Neurogenesis in the Dentate Gyrus Depends on Ciliary Neurotrophic Factor and Signal Transducer and Activator of Transcription 3 Signaling§


  • Author contributions: S.M. and H.S.: collection and/or assembly of data, data analysis and interpretation; B.P.S.C.: collection and/or assembly of data; H.-D.H. and M.K.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript.

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

  • §

    First published online in STEM CELLSExpress November 20, 2008.


In the neurogenic areas of the adult rodent brain, neural stem cells (NSCs) proliferate and produce new neurons throughout the lifetime. This requires a permanent pool of NSCs, the size of which needs to be tightly controlled. The gp130-associated cytokines ciliary neurotrophic factor (CNTF) and leukemia inhibitory factor (LIF) have been implicated in regulating NSC self-renewal and differentiation during embryonic development and in the adult brain. To study the relevance of the two cytokines in vivo, we analyzed precursor cell proliferation and neurogenesis in the dentate gyrus of CNTF- and LIF-deficient mouse mutants. The number of radial glia-like NSCs, proliferative activity, and generation of new neurons were all reduced in CNTF−/− mutants but unaltered in LIF−/− animals. Conditional ablation of the signal transducer and activator of transcription 3 (STAT3) gene under the control of the human glial fibrillary acidic protein promoter resulted in a reduction of neurogenesis similar to that in CNTF−/− mice. The size of the granule cell layer was decreased in both mutants. Treatment of neurosphere cultures prepared from adult forebrain with CNTF inhibited overall proliferative activity but increased the number of NSCs as indicated by enhanced secondary neurosphere formation and upregulated expression of stem cell markers. Knockdown of STAT3 with short interfering RNA inhibited CNTF effects on neurospheres, and knockdown of suppressor of cytokine signaling 3 (SOCS3) enhanced them. Our results provide evidence that CNTF-induced STAT3 signaling is essential for the formation and/or maintenance of the neurogenic subgranular zone in the adult dentate gyrus and suggest that CNTF is required to keep the balance between NSC self-renewal and the generation of neuronal progenitors. STEM CELLS2008;27:431–441


Proliferation and self-renewal of neural stem cells (NSCs), their transition into restricted progenitor cells, and production of postmitotic neurons and glial cells must be tightly controlled. Although multiple signaling molecules and regulatory genes influencing proliferation, migration, and fate determination of neural precursors have been identified [1–4], the mechanisms regulating these processes in vivo are not clear.

The neurotrophic cytokines ciliary neurotrophic factor (CNTF) and leukemia inhibitory factor (LIF) have been implicated in the regulation of NSC self-renewal and differentiation [5]. Their effects are mediated by receptor complexes consisting of two common signal transducing components, gp130 and LIF receptor β (LIFRβ), and an additional ligand-binding subunit, ciliary neurotrophic factor receptor α (CNTFRα), in the case of CNTF [6, 7]. Ligand-binding activates janus kinase (JAK)/signal transducer and activator of transcription 3 (STAT3), Ras/mitogen-activated protein (MAP) kinase, and phosphoinositide 3-kinase/Akt signaling pathways [8–11]. In NSC cultures from embryonic rodent brain, CNTF (or LIF) was shown to enhance the generation of glial fibrillary acidic protein (GFAP)-positive cells, and in cytokine signaling mutants the number of GFAP-positive cells is reduced [12–18], suggesting that these cytokines promote the differentiation of neural precursor cells into astrocytes. Others reported that these cytokines increase the number of self-renewing cells in embryonic neurosphere cultures and that self-renewal and growth of the forebrain are reduced in LIFRβ- or gp130-deficient mice [19–22]. These observations led to the conclusion that gp130/LIFRβ signaling is important for maintaining the pool of neural stem cells. The apparent contradiction may be resolved by the observation that cultured NSCs, which express GFAP in response to LIF, remain multipotent and capable of self-renewal [12, 22]. Moreover, GFAP-positive radial glia in the embryonic central nervous system (CNS) and astrocyte-like cells in the adult brain were identified as pluripotent stem cell-like precursors in vivo [23].

In the adult brain, NSCs reside in two neurogenic areas, the subventricular zone (SVZ) of the lateral ventricle and the subgranular zone (SGZ) of the dentate gyrus (DG). Infusion of CNTF or overexpression of LIF increases NSC self-renewal [5, 19, 24] and inhibits or stimulates neurogenesis in the adult mouse forebrain [5, 25]. Since the two factors use shared signal-transducing receptor components (gp130, LIFRβ), these experiments did not provide evidence about the physiological relevance of the cytokines. In addition, there was no information about which of the alternative intracellular signaling pathways mediate CNTF or LIF actions in vivo, although previous in vitro studies had indicated that CNTF exerts its effects on NSCs mainly via the JAK/STAT3 signaling pathway [14, 26, 27]. To clarify the importance of the endogenous cytokines and the physiological relevance of JAK/STAT3 signaling, we have analyzed neurogenesis in the adult DG of LIF- and CNTF-deficient mice and in conditional STAT3-knockout mutants. We found that both the number of proliferating cells and neurons were reduced in CNTF- but not in LIF-knockout animals. Conditional ablation of STAT3 in NSCs produced a very similar reduction of neurogenesis. Together with results of vitro experiments, these results indicate that CNTF-induced STAT3 signaling is essential for achieving normal levels of adult neurogenesis in the dentate gyrus.



