• Open Access

FGF2 and Insulin Signaling Converge to Regulate Cyclin D Expression in Multipotent Neural Stem Cells


  • Adedamola Adepoju,

    1. National Institute for Neurological Disorders and Stroke, Maryland, USA
    2. University of Massachusetts School of Medicine, Massachusetts, USA
    Search for more papers by this author
  • Nicola Micali,

    1. National Institute for Neurological Disorders and Stroke, Maryland, USA
    2. Lieber Institute for Brain Development, Baltimore, Maryland, USA
    Search for more papers by this author
  • Kazuya Ogawa,

    1. National Institute for Neurological Disorders and Stroke, Maryland, USA
    2. Lieber Institute for Brain Development, Baltimore, Maryland, USA
    Search for more papers by this author
  • Daniel J. Hoeppner,

    1. National Institute for Neurological Disorders and Stroke, Maryland, USA
    2. Lieber Institute for Brain Development, Baltimore, Maryland, USA
    Search for more papers by this author
  • Ronald D.G. McKay

    Corresponding author
    1. National Institute for Neurological Disorders and Stroke, Maryland, USA
    2. Lieber Institute for Brain Development, Baltimore, Maryland, USA
    • Correspondence: Ronald McKay, Ph.D., Lieber Institute for Brain Development, 855 N Wolfe Street, Baltimore, Maryland 21205, USA. Telephone: 410-955-1000; Fax: 410-955-1044; e-mail: ronald.mckay@libd.org; or Daniel Hoeppner, Ph.D., Lieber Institute for Brain Development, 855 N Wolfe Street, Baltimore, Maryland 21205, USA. Telephone: 410–955-1000; Fax: 410–955-1044; e-mail: daniel.hoeppner@libd.org

    Search for more papers by this author


The ex vivo expansion of stem cells is making major contribution to biomedical research. The multipotent nature of neural precursors acutely isolated from the developing central nervous system has been established in a series of studies. Understanding the mechanisms regulating cell expansion in tissue culture would support their expanded use either in cell therapies or to define disease mechanisms. Basic fibroblast growth factor (FGF2) and insulin, ligands for tyrosine kinase receptors, are sufficient to sustain neural stem cells (NSCs) in culture. Interestingly, real-time imaging shows that these cells become multipotent every time they are passaged. Here, we analyze the role of FGF2 and insulin in the brief period when multipotent cells are present. FGF2 signaling results in the phosphorylation of Erk1/2, and activation of c-Fos and c-Jun that lead to elevated cyclin D mRNA levels. Insulin signals through the PI3k/Akt pathway to regulate cyclins at the post-transcriptional level. This precise Boolean regulation extends our understanding of the proliferation of multipotent NSCs and provides a basis for further analysis of proliferation control in the cell states defined by real-time mapping of the cell lineages that form the central nervous system. Stem Cells 2014;32:770–778


Neural stem cells (NSCs) are multipotent cells that differentiate into neurons, astrocytes, and oligodendrocytes. The first report that proliferating precursors could generate neurons in the laboratory was followed by the demonstration using clonal expansion that these precursors were also capable of generating astrocytes and oligodendrocytes [1, 2]. These results were obtained by growing the cells in fully defined culture conditions where the only soluble proteins added were insulin and basic fibroblast growth factor (FGF2). In this paradigm, NSCs required both growth factor as neither alone was sufficient to promote survival and proliferation [3]. Insulin was part of the N2 supplement that was combined with basal growth medium to enable the first serum-free cell culture of any mammalian cell [4]. FGF2 was initially identified as a growth factor capable of stimulating the proliferation of 3T3 fibroblasts [5] and was later shown to stimulate the in vitro proliferation of many other cells types including cells from the developing nervous system [2, 6-14]. The expression pattern of FGF2 correlates with NSC expansion during embryonic brain development and genetic deletion of FGF2 decreased the numbers of neural precursors in the brain of developing mouse embryos [15]. A role in tissue development and homeostasis was further emphasized by infusion of ligands for tyrosine kinase receptors into the brain [15-17].

