|Home | About | Journals | Submit | Contact Us | Français|
Cerebellar granule neurons are the most abundant neurons in the brain, and a critical element of the circuitry that controls motor coordination and learning. In addition, granule neuron precursors (GNPs) are thought to represent cells of origin for medulloblastoma, the most common malignant brain tumor in children. Thus, understanding the signals that control the growth and differentiation of these cells has important implications for neurobiology and neuro-oncology. Our previous studies have shown that proliferation of GNPs is regulated by Sonic hedgehog (Shh), and that aberrant activation of the Shh pathway can lead to medulloblastoma. Moreover, we have demonstrated that Shh-dependent proliferation of GNPs and medulloblastoma cells can be blocked by basic fibroblast growth factor (bFGF). But while the mitogenic effects of Shh signaling have been confirmed in vivo, the inhibitory effects of bFGF have primarily been studied in culture. Here we demonstrate that mice lacking FGF signaling in GNPs exhibit no discernable changes in GNP proliferation or differentiation. In contrast, activation of FGF signaling has a potent effect on tumor growth: treatment of medulloblastoma cells with bFGF prevents them from forming tumors following transplantation, and inoculation of tumor-bearing mice with bFGF markedly inhibits tumor growth in vivo. These results suggest that activators of FGF signaling may be useful for targeting medulloblastoma and other Shh-dependent tumors.
Normal development requires a delicate balance between proliferation and differentiation. Too little proliferation can result in impaired tissue structure and function, while excessive proliferation can lead to cancer. One salient example of this is the development of granule neurons in the cerebellum. Mutant mice that fail to generate enough granule cells develop severe ataxia (1-3), and patients with congenital granule cell degeneration have deficits in motor coordination, language use, and cognitive function (4). Conversely, when GNP differentiation fails, cells that would normally become post-mitotic continue to proliferate and give rise to medulloblastoma (5, 6). Thus, elucidating the molecular mechanisms that control growth and differentiation of granule neurons is critical for understanding both normal cerebellar development and tumorigenesis.
GNPs are generated on the surface of the cerebellum in a structure called the external germinal layer (EGL) (7, 8). At birth, the EGL contains a single layer of undifferentiated cells, but over the next few days, these cells undergo extensive proliferation to generate a large pool of GNPs (9, 10). As new GNPs are generated, older cells exit the cell cycle and differentiate, and their cell bodies migrate inward to the internal granule layer (IGL) (11, 12). The waves of proliferation and differentiation continue until about 3 weeks of age (in mice), when all GNPs mature into granule neurons and the EGL disappears (13).
Studies from our lab and others have demonstrated that the major mitogen for GNPs is Sonic hedgehog (Shh) (14-16). Produced by Purkinje neurons, Shh acts by binding to the transmembrane protein Patched (Ptc). In the absence of Shh, Ptc acts as an antagonist of the hedgehog pathway, inhibiting the function of the signal transducing protein Smoothened (Smo). When Shh binds to Ptc, inhibition is alleviated, and Smo initiates a signaling cascade that results in activation of Gli family transcription factors and expression of hedgehog target genes (17, 18). In GNPs, these include N-myc and cyclins D1 and D2 (19, 20), and thus Shh signaling results in a robust proliferative response. Notably, mice with targeted mutations that impair Shh signaling (e.g. deletion of Shh, Smo or Gli2) exhibit decreased GNP proliferation (21-23); conversely, animals with mutations that activate the pathway (e.g. deletion of ptc) develop medulloblastoma (5, 6, 24-26). The fact that ~30% of human medulloblastomas contain activating mutations in the Shh pathway (27-29) also implies a key role for this pathway in proliferation and tumorigenesis.
But while the importance of Shh in GNP proliferation is well established, the signals that cause GNPs to exit the cell cycle and differentiate remain much less clear. One potential explanation might be decreased exposure of GNPs to Shh ligand. However, previous studies have shown that Shh continues to be expressed by Purkinje cells into adulthood (30). The fact that GNPs can exit the cell cycle despite the continued presence of Shh suggests that some other signal may override the mitogenic effects of Shh. We have previously shown that bFGF can block proliferation and promote differentiation of GNPs in the presence of Shh (14, 31), raising the possibility that FGFs might serve as differentiation factors for these cells. In addition, we have shown that bFGF can inhibit the growth of tumor cells from ptc mutant mice (31), suggesting that this pathway may hold promise for medulloblastoma therapy.