Breeding pairs of CNTF−/− and LIF+/− mice [28, 29] were obtained from Dr. Sendtner (University of Würzburg, Würzburg, Germany). They were kept on a C57BL/6-background by regular back-crossing with wild-type mice (Charles River Laboratories, Sulzfeld, Germany, STAT3loxP/loxP mice were provided by Drs. Takeda and Akira (Osaka University, Osaka, Japan [30]). To increase the efficiency of Cre-mediated deletion of the STAT3 gene and to prevent Cre-mediated interchromosomal recombination, a STAT3-null allele was introduced in addition to the floxed allele [31]. Mice expressing Cre recombinase under the control of the human GFAP promoter, a generous gift from Dr. Messing (Waisman Center, Madison, WI; [32]), were used to introduce Cre. GFAP-Cre/STAT3flox/−, GFAP-Cre/STAT3flox/+, and their respective controls not carrying the GFAP-Cre allele obtained by crossing STAT3loxP/loxP mice with GFAP-Cre/STAT3+/− mice were used in the experiments. Animals were kept under a 12-hour-light/12-hour-dark cycle with ad libitum access to food and water and were used at approximately 3 months of age. Animal experiments were in accordance with German federal law and were approved by the local animal care committee (Regierungspräsidium Freiburg, Freiburg, Germany).

Bromodeoxyuridine Labeling and Immunocytochemistry

To label proliferating cells, a saturating labeling protocol was used [33]. Briefly, mice were given four intraperitoneal (i.p.) injections of bromodeoxyuridine (BrdU, 100 mg/kg) at 4-hour intervals over the course of 12 hours. This labeling paradigm labels the majority of proliferating cells during one cell cycle, avoiding problems associated with relabeling cells randomly. After survival times of 1 or 4 weeks, mice were anesthetized with 100 mg/kg pentobarbital (i.p.) and fixed by transcardial perfusion with 0.9% saline in 10 mM phosphate buffer, pH 7.35, followed by 4% paraformaldehyde. Brains were postfixed for 24 hours, transferred to 20% sucrose, embedded in optimal cutting temperature compound (Leica Microsystems, Wetzlar, Germany,, cut frozen 30 μm in the coronal plane, and processed free-floating for immunodetection. Sectioning was such that eight parallel series, each consisting of sections spaced apart by 8 × 30 μm through the entire hippocampus, were collected into 8 wells of a 12-well plate. In coronal sections of the dorsal hippocampus, dimensions of the SGZ can be precisely defined, whereas this is not possible in the oblique or tangential sections through the ventral part of the hippocampus. Therefore, we restricted the quantitative analysis to the dorsal hippocampus [34, 35]. To ensure that representative cell counts were obtained by this procedure, series containing the entire hippocampus were stained with neurogenic differentiation 1 (neuroD), and the number of immunoreactive neurons was evaluated separately in the rostral, middle, and caudal thirds of the hippocampus comparing wild-type and CNTF-deficient mice (supporting information Fig. 1). Although there was no significant difference between the hippocampal subregions in either wild-type or CNTF−/− animals, the difference between the genotypes was highly significant. Sections from mutant and wild-type animals were processed in parallel for immunocytochemistry to ensure equal staining conditions. Staining protocol and antibodies used can be found in supporting information.

Quantification of Cell Numbers, Dentate Gyrus, and Cortex Measurements and Statistical Analysis

Immunostained cells in the DG were counted in all intact sections of a series containing the dorsal part of the hippocampus and expressed as cell density (cells per mm3) calculated from length and width of the SGZ (two-cell-thick region from the inner margin of the dentate granule cell layer) as determined with the help of ImageJ ( and section thickness [35, 36]. Values from individual animals were averaged and treated as an independent data point. Hoechst-stained sections were used for measuring the volume of the DG granule cell layer and cortical thickness. Granule cell area was measured in all sections of a series with the help of ImageJ; values were summed and multiplied by 8 (the inverse of the sampling fraction) and section thickness. Cortical thickness was measured in three or four sections located at identical rostro-caudal positions with the habenular commissure as a landmark. Statistical analysis was by two-tailed t tests or analysis of variance followed by post tests as appropriate using GraphPad Prism (GraphPad Software, Inc., San Diego,

Neurosphere Cultures

Neurosphere cultures were prepared from the SVZ of adult mice and maintained in NeurobasalTM A (Invitrogen, Karlsruhe, Germany,, 2% B27 supplement, 2 mM GlutaMaxTM 1 (Invitrogen), 1 mM L-glutamine, penicillin/streptomycin (both at 100 μg/ml), 20 ng/ml recombinant murine epidermal growth factor (EGF), and 20 ng/ml recombinant human basic fibroblast growth factor (bFGF) (medium and supplements from Invitrogen, growth factors from Peprotech [London,]; details are given in supporting information).

To measure proliferation and to count the number of neurospheres, dissociated cells were seeded at 500 or 1,000 cells per 100 μl in 96-well plates (Greiner Bio-One, Frickenhausen, Germany, After various times in culture, 10 μl of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) solution (Roche Applied Science, Mannheim, Germany was added, and cells were incubated for 4 hours at 37°C/5% CO2. The wells were then photographed. One hundred microliters of lysis solution (10% SDS, 0.01 M HCl) was added to each well, and incubation was continued overnight. Spectrophotometric measurement according to the manufacturer's instructions was used to quantify cell numbers. Micrographs of the MTT-stained neurospheres before lysis were analyzed with ImageJ. Because of the high contrast of MTT-stained spheres (e.g., Fig. 6B), image processing could be limited to a minimum and consisted of binarization and removal of single-pixel artifacts by median filtering (2 × 2 filter). Thresholded objects in 30–50 images could then be counted automatically. The diameter of all spheres was determined to obtain average sphere size. Quantitative reverse transcription-polymerase chain reaction, short interfering RNA (siRNA) treatment, immunoblotting, and behavioral testing are described in supporting information.


