|Home | About | Journals | Submit | Contact Us | Français|
Precise control of proliferation and differentiation of multipotent neural stem cells (NSCs) is crucial for proper development of the nervous system. Although signaling through the cell surface receptor Notch has been implicated in many aspects of neural development, its role in NSCs remains elusive. Here we examined how the Notch pathway cross talks with signaling for growth factors and cytokines in controlling the self-renewal and differentiation of NSCs. Both Notch and growth factors were required for active proliferation of NSCs, but each of these signals was sufficient and independent of the other to inhibit differentiation of neurons and glia. Moreover, Notch signals could support the clonal self-renewing growth of NSCs in the absence of growth factors. This growth factor-independent action of Notch involved the regulation of the cell cycle and cell-cell interactions. During differentiation of NSCs, Notch signals promoted the generation of astrocytes in collaboration with ciliary neurotrophic factor and growth factors. Their cooperative actions were likely through synergistic phosphorylation of signal transducer and activator of transcription 3 on tyrosine at position 705 and serine at position 727. Our data suggest that distinct intracellular signaling pathways operate downstream of Notch for the self-renewal of NSCs and stimulation of astrogenesis.
In the developing mammalian central nervous system (CNS), neural stem cells (NSCs) serve as the common source of the three major neural cell lineages, i.e., neurons, astrocytes, and oligodendrocytes (41). In early development, NSCs continue self-renewal and expand a pool of undifferentiated cells. These actively proliferating NSCs first give rise to neurons and subsequently differentiate into glia at later stages. Thus, developmental stage-dependent control of the balance between growth and differentiation of NSCs is crucial for proper morphogenesis of the CNS.
A number of extracellular signals have been shown to participate in this control of NSCs. Members of the fibroblast growth factor (FGF) and epidermal growth factor (EGF) families act as mitogens for NSCs (35, 43). These growth factors also regulate the responsiveness of NSCs to gliogenic signals late in development (32, 37, 44). In particular, bone morphogenetic proteins (BMPs) and the interleukin-6 (IL-6) family of cytokines (IL-6, leukemia inhibitory factor, cardiotrophin-1 [CT-1], and ciliary neurotrophic factor [CNTF]) have been shown to act as signals for differentiation of astrocytes (3, 6, 17, 26, 37, 44).
Another important regulatory mechanism in NSCs is signaling through the transmembrane receptor Notch. Notch receptors are activated by specific ligands expressed on the surface of neighboring cells, thereby mediating signals through cell-cell interactions (8, 45). Recent studies have demonstrated that Notch signaling plays important roles in many aspects of CNS development (22, 49). Notably, some studies have proposed that Notch signals induce astrogenesis (9, 14, 15, 21, 40), whereas others reported that the Notch pathway is involved in the maintenance of NSCs (10, 19, 28, 48). Thus, the role for Notch signaling in NSCs remains elusive.
To better understand how the Notch pathway controls NSCs, we performed a series of gain- and loss-of-function experiments using neurosphere culture of embryonic forebrain NSCs. We show that Notch signaling controls multiple aspects of growth and differentiation of NSCs through interactions with growth factor and cytokine signals in a stage-dependent manner. In actively proliferating NSCs, the Notch pathway contributed to the maintenance of the undifferentiated state and active self-renewing growth in collaboration with growth factors. Notch signals also regulated the cell cycle progression and cell-cell interactions of NSCs independently of growth factor signaling. During differentiation of NSCs, however, the Notch pathway acted as a potent inducer of astrocytes in collaboration with the gliogenic signal CNTF. Importantly, the RAM domain of Notch1 receptors was required for the self-renewal and differentiation inhibition of NSCs, whereas it was dispensable for promoting astrogenesis. Thus, different intracellular signaling pathways appear to operate downstream of Notch to regulate NSCs at distinct steps.
All animal procedures were performed according to the guidelines of the Institutional Animal Care and Use Committee and the National Institutes of Health.
Neurosphere culture was performed as described previously (29, 42, 47) with some modifications. Rat embryos at embryonic day 13.5 (E13.5) were collected from timed-pregnant Sprague-Dawley rats and placed in an artificial cerebrospinal fluid (124 mM NaCl, 5 mM KCl, 1.3 mM MgCl2, 2 mM CaCl2, 26 mM NaHCO3, and 10 mM d-glucose). The forebrain neuroepithelium was removed from the rest of embryos under a dissection microscope (Zeiss SV-11) as described previously (42). The resultant tissue was dissociated by incubation in a low-Ca2+, high-Mg2+ artificial cerebrospinal fluid (124 mM NaCl, 5 mM KCl, 3.2 mM MgCl2, 0.1 mM CaCl2, 26 mM NaHCO3, 10 mM d-glucose, 100 units/ml penicillin, and 100 μg/ml streptomycin [Mediatech, Inc., Herndon, VA]) containing 0.05% (wt/vol) trypsin (Sigma-Aldrich, St. Louis, MO), 0.67 mg/ml hyaluronidase (Sigma-Aldrich), and 0.1 mg/ml DNase I (Roche, Indianapolis, IN) at 37°C for 10 min. Subsequently, trypsin was neutralized with 0.7 mg/ml ovamucoid (Sigma-Aldrich), and the resultant tissue suspension was triturated mechanically to yield a single-cell suspension. The cells were filtered through a sterile nylon mesh (40 μm; Becton Dickinson and Company, Franklin Lakes, NJ) and washed twice with a basal medium (a 1:1 mixture of Dulbecco's modified Eagle's medium and F-12 medium [Invitrogen Corp., Carlsbad, CA] containing 100 units/ml penicillin and 100 μg/ml streptomycin). Numbers of viable cells were determined by staining with trypan blue (Sigma-Aldrich).