Although FGFs have been shown to inhibit proliferation of GNPs and tumor cells in vitro, their effects on these cells have not been rigorously tested in vivo. In the current study, we used conditional knockout mice to investigate the importance of FGF signaling in GNP development in vivo, and intratumoral injections to test the efficacy of bFGF as a treatment for medulloblastoma. Our results suggest that genetic ablation of FGFRs does not alter GNP differentiation in vivo. However, exposure of medulloblastoma cells to bFGF dramatically inhibits their growth. Thus, while bFGF may not be critical for GNP development, activation of the FGF pathway may be useful in the treatment of Shh-dependent tumors.
GNPs undergo proliferation and differentiation in the EGL during the first 2-3 weeks after birth. Our previous studies (31) have shown that bFGF inhibits proliferation and promotes differentiation of GNPs. To determine whether the inhibitory effects of bFGF are age-dependent, we isolated GNPs at a range of postnatal ages and measured incorporation of tritiated thymidine in response to Shh and bFGF. Consistent with our previous observations, P7 GNPs proliferated robustly in response to Shh, and this response was abolished by addition of bFGF (Figure 1C). Marked inhibition of Shh-induced proliferation was also seen at P2, P4 and P11 (Figures 1A, B and D). Thus, bFGF can inhibit proliferation of GNPs throughout postnatal development.
Our previous studies suggested that bFGF promotes cell cycle exit of GNPs and accelerates their differentiation into granule neurons (31). To better understand the mechanisms by which bFGF inhibits proliferation, we performed cell cycle analysis on cells treated with Shh ± bFGF. As shown in Supplementary Figure 1A-C, in the absence of growth factors the majority of GNPs exit the cell cycle within 48 hours (< 2% of cells in S, G2 and M phases of the cell cycle). Treatment with Shh maintains a population of GNPs in cycle (~11% of cells in S/G2/M). However, co-treatment with bFGF inhibits the effects of Shh and causes cells to accumulate in the G0/G1 phase of the cell cycle (~2% of cells in S/G2/M). We also examined whether bFGF increased apoptosis by staining cells with antibodies specific for cleaved caspase 3. As shown in Supplementary Figure 2A, bFGF did not cause a significant increase in apoptosis. Together, these data suggest that bFGF acts by promoting cell cycle exit and differentiation rather than by inducing cell death.
We previously showed that bFGF can inhibit proliferation of GNPs in response to stimulation with Shh ligand, as well as of tumor cells resulting from mutations in patched (31). These data suggest that bFGF acts on the Shh pathway at a level downstream of patched. To gain further insight into the mechanisms by which bFGF interferes with Shh signaling, we tested the effects of bFGF on cells in which the Shh pathway is activated as a consequence of a Smo mutation. NeuroD2-SmoA1 transgenic mice exhibit increased Shh pathway activation and proliferation of GNPs and develop medulloblastoma (32). As shown in Supplementary Figure 3A, we found that bFGF inhibits the basal proliferation of these cells, suggesting that it acts downstream of Smo.
Shh target genes can also be activated by mutations in Suppressor of Fused (Sufu), a negative regulator of the pathway that regulates processing, localization and transcriptional activity of Gli proteins (33-35). To determine whether bFGF could suppress activation of the pathway resulting from loss of Sufu, we tested the effects of bFGF on mouse embryo fibroblasts (MEFs) from Sufu-deficient mice (36). As shown in Supplementary Figure 3B, Sufu−/− MEFs exhibit constitutive expression of gli1 that is significantly higher than that seen in wild type MEFs. Importantly, addition of bFGF to Sufu−/− MEFS results in a 3-fold reduction in gli1 expression, suggesting that bFGF can inhibit activation of the pathway mediated by loss of Sufu. Similar results were observed when a Gli-luciferase reporter gene was used to monitor gli1 expression (data not shown). Together these data suggest that bFGF inhibits Shh signaling at a level downstream of Smo and Sufu, and proximal to the nucleus.