Neurogenesis Is Reduced in the Dentate Gyrus of CNTF−/− Mice

Generation of granule cells in the DG, one of the germinal zones in the adult mammalian brain [37, 38], originates from slowly dividing, multipotent radial glia-like cells in SGZ (B or type 1 cells) and progresses via proliferating progenitors (D or type 2 cells) to mature neurons. The expression of marker proteins defining different stages of this lineage has been established (reviewed in [37, 39]).

To investigate whether neurogenesis in the adult mouse hippocampus is altered in the absence of CNTF or LIF, we immunolabeled hippocampal sections of wild-type, CNTF−/−, and LIF−/− mice with antibodies against neuroD, doublecortin (DCX), and PSA-NCAM (Fig. 1). These markers are transiently expressed in progenitor cells and immature neurons but are downregulated in mature neurons [39]. The number of neuroD-positive cell nuclei located in the SGZ and innermost part of the granule cell layer was reduced in CNTF−/− animals by 45.9% compared with wild-type mice (Fig. 1A, 1B, 1G). A virtually identical reduction (46.1%) was observed for DCX-positive cell bodies in the SGZ (Fig. 1D, 1E, 1G). Labeling for polysialylated neural cell adhesion molecule (PSA-NCAM), another marker for D-type progenitors [40], confirmed that the density of these cells is significantly lower in CNTF-deficient mice (Fig. 1G). In contrast, the labeling pattern for neuroD (Fig. 1C) and DCX (Fig. 1F) in LIF−/− animals was indistinguishable from that in C57BL/6 mice (Fig. 1A, 1D) or LIF+/+ mice (not shown; Fig. 1H). These findings indicated that normal levels of neurogenesis in the adult hippocampus depend on the presence of CNTF, whereas LIF does not have a crucial role.

Figure 1.

Neurogenesis is reduced in the dentate gyrus (DG) of CNTF−/− but not of LIF−/− mouse mutants. (A–F): Coronal sections of the DG stained with antibodies against neuroD (AC) or DCX (DF) are shown. Numbers of neuroD- and DCX-immunoreactive cells in the subgranular zone of the DG were reduced in CNTF−/− animals (B, E) compared with WT animals (A, D). LIF−/− animals (C, F) did not show reduced cell numbers. (G): Quantification of cells stained with immunocytochemical markers for immature granule cells. Cell density in the DG (cells per mm3) of newly generated neurons labeled for neuroD, DCX, and PSA-NCAM was reduced in CNTF−/− animals (filled bars) by 45.9%, 46.1%, and 37.1%, respectively, compared with WT mice (open bars). Data are means ± SEM (n = 5–10 animals; six or seven sections analyzed per animal; statistical analysis was by two-way analysis of variance with Bonferroni post tests: degrees of freedom = 43, residual mean square = 1.162; significant difference between genotypes: F = 18.49, p < .0001; ∗, p < .05). (H): Quantitative evaluation of stained cells in LIF−/− (filled bars) and WT littermates (LIF+/+; open bars). Scale bar = 100 μm (A) (also valid for [B–F]). Abbreviations: CNTF, ciliary neurotrophic factor; DCX, doublecortin; LIF, leukemia inhibitory factor; neuroD, neurogenic differentiation 1, PSA-NCAM, polysialylated neural cell adhesion molecule; WT, wild-type.

Proliferation and the Stem Cell Pool Are Reduced in the Dentate Gyrus of CNTF−/− Mice

To determine whether the difference in the number of newborn granule cells results from a reduced number of proliferating progenitors in the DG of CNTF−/− mice, we stained actively dividing cells with antibodies against Ki67. The number of labeled cells was reduced by 50.1% in CNTF−/− animals (Fig. 2 A), that is, to approximately the same extent as seen with markers for newborn granule cells (described above). Analyzing BrdU-labeled cells 1 and 4 weeks after injection confirmed that proliferative activity is reduced in the DG of CNTF−/− mice. This experiment further revealed that the number of BrdU-labeled cells decreased at a similar rate in wild-type and CNTF−/− animals between these two time points (Fig. 2A). This decrease represents the combined effect of cell death and of label dilution due to continued division of the initially labeled cells. Our finding that the relative loss of BrdU-labeled cells was not significantly different between wild-type and CNTF−/− mice would argue that neither apoptosis nor production of postmitotic differentiating cells is affected by the absence of CNTF. In line with this, the relative numbers of newly generated (BrdU-labeled) immature DCX-positive neurons and mature granule cells expressing neuron-specific nuclear antigen and Prox1 (homeobox prospero-like protein 1) were very similar in wild-type and CNTF−/− animals (Fig. 2B). Four weeks after BrdU injection, two types of BrdU-labeled cells could be detected: (a) heavily labeled cells, lying at the hilar margin of the granule cell layer (Fig. 2C, arrows), and (b) cells showing a punctuate staining pattern (Fig. 2C, arrowheads and inset). The vast majority of the latter can be shown to express neuronal markers (e.g., Prox1; Fig. 2C), thus representing differentiated granule cells. The cells that remained strongly labeled 4 weeks after BrdU injection are likely to represent a slowly cycling label-retaining population and thus show stem cell-like properties. They do not express differentiation markers, and their number is reduced by 65% in CNTF−/− mice (data not shown). Bona fide stem cells in the DG coexpress nestin and GFAP and have a long, radially oriented process (Fig. 2D). They have been identified as the earliest neuronal progenitors in the adult DG [23, 41]. Their number was reduced by 35.8% in CNTF−/− mice (Fig. 2E–2G). Taken together, these results suggest that the reduction of the number of proliferating and differentiating cells we observed in CNTF−/− animals results from a reduced size of the stem cell pool rather than from an effect on differentiation of their progeny.

