Following this, we induced Bmi1 overexpression in cultured P7 SVZ NSPC by adenoviral-mediated Cre delivery, as previously described. Similarly to what was observed with embryonic NSPC, 90%–100% of NS originating after infection of dissociated NSPC lacked LacZ expression (Supporting Information Fig.S4E). Infection with Adeno-GFP, performed in parallel, confirmed an infection rate of 100% (Supporting Information Fig.S4D) and retained LacZ expression on adenoviral infection (data not shown).

As predicted, overexpression of Bmi1 in postnatal NSPC led to increased self-renewal capacity of these cells, indicated by an increased frequency of NS arising from the same number of plated cells over nine passages (Fig. 5G) and in a clonogenic assay (Supporting Information Fig. S4F, p < .001 and p < .01). As for embryonic NSPC, the proliferation rate of the SVZ NSPC upon dissociation was significantly higher in cultures overexpressing Bmi1 as compared with control GFP-infected cultures (alamar blue assay, Fig. 5H, p < .01 or p < .05). However, when the apoptotic rate was measured in NPC contained within NS, a significant reduction in the percentage of apoptotic cells was seen in NS-overexpressing Bmi1 (Fig. 5I, p < .05). Similar results were obtained at p70 (data not shown).

In vivo analysis of the SVZ of Nestin-Cre;STOPFloxBmi1 mice at P7 revealed increased number of label retaining BrdU+ cells located in the subependymal area of the lateral wall of the lateral ventricle (Fig. 5J, 5K, 5L [quantification]) as well as small clusters of GFAP+ and GLAST+ cells in the same areas (Fig. 5M–5N, arrowheads) which were not seen in control animals. These finding are in keeping with the in vitro data and raise the possibility that the resident NSC pool is increased in vivo also postnatally.

We show here that increased Bmi1 expression in postnatal SVZ NSC increases their in vitro proliferation and self-renewal capacity, but does not predispose NPC to apoptotic death. Increased NSC pool in vivo is also observed postnatally.

Reduced Overall Brain Size in Nestin-Cre; STOPFloxBmi1 Mice But No Tumors in Both Nestin-Cre;STOPFloxBmi1 and in A-Cre;STOPFloxBmi1 Mice

Adult Nestin-Cre;STOPFloxBmi1 animals consistently exhibited a reduced overall brain size (Fig. 5O, 5P) and brain weight (Fig. 5Q, p < .001), compared with age-matched and sex-matched controls. Histological and immunohistochemical analysis of their brains by H&E and immunolabeling for neuronal (Map2, NeuN, neurofilament, Synaptophysin), glial (GFAP, myelin basic protein), and microglial (Iba1) markers revealed that overall architecture and organization of the brain was normal despite its reduced size (Fig. 5O, 5P and data not shown). Gross examination and histological analysis of all brains (n = 12) excluded the occurrence of brain tumors in these mice.

We observed a dual effect of Bmi1 overexpression on apoptosis in NS, depending on whether the cultures were obtained from embryonic or postnatal cells with postnatal NS showing a lower apoptotic rate, therefore, we reasoned that activation of Bmi1 overexpression postnatally is more likely to elicit tumor formation. We induced Bmi1 overexpression in adult mice by means of intraventricular injection of Adeno-Cre in STOPFloxBmi1 mice. This is a highly efficient method to obtain recombination of LoxP alleles in cells of the SVZ, as reported in the literature [26] and as we have seen in our experiments in R26R mice (Supporting Information Fig. S5A, S5B). A cohort of 15 mice was kept under observation for 20 months, all brains were analyzed histologically at the termination of the experiment and no CNS tumors were observed.

We conclude that reduced brain size of Nestin-Cre;STOPFloxBmi1 is probably a consequence of the depletion of differentiating progenitors during embryonic cortical development. Bmi1 overexpression is not sufficient to induce NSC/NPC or their progeny to undergo neoplastic transformation in the mouse.

Downregulation of Survivin Is Responsible for the Increased Apoptosis Observed in Neuronal Progenitors-Overexpressing Bmi1 In Vitro and In Vivo

We analyzed the downstream mechanisms mediating the observed effects of Bmi1 overexpression. First, we assessed the expression of known Bmi1 target genes, p16ink4a, p19Arf and p21WAF1/Cip1 on NS isolated from Nestin-Cre;STOPFloxBmi1. We show reduced levels of all three cell cycle inhibitors in embryonic NS-overexpressing Bmi1 independent of whether Bmi1 expression was activated chronically during embryonic development in vivo (+NCre) or acutely in vitro (+Acre, Fig. 6A). A similar reduction of the expression levels of these proteins was observed in NS derived from postnatal NSPC, and this was independent of the timepoint analyzed (Fig. 6A). In accordance with recent data supporting a correlation between the expression of Bmi1 and Foxg1 [17], a forebrain specific transcription factor, we also found increased levels of Foxg1 in embryonic NS (Fig. 6B, (p < .01). This was independent of whether Bmi1 activation occurred during embryonic development in vivo or acutely in vitro. On the contrary, no increased expression of Foxg1 was observed in postnatal NS at P7 and P70 (Fig. 6B).