Interest in NSCs has been spurred by the prospect of using them to understand and treat neurological disease. Expansion outside the body is a central feature of stem cell-based technologies. To harness the potential of NSCs in medical research, it is important to understand the signaling that regulates survival, proliferation, differentiation, and quiescence. Tyrosine kinase signaling from the ligands insulin and FGF2 is sufficient for NSC expansion in vitro [1, 2]. Insulin and FGF2 interact with different cell surface receptors. FGF2 binds FGF receptor 1, 2, and 3 (FGFR1, 2, and 3) [11]. Insulin and related protein insulin-like growth factors (IGF-1 and IGF-2) bind the insulin and IGF-1 receptors (InsR and IGF-1R) [18, 19]. These cell surface receptors are receptor tyrosine kinases (RTKs) that activate similar signaling cascades [20]. Using in vitro cultures of oligodendrocyte progenitor cells, FGF2 and IGF-1 were shown to synergistically promote elevated cyclin D1 levels through increased transcription and translation, respectively [21, 22]. Adult rat neurospheres cultured with FGF2 and IGF-1 have elevated cyclin D1 protein levels [23]. FGF2 and its receptors have been shown to regulate NSC numbers in vivo during brain development [24, 25]. IGF2/insulin signaling also has a major developmental role [26] and IGF2 secreted by the choroid plexus regulates neurogenic precursors of the developing cerebral cortex [27].

The midgestation cerebral cortex is a large epithelial field containing precursor cells that generate neurons in vitro and in vivo [1, 28, 29]. Because this dorsal telencephalic structure contains large numbers of cells, it is widely used to define the molecular and cellular mechanisms that regulate brain development. The multipotent nature of cells in the developing cortex has been established first using clonal analysis and then with the more precise assessment afforded by real-time imaging [2, 30, 31]. In principle, multipotent cells may generate specific progeny by either positive or negative selection. The precision of real-time imaging established that external signals positively influence fate choice by tripotent stem cells that are only present for a restricted period immediately following passage [30]. Lineage imaging provides two insights that are important to the study reported here. These data show that multipotency is regenerated when the cells are passaged and that sister NSCs reach metaphase within minutes of each other for several generations [30]. Here, we apply high-content screening methods to demonstrate that proliferation is achieved through the synergistic action of FGF2 and insulin on the mRNA and protein levels of cell cycle regulators in multipotent NSCs.

Materials and Methods

Cell Culture

Embryonic day 13.5 embryos were isolated from pregnant C57BL/6 mice according to the National Institute for Neurological Disorders and Stroke (NINDS) and Lieber Institute for Brain Development (LIBD) guidelines, and cerebral tissues were collected as previously described in Johe et al. [2]. To prepare cells for all studies, tissues were separated into single cells by trituration and the cells were plated at 2 × 106 cells per 10-cm culture plates (Falcon 35-3003; http://www.thermofisher.com, Waltham, MA) coated with poly(l-ornithine) (Sigma P3655; Sigma-Aldrich, St. Louis, MO, USA; http://www.sigmaaldrich.com) and fibronectin (R&D Systems 1030FN; http://www.rndsystems.com, Minneapolis, MN). Cells were incubated at 5% O2, 5% CO2, and 90% N2 for 5 days in Dulbecco's modified Eagle's medium (DMEM)/F-12 medium (Mediatech 10-090-CV; http://www.cellgro.com, Herndon, VA) with N2 supplement containing 100 µg/ml apotransferrin (Sigma T1147-54), 20 nM progesterone (Sigma P8783), 100 µM putrescine (Sigma P5780), 30 nM sodium selenite (Sigma S5261), and 25 µg/ml insulin (Sigma I6634-1G). FGF2 (10 ng/ml) (human, R&D Systems 4114) was added daily and the medium was changed every other day until cells became 80% confluent, normally on the fifth day. Cells were lifted with Hanks' balanced saline solution (HBSS) (passage 1) and frozen in 10% Dimethyl Sulfoxide then stored at −80°C for future use. Thawed samples (passage 1) were expanded to 80% confluence, normally 3–4 days then passaged (passage 2). For all experiments, cells were then plated in media containing no FGF2 and no insulin for 3 hours. If inhibitor was used, it was added 2 hours after plating. We define time t = 0 as 3 hours after plating without FGF2 or insulin. For t = 0 readings, media were aspirated then samples were immediately harvested with lysis buffer because media exchange alone causes transient (3 minutes) Erk1/2 activation (data not shown). For all later time points, media are exchanged at t = 0 and replaced with experimental combinations. For experiments requiring FGF2 alone, insulin was excluded from the N2 supplement. Cells were passaged with HBSS containing HEPES (Sigma H-3784) pH 7.2 and either frozen at −80°C for storage or plated on six-well plates for experiments.