Neurosphere culture was initiated by seeding cells at a density of 1 × 105 to 2 × 105 viable cells/ml in the basal medium supplemented with the B-27 supplement (Invitrogen Corp.), 20 ng/ml bovine FGF2 (R&D Systems, Inc., Minneapolis, MN), 20 ng/ml mouse EGF (Roche), 2 μg/ml heparin (molecular weight of 3,000; Sigma-Aldrich), and 1 mg/ml bovine serum albumin (BSA) (Sigma-Aldrich) (growth medium). The surfaces of culture dishes were coated with poly(2-hydroxy-ethyl methacrylate) (polyHEMA) (1.6 mg/cm2; Sigma-Aldrich) to prevent cell attachment as described previously (29, 42, 47).
The following cDNA fragments were obtained by PCR and cloned into the replication-defective recombinant retrovirus vector pMXIG as described previously (29, 46, 47): constitutive-active Notch1 (ca-Notch1), amino acid residues 1744 to 2372 of mouse Notch1 (GenBank accession number Z11886); ca-Notch1ΔRAM, amino acid residues 1848 to 2372 of mouse Notch1; and dominant-negative Delta-like 1 (dn-Dll1), amino acid residues 1 to 582 of mouse Dll1 (GenBank accession number NM_007865). cDNAs for rat Hes1 and Hes5 and mouse Id1 and Id3 were kind gifts from R. Kageyama (Kyoto University, Japan) and T. Taga (Kumamoto University, Japan), respectively. The retrovirus packaging cell line Plat-E (24) was transiently transfected with these virus constructs by the FuGene-6 lipofection method (Roche), and viruses were collected as the culture supernatant at 2 days after transfection as described previously (24). The expression of transgene protein products in virus-infected cells was confirmed by Western blotting and immunostaining.
Neurosphere cells were infected with retroviruses under two different conditions. In the early-infection experiment, neurospheres were dissociated and incubated with retroviruses in the presence of 4 μg/ml hexadimethrine bromide (Sigma-Aldrich) for 8 h. Subsequently, the cells were maintained in floating culture for 3 days in the growth medium. During this culture period, 30 to 60% of the cells expressed enhanced green fluorescent protein (GFP). The resultant neurospheres were subjected to either neurosphere formation or differentiation assay. In the late-infection experiment, neurospheres were dissociated and seeded onto poly-d-lysine (PDL) (100 μg/ml; Sigma-Aldrich)-coated chambers. At 2 days after plating (DAP2), the cells were infected with retroviruses for 8 h. Subsequently, the cells were maintained for additional 6 days in growth medium without FGF2 and EGF. In all infection experiments, the pMXIG virus without cDNA inserts was used as a control.
Virus-infected neurosphere cells were dissociated, and GFP-positive cells were purified by fluorescence-activated cell sorting using FACS Vantage (Becton Dickinson). The resultant single cells were plated in polyHEMA-coated six-well plates at a density of 2 × 104 cells per well in growth medium containing 0.8% (wt/vol) methylcellulose (Nacalai Tesque, Inc., Kyoto, Japan). The number of GFP-positive cell clusters with a diameter of over 30 μm was counted at 1 week after plating under a DMIRB/SLR Leica inverted fluorescence microscope equipped with a Hamamatsu Photonics C5985 charge-coupled device camera (Hamamatsu, Japan). In the neurosphere formation assay in the absence of growth factors, virus-infected cells were seeded at a density of 2 × 104 cells (low density) or 1 × 105 cells (high density) per well on polyHEMA-coated six-well plates containing methylcellulose. The number of GFP-positive cell clusters was counted at 3 weeks after plating. To culture neurospheres on the adhesive surface, culture dished without polyHEMA coating were used, and clustering and dispersed GFP-positive cell colonies were distinguished under a fluorescence microscope. In some experiments, cells were treated with the following reagents: human CNTF (50 ng/ml; Sigma-Aldrich), human BMP4 (10 ng/ml, R&D Systems, Inc.), the γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl)-l-alanyl]-(S)-phenylglycine t-butyl ester (DAPT) (1 μM; Sigma-Aldrich), the FGF receptor inhibitor 3-[3-(2-carboxyethyl)-4-methylpyrrol-2-methylidenyl]-2-indolinone (SU5402) (25 μM; EMD Biosciences, Inc., San Diego, CA), and the EGF receptor inhibitor 4-[(3-bromophenyl)amino]-6-acrylamidoquinazoline (PD168393) (25 nM; EMD Biosciences, Inc.).
Virus-infected neurospheres were dissociated and seeded onto PDL-coated glass chambers at a density of 6 × 104 cells/cm2. The cells were incubated for defined periods in growth medium without FGF2 and EGF and subjected to immunostaining. In some experiments, dividing cells were labeled with 5-bromo-2′-deoxyuridine (BrdU) (1 μM, Sigma-Aldrich) for 2 h.
pHes1-Luc contained the 354-bp genomic fragment of the mouse Hes1 gene (positions −193 to +161; GenBank accession number D16464) upstream of the luciferase reporter pGVB (TOYO-INKI) (46). pE7βA-Luc contained seven repeats of the E-box motif (CAGGTG) upstream of the β-actin core promoter (46). Luciferase reporter assays were performed using the NSC line MNS-70 as described previously (46). The cells were grown at a density of 6 × 104 cells per well of six-well plate, and transient transfection was performed by the FuGene-6 lipofection method (Roche). Each well received an appropriate reporter plasmid (pHes1-Luc [0.5 μg] or pE7βA-Luc [2 μg]) in combination with pEF-BOS plasmids harboring ca-Notch1, ca-Notch1ΔRAM, and the full-length rat Mash1. The β-galactosidase expression plasmid pEF-BOS-β-gal was included to normalize the results for transfection efficiency as described previously (46).