The FGF receptor family consists of four members (FGFR1-4), three of which can undergo alternative splicing to generate multiple receptor isoforms (FGFR4 exists in only one isoform) (37). To determine which receptors might mediate inhibition of Shh signaling and proliferation, we first examined the receptors expressed by GNPs. Cells were FACS-sorted from the cerebellum of Math1-GFP transgenic mice (which express green fluorescent protein (GFP) in their GNPs (38, 39)), and RNA was isolated and subjected to RT-PCR using primers for FGFR receptor isoforms. Consistent with previous studies (40), we found that GNPs express FGFR1, 2 and 4; FGFR3 was also detected in some samples, but usually at lower levels than the other receptors (Figure 2A).
We have previously shown that FGF-mediated inhibition of Shh signaling can be blocked by pharmacological antagonists of FGFR kinase activity (31). Because these antagonists can act on all FGF receptors, the particular receptor or receptors required for inhibition of hedgehog signaling remained unclear. To determine which FGFRs were required for the inhibitory effects of bFGF in GNPs, we used mice lacking FGFR1, 2 and 4, the predominant FGFRs expressed in these cells. Complete loss of FGFR1 or FGFR2 results in embryonic lethality (41, 42); therefore we crossed mice carrying loxP-flanked alleles of these genes (43, 44) with Math1-Cre transgenic mice (5, 6) to generate animals lacking FGFR1 or FGFR2 in GNPs. These animals, along with germline FGFR4 knockout mice (which remain viable into adulthood (45, 46)), were used to examine the effects of loss of FGFRs on GNP responses to Shh and bFGF. GNPs from single knockout mice were cultured in the presence of Shh ± bFGF for 48 hours and then assayed for incorporation of tritiated thymidine. As shown in Figure 2B-D, GNPs from mice lacking FGFR1, FGFR2 or FGFR4 all showed robust proliferation in response to Shh. But while loss of FGFR2 or FGFR4 did not affect FGF-mediated inhibition of Shh-induced proliferation (Figure 2 C-D), loss of FGFR1 completely abrogated the inhibitory effects of bFGF (Figure 2B). Consistent with its inability to suppress Shh-induced proliferation, bFGF was also unable to inhibit Shh induction of gli1 in FGFR1-deficient GNPs (Supplementary Figure 4). FGFR1-deficient GNPs treated with Shh also showed no change in cell cycle distribution following exposure to bFGF (Supplementary Figure 1D-F). Together, these data indicate that the inhibitory effects of bFGF require signaling through FGFR1.
Our previous studies (31) demonstrated that FGF-mediated inhibition of Shh signaling is mediated, at least in part, by activation of the extracellular signal-regulated kinase (ERK). To determine whether GNPs lacking FGFR1 were still capable of activating this kinase, we stimulated cells with bFGF and examined the phosphorylation status of ERK. As shown in Supplementary Figure 5, wild type (WT) GNPs exhibited robust ERK phosphorylation in response to bFGF. In contrast, GNPs from mice lacking FGFR1 showed no increase in ERK phosphorylation. These studies suggest that loss of FGFR1 renders GNPs unresponsive to bFGF.
Since our studies indicated that FGFR1 was a key mediator of FGF-mediated inhibition in vitro, we next examined the effects of loss of FGFR1 on GNP differentiation in vivo. We isolated cerebella from WT and FGFR1-deficient mice at P7 and stained them with antibodies specific for markers of proliferation and differentiation. As shown in Figure 3A, WT cerebella contained a broad band of proliferating (Ki67+) cells in the outer EGL and discrete regions of differentiating (NeuN+) granule neurons in the inner EGL and IGL. Surprisingly, expression of Ki67 and NeuN in FGFR1 knockout mice was indistinguishable from that seen in WT littermates (Figure 3B). These results suggested that loss of FGFR1 does not impair GNP cell cycle exit or differentiation.