FGF2/Insulin Dose-Response

Cells (5 × 103) were seeded in each well of a 96-well high-content imaging plate (Perkin Elmer 6005558; http://www.perkinelmer.com, Waltham, MA) using DMEM/F-12 media as above but without FGF2 or insulin. After 3 hours of FGF2 and insulin starvation, media containing increasing doses of FGF2 and insulin were replaced in every well. FGF2 was added daily while medium with varying concentration of insulin was replaced every second day. After 96 hours, cells were fixed with 4% paraformaldehyde for 15 minutes then permeabilized with 0.1% Triton X-100 with 3% bovine serum albumin for 20 minutes. 4′,6-diamidino-2-phenylindole (Sigma D8417) was used as a nuclear counterstain. Images were acquired with the Operetta automatic microscope (Perkin Elmer) and image analysis was performed using custom building blocks with Columbus software (Perkin Elmer). Plotted data come from counting all cells from 15 nonoverlapping images per well.

EdU Incorporation

For six-well plates, 2 × 105 cells were plated in each well. For 96-well plates, 2.5 × 104 cells were plated per well and treated with FGF2, insulin, and FGF2 plus insulin. Cells were treated and processed according to Click-IT EdU Alexa fluor 488 Imaging kit (Invitrogen C10337; www.lifetechnologies.com, Carlsbad, CA). After 11 hours of incubation in 5% O2, 5% CO2, and 90% N2, the cells were treated for 60 minutes with 10 µM 5-ethynyl-2′-deoxyuridine (EdU). The media were aspirated and the cells were fixed with 4% paraformaldehyde for 15 minutes. To identify the S-phase population, cells were permeabilized with 0.5% Triton X-100 (Sigma T9284) and processed according to the manufacturer's protocol using Hoechst 33342 as a nuclear counterstain and EdU detection by Alexa 488 fluorescence. Fluorescent signals were captured using unbiased replicate acquisition patterns on six-well plates (with a Zeiss AxioObserver with a scanning stage; www.zeiss.com) or 96-well plates (with the Operetta). Data were analyzed using Columbus software and confirmed by manual spot check. EdU+ values reflect the percentage of nuclei showing Alexa 488 signal greater than two standard deviations above background.

Protein Preparation and Western Blotting

Cells (1 × 106)/well in six-well plates were plated in medium with no insulin or FGF2 for 3 hours. The medium was gently aspirated with pipette and replaced with medium containing FGF2 and insulin as indicated in the figure legends. For pharmacological inhibition, SU5402 (Tocris 3300; www.tocris.com. Minneapolis, MN), U0126 (EMD Chemicals 662005, www.emdmillipore.com/), and LY294002 (EMD Chemicals 440202) were added 1 hour prior to media change. At the end of the experiment, cells were lysed with lysis buffer containing 10% glycerol, 2% sodium dodecyl sulfate, 125 mM Tris-HCl pH 7.2, and 0.02% bromphenol blue. Total protein was determined with bicinchoninic acid (BCA) protein assay (Thermo Scientific 23221; www.thermofisher.com, Waltham, MA) and Western blot was performed with the following antibodies from Cell Signaling (http://www.cellsignal.com, Danvers, MA): anti-phospho Akt (S473 cat no. D9E), anti-phospho Akt (T308 cat no. C31E5E), anti-Akt, anti-phospho-ERK1/2, anti-cyclin D2, anti-c-fos, anti-c-jun, anti-phos-c-Jun (S63 and S73), and anti-phospho Rb (S780) (Cell Signaling); anti-cyclin D1 (Millipore; http://www.millipore.com, Billerica, MA); anti-phospho Rb (S780) and anti-phospho Rb (S795) (Abcam; http://www.abcam.com, Cambridge, MA).


Total RNA was extracted using TRIzol reagent (Invitrogen). cDNA was synthesized with Oligo dT primers and reverse transcribed with superscript III (Invitrogen) according to the manufacturer's protocol. Real-time PCR was performed on Applied Biosystems 7900HT Fast Real-Time System using TaqMan gene expression assays (Applied Biosystems). Primers for cyclin D1 (Mm-00432359), cyclin D2 (Mm-00438072), and cyclin D3 (Mm-01612362) were purchased from Applied Biosystems. Values plotted are the changes relative to control (ΔΔCt), which determines the fold increase in mRNA levels relative to the internal control (mGAPDH).

Flow Cytometry and Cell Cycle Analysis

Cells were plated at 3 × 106 cells per 10-cm tissue culture plate and treated with growth factors as indicated in the figure legends. After 24 hours, the adherent cells were passaged into suspension and fixed with 70% ethanol at −20°C overnight. The following day, the ethanol was removed and the cells were suspended in 500 µl PBS containing 1 µg/ml propidium iodide (sigma), and 1:50 RNase (Promega; http://www.promega.com, Madison, WI). Flow cytometry was performed with FACScaliber system and data were analyzed with the software Flowjo (http://www.flowjo.com, Ashland, OR).