Figure 2.

The number of proliferating cells in the dentate gyrus of CNTF-knockout animals is reduced. (A): Determination of the density of proliferating cells by immunostaining for Ki67 showed a reduction by 50.1% in CNTF−/− animals (filled bars) compared with WT animals (open bars). Comparison of BrdU+ cells 1 and 4 wk after BrdU injection further shows that their number is correspondently reduced in knockout versus control animals between those two time points: 54.6% at 1 wk, 65.1% at 4 wk (∗, p < .05; ∗∗, p < .01; n = 7–10 animals for Ki67; n = 3 animals for each of the survival times following BrdU injection; Student's t test). (B): Quantification of cells coexpressing BrdU and markers for differentiated granule cells 1 and 4 wk after saturating labeling of proliferating cells. More than 70% of the BrdU-labeled cells coexpress DCX, a marker of newborn granule cells 1 wk after labeling. Four wk after BrdU injection more than 70% of BrdU-stained cells coexpressed the mature markers NeuN and Prox1. There is no significant difference between CNTF−/− and WT animals, suggesting that differentiation of newborn cells is not affected by the absence of CNTF. (C): BrdU (red)/Prox1 (green) double-stained section of the dentate gyrus 4 wk after BrdU injection illustrating the two principal BrdU-staining patterns: heavily labeled cells in the subgranular zone, stained only for BrdU (arrows), and cells displaying a punctate BrdU-staining pattern, which are also positive for Prox1 (arrowheads; additional examples shown at higher magnification in the inset). (D): High magnification of putative stem cells in the granular cell layer, cells double-labeled for glial fibrillary acidic protein (green) and nestin (red). (EG): Comparison of the number and density of nestin-stained cells in WT (E) and CNTF−/− animals (F) showed a significantly reduced density ([G], by 35.8%; p < .01; n = 7–9 animals; Student's t test) of nestin-positive cells with a radially oriented process in the granule cell layer. All data shown are means ± SEM. Scale bars = 100 μm (C, F) (also valid for [E]) and 50 μm ([D] and inset of [C]). Abbreviations: BrdU, bromodeoxyuridine; CNTF, ciliary neurotrophic factor; wk, week; DCX, doublecortin; NeuN, neuron-specific nuclear antigen; Prox1, homeobox prospero-like protein 1; WT, wild-type.

To investigate whether reduced neurogenesis in adult CNTF−/− mice is associated with neuroanatomical changes, we compared the neocortex and the DG of wild-type and knockout animals. CNTF-knockout animals do not show gross neuroanatomical alterations, and no behavioral differences were observed in open field or radial arm maze tests (supporting information Fig. 2). Cell density, layering, and thickness of the cortex are not distinguishable, as demonstrated in Hoechst-stained sections through the visual cortex (Fig. 3A, 3B). Quantitative analysis also did not reveal differences in cortical thickness (Fig. 3C). In contrast, when comparing the total volume of the granule cell layer, it was apparent that it was significantly smaller in CNTF−/− animals, by 21.6% (Fig. 3D–3F).

Figure 3.

Cortical cytoarchitecture is unaltered and size of the dentate gyrus (DG) granule cell layer is decreased in CNTF−/− mice. (A, B): Representative coronal sections of the visual cortex from WT and CNTF−/− animals stained with Hoechst 33258. No differences in cell density and layering were detectable. (C): Quantification of thickness of the neocortex in WT and CNTF−/− mice. (D, E): Comparison of the DG in wild-type and CNTF−/− mice. The coronal sections taken from identical rostro-caudal positions and stained with Hoechst showed a reduction of DG size in CNTF-deficient animals. (F): Quantification of the total volume (mm3) of the DG granule cell layer. The volume is significantly reduced (by 21.6%) in CNTF-deficient mice. Values in (C) and (F) represent means ± SEM (n = 3–4 animals; ∗∗, p < .01; Student's t test). Scale bars = 100 μm (A) (also valid for [B]) and 200 μm (D) (also valid for [E]). Abbreviations: CNTF, ciliary neurotrophic factor; WT, wild-type.

Conditional STAT3 Ablation Suppresses CNTF-Induced STAT3 Activation in Neural Stem Cells

Three main intracellular signal transduction pathways are activated by CNTF and other gp130-cytokines: the JAK/STAT3, mitogen-activated protein kinase (MAPK), and AKT pathways [10]. We used conditional STAT3 mutants to study whether the JAK/STAT3 pathway is involved in mediating the observed effects of CNTF on neurogenesis. For this we crossed Stat3loxP mice [31] with human glial fibrillary acidic protein (hGFAP)-Cre transgenic mice [32]. The hGFAP gene promoter directs Cre expression to all GFAP-expressing cells, including GFAP-positive neural stem/progenitor cells [33, 42]. In the adult DG of GFAP-Cre/STAT3+/+ animals, Cre recombinase can be shown to be expressed in astrocytes and GFAP+ cells with cell bodies in the SGZ and radially oriented processes in the granule cells layer (Fig. 4 A, 4B, arrows), identifying them as B or type 1 cells. In GFAP-Cre/STAT3flox/− mice, STAT3 mRNA levels in the hippocampus, as determined by quantitative polymerase chain reaction (qPCR), were decreased by 54% compared with STAT3loxp/− control animals, demonstrating Cre-mediated deletion of the STAT3 gene (Fig. 4C). To confirm that STAT3 protein synthesis and activation is abolished in NSCs of GFAP-Cre/STAT3flox/− mice, we prepared neurosphere cultures from the adult SVZ and analyzed them by immunoblotting. In cultures from the conditional knockout mutants, STAT3 protein levels were found to be very low (Fig. 4D). In addition, although cultures from control littermates (STAT3loxP/Cre) displayed robust tyrosine phosphorylation of STAT3 upon CNTF stimulation, no phospho-STAT3 could be detected in cultures of Cre+ mutant animals. Activation of the MAPK pathway by CNTF, as determined by analyzing phospho-MAPK in CNTF-treated cultures, was not affected in the STAT3 mutants (Fig. 4E), indicating a specific impairment of JAK/STAT3 signaling.