One possible explanation for the lack of an in vivo phenotype in FGFR1-deficient mice was compensation by other FGFR receptors. To address this possibility, we generated triple knockout (TKO) mice lacking FGFRs 1, 2 and 4 in GNPs. Similar to GNPs from FGFR1 knockout mice, Shh-induced proliferation of GNPs from TKO mice was not inhibited by bFGF in culture (Supplementary Figure 6), nor did TKO GNPs treated with bFGF exhibit an increase in apoptosis (Supplementary Figure 2B). In contrast, BMP2, which inhibits GNP proliferation through a distinct mechanism (47, 48)) remained capable of inhibiting Shh-induced proliferation of TKO GNPs (Supplementary Figure 6). However, analysis of TKO cerebella showed no significant differences in proliferation or differentiation when compared to WT littermates (Figure 3C-D). Similar results were seen when cerebella were analyzed at P2, P4, P11 and P14 (data not shown). Based on these findings we conclude that FGF signaling, while capable of promoting differentiation in vitro, is not required for GNP differentiation in vivo.
While the above experiments suggest that FGF signaling is not essential for GNP development, they leave open the possibility that it might be useful as a treatment for Shh-dependent tumors. Recent studies from our lab have indicated that tumors from ptc+/− mutant mice are propagated by a subpopulation of tumor cells that expresses the carbohydrate antigen CD15/SSEA-1 (49). To determine whether CD15+ cells are sensitive to FGF-mediated inhibition, we purified these cells by flow cytometry, cultured them in the presence or absence of bFGF, and then measured their incorporation of tritiated thymidine. As shown in Figure 4A, bFGF markedly inhibited the growth of both unfractionated and CD15+ tumor cells. The basal proliferation of CD15− cells was much lower than their CD15 counterparts, but these cells were inhibited by bFGF as well. Moreover, we did not see an increase in apoptosis when unsorted tumor cells from ptc+/− mice were cultured in the presence of bFGF, indicating that the reduced proliferation observed following bFGF treatment was not due to increased cell death (Supplementary Figure 7). These results indicated that tumor-propagating cells from ptc+/− mice are sensitive to the inhibitory effects of bFGF.
Although bFGF inhibited the growth of tumor cells, it was possible that this inhibition was transient, or that a subpopulation of tumor cells remained capable of forming tumors even after exposure to bFGF. To test this possibility, we isolated tumor cells from ptc+/− mice and cultured them in the presence or absence of bFGF for 24hr. Cells were then transplanted into the cerebellum of SCID-beige mice and recipients were monitored for symptoms of tumor development (Figure 4B). As shown in Figure 4C, 93% of mice that received tumor cells cultured in control media went on to form tumors. In contrast, no tumors resulted from transplantation of FGF-treated cells. These studies demonstrated that the growth-inhibitory effects of bFGF are not transient, but rather, result in a long-lasting inability of cells to form tumors in vivo.
Having shown that bFGF pretreatment could abolish the tumorigenic potential of ptc+/− medulloblastoma cells, we wondered whether bFGF could also inhibit the expansion of tumor cells that were already growing in animals. To test this, we transplanted tumor cells into the cerebellum of SCID-beige mice, and after 8-12 days, injected PBS (vehicle) or bFGF into the implantation site. Animals received an additional injection 3-5 days later, and were then monitored for signs of tumor development (Figure 5A). For these experiments, we used tumors from Math1-GFP;ptc+/− mice, which allowed us to detect tumor bulk based on GFP expression (50). As shown in Figure 5B, tumors injected with PBS typically grew unchecked, encompassing a large portion of the cerebellum. In contrast, bFGF-treated tumors were barely detectable in intact cerebella (Figure 5C). Analysis of cerebellar sections revealed large numbers of GFP+ cells in PBS-treated animals, and few such cells in animals treated with bFGF (Figure 5D-G). H&E staining confirmed that the area of each tumor was comparable to the expression of GFP, both in the control and bFGF-treated animals (Supplementary Figure 8). To quantitate the reduction in tumor size following FGF treatment, we counted the number of sections in which tumor cells could be found. As shown in Figure 5H, FGF treatment resulted in a significant reduction in tumor bulk. Based on these results, we conclude that FGF is a potent inhibitor of tumor growth and may be useful for targeting ptc+/− mutant medulloblastomas and other hedgehog-dependent tumors.