Gene Silencing by siRNA

Cells (5 × 106) were transfected with 1 µM siRNA using the Mouse NSC Nucleofector kit from Lonza (http://www.lonza.com, Basel Switzerland). The manufacturer's protocol was followed accordingly. For cyclin D gene silencing, the surviving cells were plated on 60-mm plates after transfection in N2 medium containing 10 ng/ml FGF2. Forty-eight hours post-transfection or when the cells reach 90% confluence, the cells were passaged and plated at 1 × 106 cells per well on six-well plates for the time course experiment as indicated in the figure legends. siRNAs for cyclin D1 (S100943642 and S100943649), cyclin D2 (S100943670 and S102736069), and cyclin D3 (S100943691 and S100943705) were purchased from Qiagen. For each siRNA pool, 3 × 106 cells transfected with siRNA were then plated onto two six-well plates in N2 media containing 10 ng/ml FGF2. Forty-eight hours later, all cells were grown in N2 medium (-Insulin –FGF2) for 12-hour. One six-well plate was later stimulated with FGF2 and insulin for 2 hour to examine c-Fos expression level, and the other was stimulated for 12 hours to examine the expressions of c-Jun, cyclins D1, and D2 levels.


NSC Proliferation Depends on FGF2 and Insulin

To study how FGF2- and insulin-initiated signals act, NSCs were obtained from the cerebral cortex of embryonic day 13.5 C57/Bl6 mice. In every passage of cell culture, the developmental potential of NSCs progressively changes from tripotent to bipotent, and unipotent [30]. As a result, cells from the earlier period following passage are multipotent and during this specific period cell fate can be controlled by external signals [30]. To reduce the variability of time-dependent changes in cell state, we performed all experiments immediately following the second passage. After the initial expansion (Passage 0), cells were frozen at −80°C. They were then thawed and expanded (Passage 1) and all the experiments reported here were conducted at the next (second) passage.

To measure the relative contributions of insulin and FGF2 on NSC proliferation, we counted cell number after 4 days of expansion across 5 log concentrations of FGF2 and insulin in addition to no ligand controls (FGF2 zero concentrations not shown in the log plot) (Fig. 1A). Prior to adding growth factors, cells were cultured in N2 (minus insulin) without growth factors for 3 hours to reduce previous signaling activity. Applying an automatic microscope and high-content screening analysis, we achieved robust dose-response curves for these two ligands. The curves fall into four categories based on insulin dose: (a) in the absence of insulin, FGF2 increases cell number only weakly at 100 ng/ml; (b) the next two lowest doses of insulin allow greater proliferation, but the dose-response curve is not saturating at 100 ng/ml FGF2, suggesting that the effect of FGF2 is being limited by low insulin signaling; (c) increasing insulin to 0.25 µg/ml results in the first plateau of cell number at the highest FGF2 levels and a half-maximal effect at 1 ng/ml FGF2; (d) at the highest insulin dose of 25 µg/ml, the half-maximal dose of FGF2 is shifted left to 0.5 ng/ml. The robust ability of insulin level to tune the FGF2 dose-response curve supports the hypothesis of a two-step process where FGF2 action is required for the insulin response.

Figure 1.

FGF2 and insulin synergistically induce DNA synthesis and promote proliferation in neural stem cells. (A): Proliferation response to increasing doses of FGF2 and insulin after 96 hours in the labeled conditions. Error bars reflect SEM of four replicates for each condition (B) EdU incorporation response to the same doses of growth factor after 12 hours (C) EdU incorporation analysis for control (N2 − insulin − FGF2), insulin (N2 + insulin − FGF2), FGF2 (N2 – insulin + FGF2), F+I (N2 + insulin + FGF2), F + I + SU5402 (N2 + insulin +FGF2 + 5 µM SU5402). Significance limits *, p < .05; **, p < .01; ****, p < .0001. (D): Cell cycle analysis with flow cytometry. Abbreviation: FGF, fibroblast growth factor.