Figure 6.

Downstream effectors of Bmi1 overexpression in neural stem cells (NSC) and neural progenitor cells (NPC). Western blot analysis reveals decreased p16ink4a, p19Arf and p21 WAF1/Cip1 in NS-overexpressing Bmi1 (A). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) shows increased expression of Foxg1 in neurospheres (NS) originating from E16.5 neural stem/progenitor cells (NSPC)-overexpressing Bmi1 (B). However, similar expression levels of Foxg1 are seen in NS-overexpressing Bmi1 and controls at both postnatal time points analyzed (B). qRT-PCR show downregulation of survivin in NS-overexpressing Bmi1 but not in NSC cultured as adherent monolayer or in NS isolated postnatally (C). The percentage of cells showing nuclear positivity for survivin is also reduced in NS-overexpressing Bmi1 (D, E, F). Striking reduction of survivin+ cells in the developing neocortex of E12.5 Nestin-Cre;STOPFloxBmi1 embryos ([G, H]; [I], quantification). **, p < .01; ***, p < .001 with Student's t test. Error bars represents SD, n ≥ 4. Scale bar = 250 μm. Abbreviation: GFP, green fluorescent protein.

Although deregulation of cell cycle inhibitors is seen both in embryonic and postnatal NS, and it is likely to be responsible for the increased proliferation of NSPC-overexpressing Bmi1, it is intriguing that increased apoptosis upon neuronal commitment seems to be developmental stage-specific. Several genes are known to be involved in the control of apoptotic death of progenitors during embryonic CNS development, among them Survivin [27]. The expression of Survivin, a member of the inhibitor of apoptosis family, has been shown to be upregulated in progenitor cells upon neuronal commitment (EURExpress). Moreover, conditional inactivation of this gene during embryonic CNS development in Nestin-expressing progenitors induces massive apoptosis of developing neurons [28] with a phenotype very similar to what we observed in our Bmi1-overexpressing double-mutant mice. Consequently, we analyzed the expression of survivin by means of qRT-PCR in E16.5 NS and ADH cultures. We found a significant decrease in the expression of survivin in NS culture but not in ADH (Fig. 6C, p < .001). Also at the protein level, reduced number of cells expressing survivin was counted in NS upon immunolabeling (Fig. 6D, 6E, 6F). Interestingly, no significant change in the expression of survivin was noted in NS cultures isolated from P7 mice (Fig. 6C). In the developing embryo at E12.5 a discrete layer of cells expressing survivin is seen and this coincides with the layer of Tbr2+ neuronal committed progenitor cells (Fig. 6G and Eurexpress). We noticed a reduced number of cells expressing this protein in Nestin-Cre;STOPFloxBmi1 embryos (Fig. 6H), in accordance with downregulation of survivin being an effect of Bmi1 overexpression also in vivo. Quantification of the finding is shown in Figure 6I.

The data showed that although increased proliferation of NSC/NPC is likely to be controlled by Bmi1 through repression of its canonical cell cycle inhibitor targets, the proapoptotic role is developmental stage-specific, and it is mediated by a novel downstream effector, survivin.


We generated a transgenic mouse model, which allowed us to activate Bmi1 expression in a cell-specific and time-controlled fashion by means of Cre-mediated recombination of LoxP sites. With this technology, we achieved a mild enhancement of Bmi1 expression in embryonic and postnatal NSC and in differentiating NPC, in a way similar to the physiological upregulation of transcription factors. This makes our model ideally suited to achieve a good characterization of Bmi1 influence on stem cells properties.

Overexpression of Bmi1 in embryonic NSC leads to increased self-renewal and proliferation both in vivo and in vitro. This effect is independent of whether a protracted activation of Bmi1 expression was induced in vivo by means of nestin-driven Cre expression or whether acute induction of Bmi1 expression was achieved in vitro by adenoviral-mediated Cre delivery. Our data support and extend a recent study where lentiviral-mediated overexpression of Bmi1 by intraventricular injection at E14 led to similar results both in vitro and in vivo, although the in vivo observation in this study was limited to 3 days after injection [17]. In contrast, overexpression of Bmi1 driven by the Nestin promoter did not elicit a similar effect in vivo [16]. Although it is currently unclear why Bmi1 overexpression leads to only partially overlapping phenotypes in different experimental model systems, it can be speculated that either the developmental time point at which overexpression is achieved or the specific subpopulation of cells targeted or gene dosage, may all contribute to explain these differences.