Immunostaining of cultured cells was performed as described previously (29, 47). Affinity-purified rabbit polyclonal antibodies (PAbs) neurogenin 2 (Ngn2) (1:10,000), Pax6 (1:1,000) (47), and Prox1 (1:3,000) (42) were described previously. Rabbit PAb for Sox2 (1:1,000) was prepared by immunization with a synthetic oligopeptide that correspond to the N-terminal amino acid sequence of rat Sox2. Mouse monoclonal antibody (MAb) against nestin (Rat401; 1:500) was obtained from the Developmental Studies Hybridoma Bank of the University of Iowa (Iowa City, IA). Antibodies for the following antigens were purchased from commercial sources: Mash1 (mouse MAb; 1:1,000; BD Biosciences, San Jose, CA), β-tubulin type III (TuJ1) (mouse MAb; 1:2,000; Babco, Richmond, CA), HuC/D (mouse MAb, clone 16A11; 1:1,000; Invitrogen Corp.), glial fibrillary acidic protein (GFAP) (mouse MAb [clone G-A-5; 1:1,000] and rabbit PAb [1:1,000; Millipore, Billerica, MA]), S100β (mouse MAb, clone SH-B1; 1:1,000; Sigma-Aldrich), O4 (mouse immunoglobulin M MAb; 1:400; Millipore), galactocerebroside (mouse MAb; 1:200; Millipore), GFP (mouse MAb [1:500] and rabbit PAb [1:5,000] [Invitrogen Corp.] and rat MAb [1:5,000, Nacalai Tesque, Inc.]), BrdU (mouse MAb; 1:200, BD Biosciences), p27 (mouse MAb; 1:500; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), signal transducer and activator of transcription 3 (STAT3) phosphorylated at the serine at position 727 (STAT3-pS727) (mouse MAb; 1:1,000; BD Biosciences), and STAT3-pY705 (rabbit PAb; 1:1,000; Cell Signaling Technology, Inc., Danvers, MA). Immunoreactive cells were visualized by staining with appropriate sets of secondary antibodies conjugated with Alexa Fluor 350, 488, and 594 (1:400; Invitrogen Corp.). To count cell numbers, cell nuclei were stained with 1 μg/ml bis-benzimide (Invitrogen Corp.). To examine cell death, DNA fragmentation was detected by the terminal deoxyribonucleotidyl transferase-mediated dUTP-digoxigenin nick end labeling (TUNEL) method using the In Situ Cell Death Detection Kit (Roche). Fluorescent images were observed and captured using a Zeiss Axiophoto2 instrument equipped with a Hamamatsu Photonics C5810 charge-coupled device camera.
Forebrain neuroepithelia were dissected from rat embryos at E13.5, E15.5, and E18.5. The resultant tissues and neurospheres derived from E13.5 forebrains were lysed in a buffer containing 50 mM HEPES-NaOH (pH 7.5), 100 mM KCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 1 mM EGTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 0.5% protease inhibitor cocktail (Sigma-Aldrich), 1 mM Na3VO4, 10 mM NaF, and 20 mM β-glycerophosphate. The resultant extracts were sonicated and centrifuged at 20,000 × g for 15 min at 4°C to obtain cleared cell lysates. Protein concentrations were determined using the Bio-Rad protein assay kit (Bio-Rad Laboratories, Inc., Hercules, CA) with BSA as a standard. Equal amounts of proteins were loaded, separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and transferred to Immobilon membranes (NC; Millipore). The membranes were blocked with 1% (wt/vol) BSA in phosphate-buffered saline containing 0.1% Tween 20 and blotted with antibodies for STAT3-pS727 (1:250), STAT3-pY705 (1:1,000), STAT3 (mouse MAb; 1:2,500, BD Biosciences), and lamin B (goat PAb; 1:200; Santa Cruz Biotechnology, Inc.). Immunoreactive bands were visualized with secondary horseradish peroxidase-conjugated antibodies (GE Healthcare, Piscataway, NJ) and ECL reagents (GE Healthcare).
The quantitative results are expressed as means ± standard deviations (SD), and the number of replicated experiments is indicated in Results or in the figure legends. Statistical analyses were performed with the two-tailed unpaired t test.
To delineate multiple regulatory steps in the growth and differentiation of NSCs, we used neurosphere culture of the forebrain neuroepithelium derived from E13.5 rat embryos (roughly corresponding to E11.5 in mice). When plated on the nonadhesive culture surface in the presence of FGF2 and EGF (here collectively called GFs), 4.7% ± 0.5% (n = 6) of the initially seeded cells formed clonally expanding neurospheres. The vast majority of cells comprising these neurospheres expressed nestin and Sox2, and virtually no cells (<0.1%) expressed markers for neurons or glia (Fig. 1A, A′, and D). After passage, 13.4% ± 0.6% (n = 3) of these cells were capable of forming secondary neurospheres (Fig. (Fig.1B;1B; also see Fig. Fig.5).5). Moreover, these passaged spheres maintained the ability to give rise to neurons, astrocytes, and oligodendrocytes (Fig. 1C and C′). Thus, these cells retained the capacity of self-renewal and multilineage differentiation, two cardinal properties of NSCs.
We next followed the time course of differentiation of these neurospheres. Upon removal of GFs, nestin- and Sox2-positive undifferentiated cells gradually decreased, and 45.8% ± 3.0% (n = 3) of the cells became TuJ1-positive neurons at DAP2 (Fig. (Fig.1D).1D). Following this early neuronal differentiation, GFAP-positive astrocytes and O4-positive oligodendrocytes emerged at later time points: 8.1% ± 0.1% and 3.4% ± 0.2% (n = 3) of total cells became GFAP positive and O4 positive, respectively, at DAP6 (Fig. (Fig.1D).1D). We observed a similar temporal sequence of neurogenesis and gliogenesis by staining for HuC/D, S100β, and galactocerebroside (data not shown). We further used three markers to detect intermediate progenitors. The proneural helix-loop-helix transcription factors Ngn2 and Mash1 and the homeodomain factor Prox1 are expressed in intermediate progenitors that emerge during differentiation of NSCs into neurons and glia (5, 42, 46). Consistent with their temporal expression patterns in vivo, Ngn2-positive and Mash1-positive cells emerged at DAP1, which was followed by the peak of Prox1-positive cells at DAP2 in neurosphere culture (Fig. (Fig.1E).1E). Thus, two sequential steps were distinguished along the course of differentiation of NSCs: the early transition from NSCs to intermediate progenitors and the subsequent differentiation of neurons and glia (Fig. (Fig.1F1F).