Figure 4.

Suppression of CNTF-induced STAT3 signaling in neural stem cells from conditional STAT3-knockout mice. (A, B): Cre recombinase expression in the dentate gyrus of adult human GFAP-Cre mice. Besides astrocytes, progenitor cells located in the subgranular zone (arrows [A]) with radially oriented GFAP-positive processes (green; arrows [B]) were strongly immunoreactive for Cre (red). (C): STAT3 mRNA levels were significantly reduced (54%) in GFAP-Cre/STAT3flox/− mouse mutants (Cre+; black bar) compared with STAT3loxP/− mice (Cre; open bars) used as controls (expression levels are expressed relative to GAPDH-mRNA expression; values are means ± SEM; Student's t test; ∗∗, p < .01; n = 3). (D, E): CNTF-stimulated (20 ng/ml) or untreated neurosphere cultures of GFAP-Cre/STAT3flox/− (Cre+) or STAT3flox/− (Cre; control) mice were immunoblotted for phosho-Stat3, Stat3 (to determine activated and total STAT3 levels, respectively), and GAPDH (D) or for pMAPK and total MAPK (E). In conditional STAT3 knockouts, STAT3 activation in neural stem cells was suppressed, whereas induction of phosphorylation of ERK1/2 by CNTF stimulation was not affected. Scale bars = 100 μm (A) and 50 μm (B). Abbreviations: CNTF, ciliary neurotrophic factor; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFAP, glial fibrillary acidic protein; MAPK, Ras/mitogen-activated protein kinase; pMAPK, phospho-MAPK; pSTAT3, phospho-STAT3, STAT3, signal transducer and activator of transcription 3.

Proliferation and Neurogenesis Are Reduced in the Dentate Gyrus of Conditional STAT3 Mice

Our results strongly suggest that CNTF is involved in controlling neurogenic activity in the SGZ. If the CNTF action is mediated by the JAK/STAT pathway, as suggested by previous in vitro studies, the phenotype of the mouse mutants with the STAT3 gene ablated in neural stem cells/progenitors is expected to resemble that in CNTF-deficient mice. We therefore examined cell proliferation and the expression of markers of the granule cell lineage in adult mutant mice and control littermates (Fig. 5). Figure 5A and 5D shows sections of the DG stained for Ki67. There was a clear reduction in the number of Ki67+ cells in Cre+ animals (Fig. 5D) compared with Cre animals (Fig. 5A). Quantification of cell counts for Ki67+ cells and for cells containing phospho-histone H3, another marker of proliferating cells [43], is shown in Figure 5G. The number of cells was reduced by 44.1% for Ki67 and by 29% for the histone. Also, the number of nestin-positive DG stem cells with radial processes (Fig. 5B, 5E) was significantly lower (47.3%; Fig. 5H) in the STAT3 mutants. The progeny of these cells, namely PSA-NCAM-positive progenitors, as well as DCX-positive (Fig. 5C, 5F) and neuroD-positive immature neurons, were also reduced in number, by 46.2%, 43.3%, and 51.7%, respectively (Fig. 5H). Thus, neurogenesis in the STAT3 mutants is impaired as in CNTF−/− mice, indicating that the influence of CNTF on neurogenesis is mediated by the JAK/STAT3 pathway. To determine the potential consequences of reduced neurogenesis in the STAT3 mutants, we measured the volume of the DG granule cell layer in Hoechst-stained sections. As in the CNTF−/− mice, the size of the DG was significantly reduced by 23.2% in GFAP-Cre/STAT3flox− mice compared with control animals (Fig. 5I–5K).

Figure 5.

Impaired proliferation and neurogenesis in the dentate gyrus (DG) of conditional signal transducer and activator of transcription 3 (STAT3)-knockout mice. (AF): Comparison of hippocampal sections from STAT3loxP/− (Cre; control) and glial fibrillary acidic protein (GFAP)-Cre/STAT3flox/− (Cre+) mice immunostained for Ki67, nestin, and PSA-NCAM. Expression of these markers was reduced in conditional STAT3 knockouts. (G, H): Quantification of cells immunolabeled with proliferation markers Ki67 and PH3 ([G]; n = 4–5; statistical analysis was by two-way analysis of variance [ANOVA] with Bonferroni post tests: degrees of freedom (df) = 16, residual mean square = 0.004091; significant difference between genotypes: F = 42.13; p < .0001) or cell lineage markers nestin, neuroD, PSA-NCAM, and DCX ([H]; statistical analysis was by two-way ANOVA with Bonferroni post tests: df = 28, residual mean square = 0.7170; significant difference between genotypes: F = 57.64; p < .0001). All markers were significantly reduced in conditional STAT3 knockouts (black bars) compared with STAT3loxP/− used as controls (∗, p < .05; ∗∗, p < .01; ∗∗∗, p < .001). (I, J): Hoechst-stained coronal sections showing the reduction of DG size in GFAP-Cre/STAT3flox/− (Cre+) compared with STAT3loxP/− (Cre) animals. (K): Quantification of the DG granule cell volume. Granule cell layer volume was significantly reduced (by 23.2%) in conditional STAT3 knockouts (black bars; n = 3–4; ∗∗, p < .01; Student's t test). Scale bars = 100 μm (A) (also valid for [BF]) and (I) (also valid for [J]); scale bar of insert in B = 50 μm (also valid for insert in [E]). Abbreviation: DCX, doublecortin; neuroD, neurogenic differentiation 1; PH3, phospho-histone H3; PSA-NCAM, polysialylated neural cell adhesion molecule.