Our experimental set up allowed us to assess the effect of Bmi1 overexpression in differentiating NPC over a protracted period of time. We show here for the first time that embryonic neural progenitors differentiating toward a neuronal lineage are highly sensitive to enhanced and/or ectopic expression of Bmi1. In fact, failure to downregulate Bmi1 expression shunted these progenitors into apoptotic death. As a consequence, reduced number of Tbr2+ intermediate neuronal progenitors were observed in the developing neocortex of the mutant mice leading to an overall reduced brain size in adult life. The apoptotic rate of NPC contained in NS derived from NSPC-overexpressing Bmi1 was similarly increased and led to a significant reduction in the number of neuronal committed Mash1+ progenitors. On the contrary, glial differentiation was favored in Bmi1-overexpressing NS. Upon induction of differentiation in vitro, the number of βIII-tubulin+ neurons originating from embryonic NSPC-overexpressing Bmi1 was reduced. Apoptotic death of NPC-overexpressing Bmi1 upon neuronal commitment is associated with downregulation of the inhibitor of apoptosis protein survivin, the expression of which is essential for the survival of developing neurons during embryonic development in vivo [28]. Downregulation of survivin expression was not observed in NS originating from postnatal NSPC and the number of βIII-tubulin+ neurons originating from these cells was not affected. NSC located in the VZ did not undergo apoptosis despite higher levels of Bmi1 expression, a phenomenon confirmed by the absence of apoptosis induction in NSC cultured under adherent conditions and the lack of downregulation of survivin in these cells both in vitro and in vivo.

Previous findings [16] showed that upregulation of Bmi1 expression led to reduced expression of p16Ink4a and p19Arf in a developmental stage-independent fashion. We also observe this effect in both embryonic and postnatal NS. Importantly, downregulation of the expression of these genes was observed when Bmi1 activation was achieved during embryonic development by means of nestin-driven Cre expression and acutely in vitro by adenoviral-mediated Cre delivery. Downregulation of p21WAF1/Cip1 expression was also observed at all timepoints analyzed and both upon chronic and acute overexpression of Bmi1. Although we could confirm upregulation of Foxg1 in embryonic NS upon activation of Bmi1, as recently reported by Fasano et al. [17], we could not detect a similar upregulation at both postnatal timepoints analyzed. We cannot exclude that this might reflect a dosage effect. Our single copy gene approach allows us to obtain a tightly controlled 1.8-fold increase of Bmi1 expression compared with a potentially greater upregulation obtained using a lentiviral transduction approach as Fasano et al. did [17]. However, it is also conceivable that downregulation of p21WAF1/Cip1, either directly through Bmi1 or indirectly in the context of p19Arf downregulation, contributes to self-renewal/proliferation control of postnatal NSPC independently of Foxg1 upregulation.

Cells with stem cell properties, so-called brain tumor stem cells, are thought to play a crucial role in initiation and maintenance of brain tumors (reviewed in [29]. These cells express high levels of Bmi1 [30], are particularly resistant to apoptosis [31, 32] and have recently been shown to be A2B5+ [33]. Overexpression of Bmi1 in postnatal SVZ NSPC in vitro led to increased self-renewal capacity as observed in embryonic NSPC. However, resilience to apoptosis was noted in NS originating from postnatal NSPC-overexpressing Bmi1. Moreover, we observed increased number of A2B5+ undifferentiated, glial progenitor cells, in these NS. It was therefore important to assess whether Bmi1 overexpression in NSC and NPC in vivo increased tumorigenicity. We show that both embryonic and postnatal activation of Bmi1 overexpression in stem cells and cells derived thereof did not lead to brain tumor formation in mice kept under observation for up to 20 months. This is in agreement with the study of He et al. where tumors were not seen when Bmi1 overexpression was induced in nestin+ progenitor cells. Fasano et al., however, reported that overexpression of Bmi1 by means of intraventricular lentiviral injections led to tumor formation in the first week after birth, although no further details were given on the type of tumor these mice developed. More studies are clearly needed to elucidate these differences. It is important to mention that He et al. and potentially also Fasano et al. achieved higher expression levels of Bmi1 in their systems as compared with our single copy gene approach.


We show that fine tuning of the expression level of the PcG gene Bmi1 is a viable tool to increase self-renewal capacity of both embryonic and postnatal NSPC in the mouse and this approach does not impair the long-term differentiation capacity of postnatal NSPC. However, neuronal differentiation of embryonic NSPC is hampered by concomitant downregulation of the expression of survivin and induction of apoptotic cell death. Upregulation of Bmi1 expression during embryonic development and in the postnatal SVZ does not lead to neoplastic transformation of these cells.


We thank Axel Behrens and Sebastian Brandner for critically reading this article. We are grateful to Gary Warnes for expert support in FACS sorting/analysis, to the BSU staff, in particular, Anthony Price, for help in the daily care of our mouse colony, and to the BICMS Experimental Pathology Facility for processing and cutting paraffin blocks. This work is supported in part by grants of Oncosuisse (OCS01636-02-2005), Cancer Research UK (C23985/A7802), and Medical Research Council UK (G0800020, 85704) to S.M.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.