For genetic manipulations of NSCs, we used the GFP-expressing recombinant retrovirus vector pMXIG (24, 46, 47). When infected with pMXIG viruses, growing NSCs and their neuronal and glial progeny stably expressed GFP (Fig. 1G to K). Importantly, virus-infected, GFP-positive cells and noninfected, GFP-negative cells within the same culture exhibited an essentially identical capacity for self-renewal and differentiation. The overall growth and differentiation of cells were also indistinguishable between virus-infected and uninfected cultures (data not shown). Thus, the virus infection procedure did not select specific subpopulations of cells or alter the properties of NSCs.
Previous studies reported that alterations of Notch signaling did not affect the cell fate choice between neurons and glia (10, 19, 28, 48). We sought to reevaluate this issue by a combination of gain- and loss-of-function experiments. To stimulate Notch signaling, we overexpressed two constitutive-active forms of Notch1: one contained the intracellular domain of Notch1 except for the C-terminal 159 amino acid residues (ca-Notch1), whereas the other had an additional deletion of the RAM domain (ca-Notch1ΔRAM) (46) (Fig. (Fig.1L).1L). To inhibit Notch signaling, we used a dominant-negative form of the Notch receptor ligand Dll1 (dn-Dll1), in which most of the intracellular domain was deleted (47) (Fig. (Fig.1L).1L). We also used the synthetic γ-secretase inhibitor DAPT to block the ligand-dependent cleavage and activation of Notch receptors (12).
Unlike control virus-infected cells, the vast majority of cells expressing ca-Notch1 remained as nestin-positive and Sox2-positive cells even after removal of GFs (Fig. 2A and B). They neither generated Mash1-positive and Prox1-positive intermediate progenitors at DAP2 (Fig. (Fig.2C)2C) nor differentiated into neurons or glia at DAP6 (Fig. (Fig.2A).2A). In contrast, when ca-Notch1ΔRAM was overexpressed, the percentage of GFAP-positive astrocytes among GFP-positive cells became much higher than that in the control culture at the expense of TuJ1-positive neurons and O4-positive oligodendrocytes. Inhibition of Notch signaling caused the opposite effects: cells expressing dn-Dll1 or treated with DAPT generated more Mash1-positive and Prox1-positive cells at DAP2 (Fig. (Fig.2C)2C) and differentiated into neurons at a higher percentage at the expense of astrocytes (Fig. (Fig.2A).2A). DAPT treatment always exerted effects similar to those of dn-Dll1 in neurosphere culture (see below), supporting the idea that it acted primarily as an inhibitor for Notch signaling in our culture. The rate of cell death, measured as the frequency of TUNEL-positive cells among total cells (2 to 3% and 18 to 20% at DAP1 and DAP6, respectively), did not significantly differ among cultures infected with different viruses, indicating that selective elimination of particular cell lineages did not account for the observed differences. Moreover, differentiation of GFP-negative, uninfected cells within the same culture occurred similarly to that in the control, demonstrating that virus-mediated modulations of Notch signaling affected the cell fate in a cell-autonomous manner. Notch ligands lacking the cytoplasmic domain, similar to dn-Dll1 used here, have been shown to inhibit Notch signaling in both cell-autonomous and non-cell-autonomous manners (18, 36, 38). In our culture, however, overexpression of dn-Dll1 affected the phenotypes of virus-infected cells but not surrounding uninfected cells, demonstrating that it acted predominantly in a cell-autonomous manner.
These results suggest that Notch signaling exerts different actions at two distinct steps: the first is to block early transition from undifferentiated NSCs to intermediate progenitors, and the second is to control cell fates at a later step. To test this idea, cells were infected with viruses 2 days after induction of differentiation. In this late-infection experiment, the majority of cells passed through the early differentiation step and many of them became postmitotic neurons. Thus, the major targets for virus infection were intermediate progenitors and cells that were committed to specific lineages but not yet fully differentiated. As described above, when cells were infected with ca-Notch1 viruses before induction of differentiation, few ca-Notch1 expressing cells could become neurons or glia (2.0% ± 0.3% of total GFP-positive cells; n = 6). In contrast, when the same viruses were used for late infection, a significant fraction of ca-Notch1-expressing cells differentiated (15.5% ± 1.9%; n = 3) (Fig. (Fig.2D),2D), and the vast majority of such differentiated progenies (about 95%) were astrocytes (Fig. (Fig.2E).2E). Moreover, late infection with ca-Notch1ΔRAM viruses markedly increased the percentage of astrocytes, in parallel with a large (2.6-fold) increase of total differentiated cells (neurons plus glia) in culture (44.5% ± 3.1%; n = 3) (Fig. (Fig.2E).2E). Conversely, dn-Dll1 suppressed astrocyte differentiation and enhanced neurogenesis under the same conditions. These results suggest that Notch signals selectively stimulate astrogenesis once NSCs are released from the early differentiation block. The marked increase of GFAP-positive cells by ca-Notch1ΔRAM also suggests that this ca-Notch1 construct not only stimulates early specification of astrocytes but also promotes subsequent expansion of astrocyte progenitors and their maturation.