CNTF Inhibits Proliferation but Stimulates Secondary Sphere Formation and Upregulation of Candidate Stem Cell Markers in Neurosphere Cultures

Our in vivo experiments have shown that impaired CNTF signaling, by elimination of the ligand or by abrogation of intracellular signaling, leads to reduced generation of new neurons in the adult hippocampus, obviously caused by a reduction of the size of the SGZ stem cell pool. This indicates that CNTF can stimulate the survival or self-renewal of neural stem cells. Studies performed in neurosphere cultures from embryonic forebrains have come to the same conclusion [12, 19, 21, 22]. They showed that CNTF promotes self-renewal of NSCs by demonstrating increased secondary sphere formation in CNTF-treated cultures. Whether CNTF has the same effect in cultures of the adult brain has not been examined so far. To investigate this, we used neurosphere cultures from the adult SVZ. Because of the very low abundance of precursor cells in the adult hippocampus, neurosphere cultures containing sufficient numbers of self-renewing stem cells are difficult to prepare in a reproducible way from this tissue and, as a consequence, results on the cellular identity and the characteristics of these cultures are still controversial [44–46]. On the other hand, precursor cells of the adult SVZ and SGZ have been shown to respond similarly to exogenous CNTF in vivo [25].

Control cultures (EGF/bFGF) showed a 20–40-fold increase in cell number during a 6-day culture period (not shown). In the presence of CNTF (20 ng/ml), however, proliferation was strongly inhibited. Quantification by measuring BrdU incorporation (data not shown) or the MTT assay (Fig. 6 A) showed that the number of cells generated during this period was reduced by 43% compared with control cultures. Morphological analysis revealed that the number of spheres generated was not significantly changed by CNTF treatment, whereas their size was reduced by 54.4% (Fig. 6B–6D; average size of EGF/bFGF (EF)-treated spheres, 2,065 ± 60.54 μm2; average size of EGF/bFGF/CNTF (EFC)-treated spheres, 942.9 ± 67.46 μm2). We then performed a neurosphere assay, in which primary spheres were dissociated and reseeded under control conditions (EGF/bFGF) to determine the number of cells capable of forming secondary spheres. In this assay, neurosphere cultures pretreated with CNTF generated 54% more secondary neurospheres compared with EGF/bFGF-treated controls (Fig. 6E), and the total cell number was increased by 57.8% (Fig. 6F). Neurospheres from CNTF-treated cultures could be further passaged, as could control cultures, and could be shown to be multipotent under differentiation conditions, indicating that CNTF promotes the self-renewal or maintenance of NSC-like cells. To clarify whether CNTF exerts its effects via JAK/STAT3 signaling, we performed knockdown experiments using siRNAs for STAT3 and for suppressor of cytokine signaling 3 (SOCS3), the later being a prototypical STAT3-target gene, which is rapidly induced by CNTF treatment and acts as a negative regulator of the JAK/STAT3 pathway. When neurospheres cultures were treated with CNTF either in the presence of a control siRNA or two different STAT3 siRNAs, the specific siRNAs were found to significantly block the effect of CNTF, by 42% and 70%, respectively (Fig. 6G). Conversely, interfering with SOCS3 expression to disinhibit the JAK/STAT signaling pathway in cultures treated with subthreshold concentrations of CNTF induced a CNTF-like effect on NSC proliferation (Fig. 6H).

Figure 6.

CNTF inhibits proliferation but promotes self-renewal of neurosphere-forming cells in neural stem cell cultures from adult forebrain. (A): Proliferation in neurosphere cultures from adult forebrain was inhibited by CNTF. Cultures were grown for 6 days in the presence of epidermal growth factor (E) and basic fibroblast growth factor (F; 20 ng/ml) with (EFC) or without (EF) CNTF (20 ng/ml), and proliferation was determined by the (MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. (B, C): Microphotographs of neurosphere cultures grown as described in (A) with (C) and without (B) CNTF, demonstrating reduced sphere size with CNTF treatment. (D): Quantification of sphere number (left y-axis) and size (right y-axis) in control (EF) and CNTF-treated (EFC) cultures. CNTF did not reduce the number of spheres but inhibited sphere growth. (E, F): CNTF enhanced the formation of secondary neurospheres. Cultures were grown as described in (A) (EF or EFC), and spheres were dissociated, reseeded, and grown for an additional 6 days in the absence of CNTF (→EF). CNTF-treated cultures produced more secondary spheres (E) and, correspondingly, more cells (F). (G): STAT3 knockdown with two different siRNAs efficiently blocked the growth inhibitory effect of CNTF. Values are means ± SEM, expressed as percentage of CNTF effect of cells treated with 5 nM control siRNA (n = 3; details are given in supporting information; ∗, p < .05; ∗∗, p < .01; one-way analysis of variance [ANOVA] with Dunnett's post test to compare treatment values with control values). (H): Knocking down SOCS3 enhanced the effect of a subthreshold concentration of CNTF. Cells were either grown in the presence of 5 nM control siRNA and 1 ng/ml CNTF (open bar) (a concentration that does not exert a significant growth inhibitory effect), with 5 nM SOCS3 siRNA (filled bar), or with 5 nM SOCS3 siRNA + 1 ng/ml CNTF. Culture in SOCS3 siRNA alone led to a small but insignificant growth inhibition. However, significantly reduced growth was observed, when 1 ng/ml CNTF was added to the SOCS3 siRNA-treated cultures (values are means ± SEM, expressed as % of average sphere size as determined in cultures grown in EF + 5 nM control siRNA; n = 3; ∗, p < .05; one-way ANOVA with Dunnett's post test to compare treatment values with control values). (I): Comparison of gene expression in control (EF) and CNTF-treated (EFC) neurospheres after 4 days in culture. mRNA levels were measured by real-time reverse transcription-polymerase chain reaction. Values represent fold increases in CNTF-treated cultures. Abbreviations: CNTF, ciliary neurotrophic factor; siRNA, short interfering RNA; STAT3, signal transducer and activator of transcription 3, SOCS3, suppressor of cytokine signaling 3.