The percentage of cells in S-phase of the cell cycle was measured by labeling newly synthesized DNA with the thymidine analog 5-ethynyl-2-deoxyuridine (EdU). To label cells, a pulse of EdU was applied 1 hour before harvesting cells. Figure 1B shows the same dose-response treatments as in Figure 1A except here cells were analyzed 12 hours after release from growth factor starvation. The dose-response curves show a relatively low S-phase population at insulin levels below 0.025 µg/ml with maximal FGF2. Increasing insulin to 0.25 µg/ml dramatically increases the S-phase population at FGF2 doses above 1 ng/ml, further supporting the model of insulin limiting the FGF2 effect when FGF2 is saturating. Having determined maximal response with insulin at 25 µg/ml and FGF2 at asymptotic levels by 10 ng/ml, these doses were used for all future studies unless otherwise indicated.

In control conditions, where neither FGF2 nor insulin was applied, only 2% of cells incorporated EdU (Fig. 1C). Insulin slightly induced DNA synthesis but it was not significantly different from control cells indicating that insulin had minimal effect on DNA synthesis. Likewise, FGF2 modestly increased the number of S-phase cells to 7% of the total cells. Combining FGF2 and insulin significantly increased EdU-positive cells to 20% and this response was abolished when the FGF receptor inhibitor SU5402 [32] was also included. Dimethyl sulfoxide, the solvent for SU5402, had no effect on cell proliferation (data not shown). In the presence of FGF2, insulin, and SU5402, the S-phase population was the same as without growth factors. The SU5402 drug shows no inhibition of insulin receptor (IRS-1) phosphorylation in NIH3T3 cells treated with insulin [33]. Pharmacological antagonism of IGF-1 receptor and insulin receptor with AG1024 (50 nM to 5 µM) [34] was rapidly cytotoxic. The lethal effect of inhibiting insulin/IGF-1 signaling is consistent with previous reports of survival signaling associated with insulin/IGF-1 signaling in neurons and NSCs [35, 36].

To provide an independent assessment of cell cycle distribution in response to growth factor treatment, we used flow cytometry (Fluorescence-activated cell sorting). In these experiments, cells were incubated with FGF2, insulin, or both growth factors for 24 hours prior to flow cytometry analysis. After treatment with insulin alone, 4% of the cells were in S-phase. Exposure to FGF2 alone resulted in 10% of cells in S-phase. Combining FGF2 and insulin increased the percentage of cells in S-phase to 40% (Fig. 1D). These data show that FGF2 and insulin act synergistically to promote S-phase entry and proliferation of NSCs in the first 24 hours after passage when ∼50% of the cells are tripotent.

FGF2 and Insulin Activate Different Intracellular Kinase Cascades

The signaling pathways downstream of the FGF and insulin receptors are known to interact in multiple systems but the nature of the interaction varies between systems studied [37]. To determine the activation of Erk1/2 and Akt in NSCs, their phosphorylation status was examined by Western blotting. FGF2 promoted phosphorylation of Erk1/2 at threonine 202 and 204 (T202/204) but did not alter the phosphorylation status of Akt on either major regulatory position (threonine 308 or serine 473; Fig. 2A). In contrast, insulin promoted phosphorylation of Akt at both T308 and S473 but had no effect on activation of Erk1/2 (Fig. 2A). These effects were rapidly established and sustained, lasting for more than 12 hours. Care was taken to minimize mechanical disturbance, which alone is sufficient to activate Erk1/2 transiently (data not shown).

Figure 2.

FGF2 and insulin act through different signaling pathways. (A): Western blot analysis of cells treated with single growth factors FGF2 or insulin. (B): Inhibition of Akt and Erk1/2 phosphorylation with 10 µM LY294002 and 15 µM U0126. (C): EdU analysis of LY294002 and U0126 treated cells (*, p < .05). Abbreviation: FGF, fibroblast growth factor.

The specificity of pharmacological inhibition was tested by Western blotting. Akt phosphorylation was blocked with the PI3 kinase inhibitor LY294002, and Erk1/2 phosphorylation was inhibited with U0126 (Fig. 2B). Each inhibitor reduced phosphorylation of the expected target (Fig. 2B). These results allow the use of small molecule inhibitors to test the contribution of these signaling cascades to DNA replication at 12 hours after growth factor addition. In the presence of FGF2, U0126 decreased EdU-positive cells from 5.6% to 1.0%. In the presence of FGF2 and insulin, U0126 decreased EdU-positive cells from 22% to 5%. LY294002 decreased the number of proliferating cells from 22% to 9% (Fig. 2C). These data show that FGF2 and insulin activate distinct second messenger pathways that both contribute to maximal NSC proliferation.