In the experiments described above, two ca-Notch1 constructs exhibited significantly different activities in NSCs. We thus compared their signaling activities using two transcriptional reporters (Fig. (Fig.3).3). We have previously shown that ca-Notch1 can both activate the transcription driven by the Hes1 enhancer and inhibit the activity of Mash1 (46). Consistent with previous reports (2, 27), ca-Notch1ΔRAM activated the Hes1 enhancer-luciferase construct pHes1-Luc to the same extent as ca-Notch1 (Fig. (Fig.3A).3A). Activation of this reporter depends on RBP-J (also called CBF-1) (45), suggesting that the RBP-J-dependent signaling pathway remains active downstream of ca-Notch1ΔRAM. Nevertheless, unlike ca-Notch1, ca-Notch1ΔRAM could not inhibit the Mash1-dependent activation of the E-box-containing reporter pE7βA-Luc (Fig. (Fig.3B).3B). Given that Mash1 plays a crucial role in neuronal differentiation (42, 46), this inhibition of Mash1 underlies, at least in part, ca-Notch1-dependent inhibition of the early differentiation of NSCs. These results can be explained by the idea that ca-Notch1 exerts its action through two distinct signaling pathways and that ca-Notch1ΔRAM is defective in activating one of these pathways. Activation of this pathway requires the RAM domain but appears to occur independently of the RBP-J/CBF-1-dependent transcriptional activation of Hes1. These data, however, do not exclude the possibility that ca-Notch1ΔRAM was a weak allele of ca-Notch1 and that its activity was sufficient for activation of pHes1-Luc but not for inhibition of Mash1.
We next examined the relationship between Notch and GF signaling. When neurosphere cells were kept exposed to GFs, they did not differentiate into either neurons or glia even after being dissociated and cultured in monolayer (Fig. (Fig.4A).4A). Like ca-Notch1, GFs suppressed differentiation at or before the step of induction of Mash1-positive intermediate progenitors (Fig. (Fig.4B).4B). Unlike the results of a previous report (13), this GF-dependent differentiation inhibition occurred under conditions in which Notch signals were blocked by either dn-Dll1 (Fig. 4C and D) or DAPT (Fig. (Fig.4E).4E). Conversely, ca-Notch1 could inhibit differentiation of neurospheres in the absence of GFs (Fig. (Fig.2A).2A). Thus, GFs and Notch signals appeared to be capable of blocking differentiation of NSCs independently of each other.
We next examined their relationship in terms of the regulation of the self-renewal activity of NSCs. Virus-infected cells were seeded at a clonal density (10 cells/μl) into a methylcellulose matrix to prevent spontaneous cell aggregation. Subsequently, the frequencies of cells capable of forming clonal secondary neurospheres were compared among cells infected with different viruses. dn-Dll1 strongly attenuated the formation of neurospheres in the presence of GFs (39% of the control level) (Fig. (Fig.5A).5A). Conversely, ca-Notch1, but not ca-Notch1ΔRAM, significantly increased the frequency of neurosphere formation in the presence of GFs (Fig. (Fig.5B5B).
The above results demonstrate that both GFs and Notch signaling are required for the active self-renewing growth of NSCs. In fact, few cells formed neurospheres without GFs in control culture (0.022% ± 0.006% of initially seeded cells) (Fig. (Fig.5C).5C). We found, however, that a small but significant fraction of cells expressing ca-Notch1 formed clonal colonies even in the absence of GFs (0.47% ± 0.1%; n = 3) (Fig. (Fig.5C).5C). These colonies grew slowly and were composed of nestin-positive and Sox2-positive undifferentiated cells (Fig. 5C′ to C and data not shown). After passage, these neurospheres could form secondary spheres at a frequency similar to that for cells expanded in the presence of GFs (8.9% ± 0.6% of total cells; n = 6). In contrast, ca-Notch1ΔRAM did not have the ability to support such GF-independent growth (Fig. (Fig.5C).5C). Overexpression of the inhibitory helix-loop-helix factors Hes1, Hes5, Id1, and Id3, which are known downstream effectors for Notch signals (30, 34), did not fully recapitulate the activity of ca-Notch1 either (Fig. (Fig.5C5C and data not shown), suggesting that molecules other than these factors are involved downstream of ca-Notch1.
The above-described neurosphere formation assays were performed at a clonal density (10 cells/μl). Under this condition, individual single cells need to survive and divide into two cells to receive Notch signals through cell-cell interactions. However, cells divide only slowly or do not survive well without GFs. Thus, it is likely that not all cells capable of forming clonal colonies can actually grow as neurospheres in such a condition. This could be a reason why more cells formed neurospheres when ca-Notch1 was expressed. To test this idea, we next asked if cells could form growing colonies at a higher frequency when they were maintained at a higher density and thereby had a higher chance to receive Notch signals during the initial period of their survival and growth. In fact, when cells were seeded at a five-times-higher density (50 cells/μl), growing colonies were formed at a much higher frequency (an 80-fold increase; 177.9 ± 14.2 colonies/1 × 104 cells) in the absence of GFs (Fig. (Fig.5D).5D). Although these colonies were not necessarily clonal, they remained as undifferentiated cells and gave rise to neurons and glia when induced to differentiate (data not shown). Importantly, overexpression of dn-Dll1 and treatment with DAPT strongly suppressed the formation of these colonies, and conversely, ca-Notch1-expressing cells formed colonies at a higher frequency under this condition. These results demonstrate that the survival and growth of NSCs without GFs strongly depend on cell-cell interactions mediated by Notch signals.
A possible explanation for this GF-independent growth is that Notch regulates the expression and/or secretion of endogenous growth-promoting factors, which in turn act on NSCs in a paracrine and/or autocrine fashion. We first tested this idea by using inhibitors for FGF and EGF signaling. The tyrosine kinase inhibitors SU5402 and PD168393 selectively blocked the FGF2- and EGF-dependent formation of neurospheres, respectively (Fig. (Fig.5E).5E). Neither single nor combined treatment with these inhibitors, however, inhibited the clonal expansion of ca-Notch1-expressing neurospheres in the absence of GFs (Fig. (Fig.5F).5F). Thus, it is unlikely that signaling for endogenous FGF2, EGF, and/or related growth factors accounts for the Notch-dependent formation of neurospheres. Yet, it could be possible that growth-promoting factors other than the members of the FGF and EGF families are involved in Notch-dependent growth. However, in culture of ca-Notch1 virus-infected cells, stimulation of clonal growth was observed only for virus-infected cells and not for surrounding uninfected cells, suggesting that the contribution of the non-cell-autonomous action of Notch is minimum.