An increase in the number of stem cells in EFC- versus EF-treated neurosphere cultures can also be inferred from qPCR analysis for marker genes that are commonly expressed by self-renewing stem cells from a variety of sources (Fig. 6I). This showed that CNTF treatment for 4 days leads to increased expression of markers for radial glial cells (GLAST) and self-renewing stem cells (NANOG, KLF4, FBX15, and c-MYC). Moreover, strongly upregulated expression of CNTF receptor components (CNTFRα, LIFRβ, and gp130) and increased STAT3 signaling, as indicated by the massive increase of SOCS3 mRNA levels, were observed. These results indicate that CNTF promotes the self-renewal of neural stem cells and prevents the generation of fast proliferating progenitors from these cells.


The SGZ of the adult DG contains a population of radial glia-like cells that express nestin and GFAP, exhibit proliferative activity, and are capable of self-renewal [40, 47, 48]. Although their identity as true NSCs is disputed [23, 49], these cells have been demonstrated to act as primary precursors of the granule cell lineage [40]. They give rise to more rapidly dividing progenitors that express markers of immature neurons (PSA-NCAM and DCX) and then exit the cell cycle to differentiate into granule cells [40, 48]. We found that the numbers of primary precursors, proliferatively active cells, and newly generated neurons in the dentate gyrus are all reduced in the DG of CNTF-deficient mouse mutants. Exogenous CNTF (or LIF) promotes neural stem cell renewal in cultures derived from the embryonic and adult SVZ ([19, 21, 22, 24], this study) and neurogenesis in the adult SGZ and SVZ [25, 50]. Together with our in vivo data, these results suggest that the lack of CNTF leads to a reduced generation or depletion of primary precursors in the DG and reduced neurogenesis. Since corresponding reductions were observed in mutants with STAT3 expression deleted in NSCs, we conclude that CNTF/STAT3 signaling, by regulating the size of the hippocampal stem cell pool, is essential for maintaining normal levels of neurogenesis in the adult DG.

Since CNTF-responsive cells expressing the functional CNTF receptor complex (gp130/LIFRβ/CNTFRα) are necessarily also responsive to LIF, which signals via the heterodimeric gp130/LIFRβ complex, previous studies that have investigated the effects of exogenous LIF or CNTF on NSC self-renewal in neurosphere cultures and in vivo could not provide conclusive information about the physiological relevance of the two cytokines. In fact, LIF has the same effects as CNTF in adult neurosphere cultures (data not shown). Our in vivo experiments now demonstrate that CNTF is an important regulatory signal of neurogenesis in the DG, whereas the absence of a phenotype in the DG of LIF−/− animals indicates that this cytokine is not involved. Recently Yang et al. [50] reported that dopamine-induced neurogenesis in both the SVZ and SGZ of adult mice is mediated by CNTF and that CNTF-knockout animals display reduced numbers of BrdU-incorporating cells (by 20%–24%) in the SVZ and SGZ, corroborating our findings of the role of endogenous CNTF in regulating neurogenesis. Interestingly, LIF, and not CNTF, has been shown to regulate olfactory sensory neuron renewal, in particular under regenerative conditions [51, 52].

In neural cells, CNTF-like cytokines can activate different intracellular signaling cascades, the JAK/STAT1 or STAT3 pathways and the Ras/MAP kinase pathway (reviewed in [9]). Our results indicate that the influence of CNTF on hippocampal proliferation and neurogenesis is specifically mediated by STAT3 signaling, since (a) they are reduced to the same extend in CNTF−/− mice and STAT3 mutants, (b) CNTF-induced phosphorylation of MAP kinase (Fig. 4) and STAT1 (not shown) was preserved in neural precursors of the STAT3 mutants, and (c) knockdown of STAT3 in NSC cultures inhibited CNTF effects, whereas reduction of SOCS3 expression potentiated CNTF action. This is in agreement with results from in vitro studies demonstrating that CNTF effects on embryonic neural precursors (interpreted in these reports as promoting gliogenesis) are STAT3-dependent [14, 26]. In addition to CNTF, effects of other, noncytokine growth factors that have been reported to activate STAT3 signaling could have been affected after conditional ablation of STAT3. However, the observation that the reduction of neurogenesis in conditional STAT3-knockout mice was very similar to that in CNTF−/− animals suggests that both effects are due to impaired JAK/STAT signaling via the gp130/LIFRβ cytokine receptor.