FGF2 Activates C-Jun and C-Fos

Among the common substrates for Erk1/2 kinase are the AP-1 transcriptional proteins composed of the Fos and Jun gene families that are involved in cell proliferation [38-40]. Western blots were used to examine the activation of AP-1 proteins including c-Fos, c-Jun, JunB, Fra-1, and Fra-2 in cells treated with FGF2 and insulin. Only c-Fos and c-Jun responded to growth factor stimulation. The other AP-1 transcription factors were either absent or the levels did not change when treated with FGF2 and insulin (data not shown).

FGF2 activated expression of c-Fos and c-Jun by 1-hour post stimulation while insulin did not (Fig. 3A). c-Fos levels peaked between 1 and 3 hours and were declining markedly at 6 hours after FGF2 stimulation (Fig. 3A). In contrast, FGF2 induced sustained expression of c-Jun that lasted for more than 12 hours (Fig. 3A). Insulin did not activate either c-Fos or c-Jun. Combining insulin with FGF2 did not change the level of c-Fos and c-Jun compared to FGF2 alone indicating that insulin had no influence on c-Fos and c-Jun expression either in the presence or absence of FGF2 (Fig. 3A).

Figure 3.

FGF2 activates c-Fos and c-Jun. (A): Western blot of c-Fos and c-Jun levels in growth factor(s) treated cells. (B): The level of phosphorylated c-Jun in FGF2 alone or in combination with insulin. (C): Effect of LY294002 and U0126 on c-Fos and c-Jun proteins. Abbreviation: FGF, fibroblast growth factor.

Phosphorylation and other post-translational modifications of c-Fos and c-Jun are known to regulate their stability and activity [38, 39, 41]. In NSCs, FGF2 treatment induced phosphorylation of c-Jun at S63 and S73. Addition of insulin did not appreciably change the phosphorylation state of c-Jun over the first 6 hours of this experiment (Fig. 3B). In the presence of FGF2 and insulin, the inhibitor U0126 decreased the phosphorylation of S63 and S73 (Fig. 3C). PI3 kinase inhibition with LY294002 did not decrease and may stimulate phosphorylation of c-Jun (Fig. 3C). U0126 decreased c-Fos and c-Jun levels in the presence of FGF2 and insulin while LY294002 had no effect on the level of c-Fos or c-Jun (Fig. 3C). These data extend the distinction between FGF2 and insulin induced signaling to the induction of the AP-1 genes. The level of phosphorylated c-Fos was also examined. None of the growth factors either alone or in combination induced phosphorylation of c-Fos (data not shown). These data indicate that FGF2-Erk1/2 signaling has a major effect on the expression and activation of the AP-1 transcriptional response system. c-Jun is well-established as a direct transcriptional activator of cyclin D expression [42, 43].

FGF2 and Insulin Regulate Cyclins

The interaction between D cyclins and cyclin-dependent kinases, Cdk4/6, exerts a major control on cell proliferation by regulating the entry into S-phase. This is achieved by Cdk phosphorylation of the retinoblastoma (Rb) protein and relief of Rb-mediated inhibition of E2F-1 that plays a central role in cell cycle progression [44]. To assess the kinetics of these regulatory events, the levels of cyclin D, cdk4, cdk6, and phosphorylated Rb (Phospho-Rb) were measured at 6 and 12 hours after growth factor treatment (Fig. 4). FGF2 induced the expression of cyclins D1 and D2 but not D3. In contrast, insulin induced the expression of cyclin D3 but not D1 or D2 at 6 and 12 hours (Fig. 4A).

Figure 4.

FGF2 and insulin act synergistically to establish cyclin D1-D3 protein levels. (A): Western blot for cell cycle proteins in response to F (FGF2 alone), I (Insulin alone), and F + I (both factors). (B): Specific pharmacological antagonism of Erk1/2 and PI3K reduces Cyclin D1, D2, and D3 levels. (C): Western blot of Cyclins after siRNA knockdown 0 and 12 hours after adding FGF2 and Insulin (25 µg/ml). siRNA are indicated along the top. (D): EdU analysis of cells transfected with the indicated siRNAs.

The combination of FGF2 and insulin showed increased expression of cyclins D1, and to a lesser extent D3 compared to either growth factor alone at both 6 and 12 hours post stimulation (Fig. 4A). No combination of FGF2 and insulin altered the levels of cdk4 or 6 (Fig. 4A). FGF2 alone increased the level of phosphorylation on Rb serine 780 (S780) at 12 hours while insulin did not. Here again, the combination of FGF2 and insulin showed elevated phosphorylation of Rb (Fig. 4A). Inhibition of Erk1/2 for 12 hours with U0126 in the presence of FGF2 and insulin showed expected decrease in cyclins D1 and D2 (not D3) while inhibition of PI3k-Akt with LY294002 showed expected decrease in cyclins D1, D2, and D3 (Fig. 4B).