We found two additional GF-independent actions of Notch. To clonally expand as neurospheres, NSCs need to be maintained on the nonadhesive culture surface. When plated on the adhesive surface, cells remained undifferentiated and continued to divide, but they could not form tightly packed floating clusters even in the presence of GFs (Fig. 6A and A′). Although some cells initially formed small aggregates, they eventually collapsed and dispersed on the culture surface (Fig. (Fig.6C).6C). Therefore, the growth of undifferentiated cells and formation of neurosphere colonies were separable under this condition, and GFs were sufficient for the former but not for the latter. Importantly, cells expressing ca-Notch1 maintained tight cell clusters even under this condition (Fig. 6B and B′). These cells formed floating colonies at a much higher frequency than control cells and kept growing as floating clusters for over 3 weeks (Fig. (Fig.6C).6C). Thus, one of the crucial functions of Notch signaling independent of GFs is to enhance the cell-cell interactions of self-renewing NSCs.
We also found that the growth of neurospheres became significantly slower when ca-Notch1 was overexpressed. In the presence of GFs, more than 50% of control cells formed clonal neurospheres with diameters larger than 50 μm at DAP7 (average diameter, 62.8 μm; n = 647) (Fig. (Fig.6D).6D). In contrast, the majority of ca-Notch1-expressing cells formed much smaller spheres under the same conditions (average diameter, 40.7 μm; n = 249). Thus, strong Notch signals appear to partially counteract the growth-promoting action of GFs. In fact, when control and ca-Notch1-expressing cells were labeled with BrdU for 2 h, the percentage of BrdU-labeled cells was significantly decreased by ca-Notch1 in both the presence and absence of GFs (Fig. (Fig.6E).6E). Nevertheless, the percentage of cells positive for the cyclin-dependent kinase inhibitor p27, which represented the fraction of cells that exited the cell cycle, was also lower in culture of ca-Notch1-expressing cells than in the control (Fig. (Fig.6E).6E). Thus, cells with a high Notch activity divided slowly, yet a large fraction of them remained in the mitotic cycle. Altogether, these results demonstrate that Notch signaling controls multiple aspects of the self-renewal mode of growth of NSCs independently of GFs.
As described above, Notch signaling promotes astrogenesis at the late differentiation step. The IL-6 and BMP families of cytokines also play important roles in differentiation of astrocytes (3, 6, 17, 26, 37, 44). We thus asked how the Notch pathway cross talks with these cytokine signals. We chose CNTF and BMP4 as representatives of IL-6 and BMP family cytokines, respectively.
As shown in previous studies, both cytokines markedly increased the percentage of GFAP-positive astrocytes among total cells (Fig. (Fig.7A).7A). BMP4 also decreased the fractions of both neurons and oligodendrocytes, whereas CNTF reduced only neurons (Fig. 7B and C). Importantly, CNTF, but not BMP4, counteracted the suppression of astrogenesis by ca-Notch1 (Fig. (Fig.7A).7A). In ca-Notch1-expressing cells, this CNTF-dependent induction of astrocytes did not accompany the release of the inhibition of neuronal and oligodendroglial differentiation (Fig. 7B and C). Thus, when CNTF and Notch signals were combined, NSCs differentiated selectively into astrocytes. Conversely, dn-Dll1 (Fig. (Fig.7A)7A) and DAPT (Fig. (Fig.7D)7D) significantly attenuated the CNTF-dependent induction of astrocytes, whereas the inhibition of neurogenesis by CNTF remained unchanged (Fig. (Fig.7B).7B). Thus, CNTF-dependent induction of astrocytes appears to require endogenous Notch signals. dn-Dll1 and DAPT also inhibited astrocyte differentiation in the control culture (Fig. 7A and D). It has recently been shown that the CNTF-related factor CT-1 is secreted from neurons and acts as the major astrocyte-inducing signal (3). Thus, the above results can be explained by the idea that endogenous CT-1 induces differentiation of astrocyte in control culture and that it requires cooperation with endogenous Notch signaling to fully exert its action.
A similar cross talk was observed between CNTF and GFs. Without CNTF, GFs strongly blocked astrocyte differentiation at both low (2 ng/ml each for FGF2 and EGF) and high (20 ng/ml) concentrations (Fig. (Fig.7E).7E). When combined with CNTF, however, GFs significantly augmented the CNTF-dependent induction of astrocytes. We also found that CNTF markedly decreased the frequency of neurosphere formation in the presence of GFs. Such an inhibitory effect of CNTF was observed even in cells expressing ca-Notch1 (Fig. (Fig.7F).7F). Thus, CNTF blocked the Notch- and GF-dependent self-renewal of NSCs. These results suggest that the mode of cross talk between CNTF, Notch, and GF signals is distinct at the early and late differentiation steps. It appears that CNTF counteracts the differentiation block by Notch and GF signals at the early step. At the late step, however, CNTF collaborates with Notch and GFs to selectively induce astrocytes.
Recent studies have shown that the transcription factor STAT3 plays important roles in both self-renewal and astrogenesis (1, 3, 7, 17, 21, 25, 33, 37, 44, 50). We thus examined whether the cross talk between the Notch, GF, and CNTF pathways occurs at the level of STAT3. We examined the phosphorylation of STAT3 at the tyrosine at position 705 (Y705) and the serine at position 727 (S727), both of which are critical for STAT3-dependent transcriptional activation (20).