The receptor component CNTFRα is highly expressed in neuroepithelial precursors [53–55], and mouse embryos lacking gp130 or LIFRβ have been shown to exhibit impaired ventricular zone precursor cell proliferation associated with reduced growth of the cerebral wall [20, 21], indicating that cytokine receptor-mediated signals are involved in regulating embryonic forebrain development. In the adult SVZ, the number of proliferating precursors and of neurosphere forming cells was shown to be reduced in heterozygous LIFRβ mutants [19]. In the present study, the adult neocortex of the CNTF-deficient animals appeared to be normal with respect to cell layering and size. Since the mouse cortex, in contrast to the DG, forms prenatally, this suggests that CNTF is not an active ligand during embryonic neurogenesis. In line with this, CNTF expression has not been observed in the embryonic brain [56]. Thus, the role of CNTF in controlling the NSC pool is obviously restricted to postnatal neurogenesis. Other CNTF-related cytokines expressed prenatally, such as neuropoietin and cardiotrophin-1 [55–57], may have similar functions in embryonic neurogenesis.

Both mutations, the CNTF knockout and the conditional STAT3 knockout, are effective already during early postnatal stages, when proliferating precursor cells immigrate into the hilar region of the dentate gyrus anlage and then rearrange to form the SGZ, which persists as a germinative zone into adulthood [1]. Thus, the decreased neurogenic activity observed in the mutants might be caused by impaired generation of the stem cell pool in the developing SGZ or by reduced self-renewal in the mature stage. Several lines of evidence support the conclusion that CNTF is of importance in the adult brain. First, CNTFRα is expressed in the adult granule cell layer and the SGZ in particular [25, 58, 59]. Second, neural precursor cells from the adult brain are obviously responsive to CNTF as shown in vitro and in vivo, in both the SGZ and SVZ ([19, 24, 25], this study). Third, injection of anti-CNTF antibodies into the adult brain interferes with neurogenesis [25]. However, a thorough developmental study or the use of other transgenic models is required to segregate potential developmental influences of CNTF and regulatory actions on NSC renewal during adulthood.

To study the effects of CNTF in vitro, we used well-established neurosphere cultures prepared from adult SVZ. Cultures from the SGZ have also been shown to contain multipotent precursor cells capable of self-renewal [45, 60], but the stem cell nature of these in vitro precursor cells has been challenged [43, 44]. In vivo, stem/primary precursor cells (called B cells) in the SVZ and SGZ are very similar: they express GFAP and nestin, proliferate slowly, and can produce all other cell types of the SVZ and SGZ cell lineage [23]. In addition, in vivo responses of proliferating SGZ and SVZ progenitors to exogenous CNTF are the same [25, 50]. Therefore, it seems to be justified to regard results from SVZ neurospheres as being representative for both germinative areas. In agreement with studies in embryonic cultures [19, 21, 22], we find that CNTF increases the number of cells capable to form secondary neurospheres. The increased expression of regulatory genes characteristic of stem cells provides additional evidence that CNTF promotes self-renewing precursor cells in these cultures. The simultaneous pronounced inhibition of cell proliferation occurring in adult cultures is in contrast to effects seen with embryonic neurospheres or after LIF injection into embryonic brain [20, 21]. On the other hand, in vitro results shown here closely agree with in vivo observations in adult mouse brain [24]. There, overexpression of LIF by viral infection was shown to inhibit proliferation by impairing the amplification of rapidly proliferating (C-type) cells, which represent the majority of dividing cells in both the SVZ and adult neurospheres [23], but to expand the pool of slowly proliferating NSC-like (B-type) cells. Thus, our results are compatible with the conclusion drawn from these experiments that CNTF promotes self-renewal of NSCs in the adult brain by preventing their progression along the neural cell lineage. The absence of such an inhibitory effect in embryonic neurospheres may be related to differences in precursor cell populations present in adult and embryonic brain [3].

In apparent contrast to our observations in vitro and to results obtained after LIF overexpression in vivo [24], CNTF injection into adult mouse brain stimulates proliferation and neurogenesis in the DG and SVZ [25] and hypothalamus [57]. However, this discrepancy may be explained by the experimental protocols used. In these studies, CNTF was applied transiently by injection or infusion into the brain, and the number of proliferating cells was determined after more than 1 week. This treatment regime corresponds to our neurosphere assay, where pretreatment with CNTF resulted in enhanced secondary sphere formation after CNTF was removed (Fig. 6E, 6F), whereas with sustained overexpression of LIF in vivo [24] or with CNTF continuously present in neurosphere cultures (Fig. 6A, 6D), proliferation was reduced. Together, these results are in line with our findings in the mouse mutants, since they lead to the conclusion that in the continuous absence of CNTF signaling, the NSC pool and neurogenesis would be reduced as a consequence of impaired self-renewal.


Our results show that CNTF, signaling via the JAK/STAT3 pathway, plays a crucial role in regulating the extent of granule cell generation from precursor cells of the SGZ. In the unlesioned brain CNTF is expressed at low levels by astrocytes, and its expression is strongly increased following lesion. As astrocytes form a major constituent of the stem cell niche, they are in a prime position to regulate the proliferation of neural stem cells and their progeny. Increased neurogenesis, as has been observed in various CNS-lesion models, could be mediated by a transient increase of CNTF expression by astrocytes of the niche, resulting in increased numbers of stem cells. Our results suggest that CNTF, related cytokines, and STAT3 signaling are rational targets in the context of manipulating the endogenous pool of neural stem cells in the lesioned or degenerating CNS.


We thank Gabriele Kaiser and Birgit Egle for excellent technical assistance. This study was supported by the Deutsche Forschungsgemeinschaft (SFB 505/B9). We thank Dr. Sendtner (University of Würzburg, Würzburg, Germany), Drs. Takeda and Akira (Osaka University, Osaka, Japan), and Dr. Messing (Waisman Center, Madison, WI) for providing the mouse mutants used in this study.


The authors indicate no potential conflicts of interest.