These data are consistent with the synergistic effect on DNA synthesis when NSCs were exposed to both growth factors (Fig. 1D). These results suggest that cyclins D1/D2 and D3 protein levels are initially separately controlled by the FGF/MAPK/ERK and Insulin/Akt pathways but that sustained expression of p-Rb and cyclins at 12 hours is achieved by integrating signals from the two pathways.

FGF2 and Insulin Synergism Requires Cyclin D1 and D2

Mouse cells can divide and proliferate without cyclins D1, D2, and D3 in embryos up to mid-late gestation [45], which indicates that there is a cyclin D-independent pathway for cell proliferation. To test whether NSCs depend on cyclins D for cell proliferation, small interfering RNA (siRNA) was used to specifically reduce the level of each cyclin D mRNA in the presence of FGF2 and insulin. Cyclin D protein levels were effectively knocked down by siRNA (Fig. 4C). The level of pRb S780 was also examined after cyclin knockdown. Knockdown of cyclins D1 and D2 clearly reduced phospho-Rb while cyclin D3 had a less marked effect. The Cyclins D1 and D2 double knockdown led to a further reduction in the level of pRb (Fig. 4C).

To test the functional consequence of cyclin knockdown at the cell-cycle level, EdU incorporation was measured. Knockdown of cyclin D1 and D3 decreased EdU-positive cells from 19% to 12% (p < .05), while cyclin D2 knockdown decreased EdU-positive cells from 19% to 6% (p < .05) (Fig. 4D). Because FGF2 specifically increased the level of cyclins D1 and D2 in Figure 4A, we mimicked FGF2 loss by targeting cyclins D1 and D2 with siRNA and this resulted in decrease of EdU-positive cells from 19% to 2% (p < .05, Fig. 4D). These data demonstrate that DNA synthesis responds to the loss of cyclins D1, D2, and D3. Unlike the early embryo, midgestation stage NSC proliferation is cyclin-dependent.

To determine whether the integration of FGF2-Erk1/2 and insulin-PI3k/Akt signaling on cyclins D1 and D2 expression was at the mRNA or post-transcriptional level, we examined the mRNA and protein level of cyclins following treatment with FGF2 or in combination with increasing doses of insulin for 6 hours (Fig. 5). At the protein level, increasing insulin dose from 0 to 25 µg/ml produced a clear response in increased cyclin D1, D2, and D3 (Fig. 5A). As expected, phosphorylation of AKT on serine 473 also showed dose-responsiveness to insulin. Thus, insulin regulates cyclin D1, D2, and D3 protein levels in a dose-dependent manner. To what extent does insulin regulate cyclin mRNA level? We performed quantitative PCR on cells grown under the same conditions as in Figure 5A (Fig. 5B, open bars). Increasing insulin dose had no effect on cyclin D1, D2, or D3 mRNA level. For direct comparison between mRNA and protein level, we plotted the quantified results above from Western blot in Figure 5A for each cyclin and qPCR. FGF2 had a significant effect on increasing cyclin mRNA level as shown by comparing the qPCR levels before FGF2 addition (t = 0, far left lane) and 6-hour after FGF2 addition (far right lane). Together, these data support the hypothesis that FGF2 promotes cyclin expression at the mRNA level and insulin promotes cyclin expression through a post-transcriptional mechanism (Fig. 6).

Figure 5.

FGF2 promotes cyclin transcription while insulin promotes cyclin translation. (A): Western blot of cyclin D1, D2, D3, and insulin-responsive phospho-AKT473 6 hours after adding indicated dose of insulin and FGF2. The left-most lane shows the protein levels after 3-hour growth factor starvation immediately before adding insulin. (B): Direct comparison of protein level (closed bars normalized to Tubulin) and mRNA level (open bars ΔΔCt normalized to GAPDH) measured by Western blot and qPCR 6 hours after growth factor treatment. Note that protein level (but not mRNA) changes in response to increasing insulin. Abbreviation: FGF, fibroblast growth factor.

Figure 6.

Proposed model showing regulation of cyclins D expression by FGF2 and insulin signaling pathways. FGF2 and insulin activate parallel pathways in Erk1/2 and PI3k/Akt to, respectively, control transcriptional and post-transcriptional levels of cyclins D1 and D2. Abbreviation: FGF, fibroblast growth factor.