We found that CNTF, GFs, and ca-Notch1 all induced nuclear accumulation of STAT3-pS727 (Fig. 8A to C and I). However, ca-Notch1ΔRAM, which lacked the ability to support GF-independent growth of NSCs, did not induce STAT3-pS727-positive cells. In contrast, STAT3 phosphorylation on Y705 was not induced by ca-Notch1 or GF treatment alone (<1.0% of total cells) and was induced only weakly by CNTF alone (15.5% ± 4.2%; n = 3) (Fig. 8E to G and I). The combination of CNTF with ca-Notch1 or ca-Notch1ΔRAM, however, resulted in a marked increase of STAT3-pY705-positive cells (Fig. 8H and I). Combinatorial actions of CNTF and ca-Notch1 were also observed regarding the phosphorylation on S727 (Fig. 8D and I). These results demonstrate that the phosphorylation status of STAT3 is differentially regulated in self-renewing cells and in cells induced to differentiate into astrocytes. When Notch and GF signals maintained undifferentiated NSCs, STAT3 was phosphorylated preferentially on S727. In contrast, when CNTF signals acted on such cells in addition to Notch and GFs, STAT3 was phosphorylated on both S727 and Y705 in parallel to astrocyte differentiation. Interestingly, ca-Notch1ΔRAM, which exhibited a strong astrocyte-inducing activity, was a very weak inducer of phosphorylation of S727 and Y705 by itself but strongly augmented the CNTF-dependent response. This result reinforces the idea that Notch-dependent induction of astrocytes occurs in collaboration with endogenous CNTF-related factors such as CT-1 (3).
Lastly, we asked if similar stage-dependent regulation of STAT3 phosphorylation occurs in vivo. In the developing forebrain, NSCs at E13.5 actively proliferate and generate predominantly neurons. Astrogenesis induced by CNTF and related factors such as CT-1 begins at around E15.5, and many GFAP-positive cells differentiate at E18.5 (reference 3 and references therein). In line with this stage-dependent gliogenesis in vivo, the phosphorylation level of STAT3-Y705 was low at E13.5 but progressively increased between E13.5 and E18.5 (Fig. (Fig.8J).8J). In contrast, phosphorylation of STAT3 on S727 was detected at a significant level at E13.5 and remained relatively unchanged between E13.5 and E18.5, in accordance with the persistence of proliferating NSCs throughout development.
Notch signaling plays vital roles in development of a variety of tissues and organs across species (8, 45). In this study, we have shown that the Notch pathway controls multiple aspects of proliferation and differentiation of NSCs through cross talk with growth factor and cytokine signaling pathways.
The formation of neurospheres has been widely used as a quantitative measure of the self-renewing activity of NSCs (references 35 and 43 and references therein). We used this in vitro assay to examine how Notch signaling controls NSCs. In accordance with previous studies (10, 19, 28, 48), inhibition of Notch signals attenuated the self-renewal of NSCs, and conversely, constitutive active Notch1 enhanced this activity (Fig. (Fig.9,9, left). Thus, Notch signaling is crucial for the maintenance of the self-renewing activity of NSCs. Yoon et al. (48) proposed that this is attributable to the regulation of responsiveness to mitogenic GFs (FGF2 and EGF) by Notch signals. In fact, both GFs and Notch were required for the active proliferation of NSCs in vitro. Our data, however, have also demonstrated the actions of Notch independent of GFs. Notch signaling alone could support the clonal self-renewing growth of NSCs by inhibiting their neuronal and glial differentiation, enhancing cell-cell interactions, and regulating the cell cycle progression independently of exogenous and endogenous GFs. Notch signals have also been shown to support NSC survival (16, 23, 31). In line with this, we found that ca-Notch1 increased the expression of a selective set of cell adhesion molecules, including neural cell adhesion molecule, vascular cell adhesion molecule, cadherin-11, and cadherin-13, and downregulated the expression of many cell cycle regulators such as cyclins A2, B1, and D1 (our unpublished data). Thus, Notch regulates multiple critical aspects of the self-renewal mode of cell divisions both independently of and in collaboration with GFs. Interestingly, Notch and GF signals have been shown to act in a mutually antagonistic manner in certain developmental events (39). Therefore, the mode of their cross talk appears to be cell type dependent.
Previous studies have reported apparently contradictory results regarding the action of Notch signaling in differentiation of NSCs. Some studies have proposed that Notch is not involved in the lineage selection between neurons and glia (19, 28, 48), whereas others have shown that Notch selectively promotes differentiation of astrocytes (9, 14, 15, 40). Our data suggest that this apparent discrepancy is likely to be due to differential actions of Notch at distinct steps during the course of differentiation (Fig. (Fig.9).9). When overexpressed before differentiation induction, ca-Notch1 inhibited both neurogenesis and gliogenesis. This early differentiation block occurred at or before the induction of intermediate progenitors such as Mash1-positive and Prox1-positive cells. In contrast, when ca-Notch1 was expressed at a later stage of differentiation, it selectively stimulated astrogenesis. Conversely, inhibition of Notch signaling by dn-Dll1 and DAPT resulted in enhanced neurogenesis at the expense of astrogenesis. These results can be explained by the idea that Notch regulates differentiation of NSCs at two steps: the first step is the transition from NSCs to intermediate progenitors, and the second step is the subsequent lineage selection between neurons and astrocytes.
Our data also suggest that distinct intracellular pathways operate downstream of Notch at these two steps. Unlike ca-Notch1, ca-Notch1ΔRAM neither blocked differentiation of NSCs at the early step nor supported the formation of neurospheres without exogenous GFs. Moreover, ca-Notch1ΔRAM did not inhibit Mash1-dependent transcriptional activation or induce phosphorylation of STAT3 on S727. Nevertheless, ca-Notch1ΔRAM promoted astrogenesis and activated the transcription of the Hes1 promoter and thus retained a part of the activity of the RAM domain-containing form. These results suggest two distinct signaling pathways downstream of Notch1: one is RAM domain dependent, and the other is RAM domain independent. Both pathways appear to be required for the maintenance of undifferentiated NSCs, whereas the former is dispensable for stimulation of astrogenesis.
A well-known molecule that binds to the RAM domain of Notch receptors is RBP-J/CBF-1 (45). It has been shown, however, that Notch1 lacking the RAM domain still binds to RBP-J/CBF-1 and activates the Hes1 and Hes5 promoters (2, 27). In fact, ca-Notch1ΔRAM could activate the RBP-J/CBF-1-dependent reporter pHes1-Luc similarly to ca-Notch1. This result suggests that the RBP-J-dependent pathway remains active downstream of ca-Notch1ΔRAM and thus does not account for the difference in the biological activity between ca-Notch1 and ca-Notch1ΔRAM. Our data have also shown that overexpression of Hes1 or Hes5 does not fully mimic the activity of ca-Notch1. Thus, a signaling pathway(s) other than the RBP-J-Hes pathway appears to underlie the difference between ca-Notch1 and ca-Notch1ΔRAM. Our data, however, do not formally exclude the possibility that ca-Notch1ΔRAM acted as a weak allele of ca-Notch1. Deltex, which binds to both ca-Notch1 and ca-Notch1ΔRAM, could also be involved the Notch-dependent regulation of NSCs (reference 46 and references therein). Overexpression of Deltex1 or Deltex2, however, did not recapitulate the action of ca-Notch1 in neurosphere culture (our unpublished data). Thus, details of this RAM domain-dependent Notch pathway currently remain unknown.
Given the two regulatory steps described above, what signal releases the Notch-dependent early differentiation block in NSCs? Our data suggest that CNTF is one such signal. CNTF could induce astrogenesis not only by control cells but also by ca-Notch1-expressing cells which otherwise differentiated into neither neurons nor glia. Yet, ca-Notch1 still blocked differentiation of neurons and oligodendrocytes in the presence of CNTF. CNTF also overcame the differentiation inhibition by GFs. These results are consistent with the idea that CNTF releases NSCs from the Notch- and GF-dependent inhibition at the early differentiation step and subsequently biases the cell fate toward the astrocyte lineage. In line with this idea, CNTF strongly inhibited the self-renewal of NSCs even in the presence of ca-Notch1 and GFs. Thus, CNTF appears to counteract two critical signaling pathways for the maintenance of NSCs. A recent study has demonstrated that CT-1 is the major CNTF-related cytokine for induction of astrocytes in the developing forebrain (3). Thus, CT-1 is a likely to be an endogenous signal that blocks Notch and GF signals in vivo. Interestingly, BMP4, which is also known to act as a potent astrocyte-inducing signal, did not counteract ca-Notch1 in terms of the induction of astrocytes. Thus, the actions of CNTF and BMPs are distinct when NSCs continue self-renewal in a Notch-dependent manner.
Despite such a counteracting action of CNTF on Notch signals regarding the self-renewal of NSCs, CNTF required endogenous Notch signals to stimulate astrogenesis. CNTF-dependent astrogenesis was also augmented by GFs. Thus, Notch and GFs appear to have two roles in gliogenesis in vivo. During early development, when NSCs actively proliferate and preferentially give rise to neurons, they act to preserve a pool of undifferentiated NSCs for later gliogenesis. Subsequently, when gliogenic signals emerge at a later stage, Notch and GF signals collaborate with such signals to actively promote gliogenesis.
It is noteworthy that recent studies have proposed that CNTF and leukemia inhibitory factor promote the maintenance of NSCs in the adult brain (4, 10). Thus, the modes of the cross talk between the CNTF and Notch pathways appear to be distinct in embryonic and adult NSCs. Interestingly, NSCs in the adult neurogenic niche have been shown to share some common properties with astrocytes (11). A possible scenario thus could be that in embryonic NSCs, CNTF counteracts the Notch intracellular pathway responsible for their maintenance and instead selectively cooperates with another pathway to promote astrogenesis. In contrast, CNTF could synergize with both pathways in adult NSCs so that NSCs maintain the self-renewing activity yet acquire some properties of astrocytes.
Recent studies have shown that STAT3 plays an important role in both the differentiation of astrocytes and maintenance of NSCs (1, 3, 7, 17, 21, 25, 33, 37, 44, 50). Our data suggest that regulation of the phosphorylation status of two critical residues of STAT3, S727 and Y705, is one of the mechanisms underlying the cross talk between Notch, GF, and CNTF signals. STAT3 phosphorylation on S727 was induced by ca-Notch1, GFs, or CNTF alone and thus was not directly associated with astrogenesis. In contrast, phosphorylation on Y705 was not induced by ca-Notch1 or GFs and was induced only weakly by CNTF. Yet, the combination of CNTF and ca-Notch1 markedly stimulated phosphorylation of STAT3 on both S727 and Y705 in parallel to their collaborative actions on astrocyte induction. These distinct phosphorylation conditions could explain how STAT3 participates in the regulation of both self-renewal and astrocyte differentiation. It could be that phosphorylation of STAT3 on both S727 and Y705 leads predominantly to astrogenesis, whereas STAT3 phosphorylated on S727, but not on Y705, contributes to the maintenance of NSCs. Further studies are necessary to reveal how such differentially phosphorylated STAT3 proteins regulate the fate of NSCs in distinct manners. It also remains to be investigated what molecules other than STAT3 are involved in the cross talk between Notch and other signaling pathways in NSCs.
We are grateful to A. Miyajima, R. Kageyama, T. Kitamura, T. Taga, T. Iwatsubo, H. Aburatani, S. Yamamoto, H. Kosako, and I. Dobashi for reagents and technical assistance. We also thank all members of our laboratories for support.
This work was supported in part by the Ohio Eminent Scholar Award of the State of Ohio and the Solution Oriented Research for Science and Technology Program, Japan Science and Technology Agency.
The authors of this study have no financial conflicts of interest that might be construed to influence the results or interpretation of the study.
Published ahead of print on 19 March 2007.