The growth of precursors from the developing telencephalon that give rise to neurons and glia is increasingly used as a model system to determine molecular mechanisms controlling cell number and fate. Clonal analysis of NSCs in fully defined media showed that multipotent cells could be maintained in culture [2]. Real-time imaging definitively established tripotent cells are only transiently present for a brief period following passage [30]. These studies define the sequence of signaling events that control NSC survival [46] and show that cell fate is regulated by a novel instructive signaling mechanism [30]. Here, we took advantage of the cell synchrony that occurs following passage to define the mechanisms regulating proliferation of multipotent NSCs.

The identification of the link between insulin and cell growth was established by pioneering studies in cell culture [4, 47]. Bottenstein and Sato [4] used the rat neuroblastoma B103 cell line to establish the now widely used N2 defined media. N2 media was sufficient to promote proliferation of B103 in the absence of additional exogenous growth factors that stimulate cyclin D expression through the Erk/MAPK pathway. B103 was originally derived from in vivo N-ethyl-n-nitrosourea (ENU) mutagenesis. Genomic DNA from B103 was used to identify the neu oncogene [48]. Neu (ErbB2) homodimers activate the Erk/MAPK pathway and heterodimers with other epidermal growth factor-receptor family members can diversify the signaling to include the PI3K > AKT > S6 kinase and PI3K > rac/rho/cdc42 pathways [49]. If B103 rat neuroblastoma cells (like mouse NSCs) require both FGF and insulin signaling for proliferation, neu signaling could potentially substitute for the FGF > Erk signal and support expansion in N2 medium with insulin as the sole exogenous RTK ligand. Expansion of nestin-positive precursor cells from the developing mouse brain required FGF2 as well as insulin but the signaling pathways linking these ligands to cell growth had not been defined [2, 50]. Using direct counts of cell number, a label for proliferating cells and cell cycle analysis, we show a synergistic interaction between FGF2 and insulin (or IGF-1). The subsequent studies reported here establish that distinct cellular pathways mediate this response.

EGF is commonly used to create suspension neurosphere cultures. In striatal NSCs cultured from the developing mouse as neurospheres, there is a similar dependence on both an insulin-like signal (IGF-1) and an FGF2 or EGF signal [3]. This suggests that the convergence of these two pathways at cyclin D may be a general phenomenon common to all neural precursors including those with a neuronal bias or glial bias.

Insulin receptor activation triggers a cascade of downstream kinases including Akt and mTOR that play a conserved role in cell growth and longevity in animals [51]. In previous work, we have reported that the precise kinetics of this signaling pathway controls NSC survival [46]. Others have shown that FGF signals through MAPK to control NSC proliferation [52]. This report is the first to report that MAPK signaling downstream of FGF2 controls proliferation through elevating cyclin mRNAs in multipotent NSCs. In contrast, stimulation of the PI3K/Akt pathway does not alter the level of cyclin RNA but rather regulates cyclin protein levels.

Beyond developing an understanding of growth control mechanisms in experimental NSC cultures, the modulation of both FGF2/MAPK and Insulin/Akt pathways has potential clinical relevance in treating Glioma. Both pathways are current active targets for clinical intervention both in adults and pediatric cases (reviewed in [53]). The clean separation of these signaling pathways in primary NSCs will be helpful in disentangling the inappropriate signaling pathways found in tumor cells.


These data support a model of signaling pathway activation that involves a Boolean AND operator coupling gene expression at the mRNA and protein levels to regulate cyclin levels and proliferation (Fig. 6). A recent genome-wide study of mRNA abundance and protein abundance demonstrates that in mouse fibroblasts, the cellular abundance of protein is largely regulated at the level of translation, not transcription [54]. Many studies have reported that Akt regulates the level of cyclin D proteins by post-transcriptional mechanisms, including promoting nuclear export of the RNA, increasing the rate of translation, and preventing cyclin D degradation [55-57]. Defining the interaction of Ras/Mapk/Erk and PI3K/Akt/mTOR signaling in different neural cell types is critical to understanding the role of these pathways in normal brain development, in cancer, and the psychiatric consequences of aberrant signaling.


This research was supported in part by the Intramural Research Program of the NIH, NINDS, and the Lieber Institute for Brain Development. A.A. was supported by the Howard Hughes Medical Institute-NIH Research Scholars Program.

Author Contributions

A.A., N.M., K.O., and D.H.: concept and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; R.M.: concept and design, financial support, data analysis and interpretation, and manuscript writing.

Disclosure of Potential Conflicts ofInterest

The authors indicate no potential conflicts of interest.