The studies described above reinforce our previous observations that bFGF can inhibit Shh signaling and Shh-induced proliferation in GNPs and tumor cells. Moreover, they provide new insight into mechanisms of bFGF-mediated inhibition, demonstrating that (1) this requires signaling through FGFR1 and (2) that it acts on the Shh pathway at a level downstream of Smo and Sufu. Interestingly, a recent report (Lauth et al., 2010) demonstrates that oncogenic Ras can also inhibit Shh signaling, and that this involves activation of the DYRK1B kinase. Since bFGF signaling is likely to involve activation of Ras, it is tempting to speculate that bFGF-mediated inhibition involves a similar mechanism. However, it is notable that oncogenic Ras and Dyrk1B did not affect signaling in Sufu−/− cells, suggesting that the mechanisms underlying inhibition, while they might overlap with those associated with bFGF, are likely distinct.
Notwithstanding the potent inhibitory effects of bFGF in vitro, our studies demonstrate that FGF signaling is not required for normal GNP development. The fact that mice lacking FGF receptors showed no obvious phenotype is surprising, given the marked anti-mitogenic effects of FGF signaling for these cells in vitro. One possible explanation for this is that some level of FGF signaling persists even in the absence of FGF receptors. FGFs are known to activate transcription of Sprouty genes, which function as negative regulators of signaling by FGFRs and other receptor tyrosine kinases (51). In the absence of FGF signaling, this negative feedback loop could be lost, resulting in a compensatory increase in activation of the FGF pathway. Alternatively, it is possible that other extracellular signals may act independently to promote cell cycle exit and differentiation of GNPs in the postnatal cerebellum. In this regard, it is worth noting that in addition to bFGF, vitronectin, pituitary adenylate cyclase activating polypeptide (PACAP) and bone morphogenetic proteins (BMPs) 2 and 4 have all been reported to inhibit Shh-induced proliferation of GNPs in vitro (47, 48, 52, 53). To date, none of these have been shown to be required for GNP differentiation in vivo. Thus, the signals responsible for GNP cell cycle exit and differentiation remain unclear, and the possibility that multiple factors cooperate to regulate these processes remains open.
Whatever its role in GNP development, our studies clearly indicate that FGF signaling can inhibit the growth of GNP-derived tumors. This is significant for several reasons. First, it highlights the unique characteristics of the tumor-propagating cells (cancer stem cells) in these tumors. In contrast to the stem-like (CD133+, multipotent, neurosphere-forming) cells that have been shown to be important for propagation of other brain tumors (54, 55), we have found that growth of ptc-associated medulloblastoma is dependent on GNP-like cells that express the surface marker CD15 (49). Although a recent report suggested that long-term culture of ptc+/− tumor cells in FGF-containing media can expand a rare population of stem-like cells with tumorigenic potential (56), our demonstration that acute bFGF treatment – either in vitro or in vivo – dramatically inhibits subsequent tumor growth raises questions about the role of these cells in tumor maintenance and their importance as targets for therapy. Although bFGF is likely to have different effects on different classes of brain tumors (and could potentially enhance the growth of some tumor subtypes), at least for Hedgehog pathway-associated medulloblastoma, it appears to be a potent inhibitor of tumor growth.
Hedgehog antagonists have recently moved into clinical trials for a variety of types of cancer, and have begun to show some promise for treatment of medulloblastoma (57). However, early reports also suggest that resistance to these antagonists can develop quite readily (58). Therefore, it is critical to develop alternative approaches to targeting these diseases. Although bFGF does not readily cross the blood-brain barrier (59), our results suggest that intratumoral injection (perhaps aided by convection-enhanced delivery (60)) could be an effective mode of therapy. Small molecules that mimic the effects of bFGF (61-63) might also be useful for overcoming the blood brain barrier. The fact that some medulloblastoma cell lines are growth inhibited by bFGF (64, 65) suggests that our results are not confined to mouse models, and would be worth pursuing in the context of human medulloblastoma.
Math1-Cre transgenics (5, 6) were from David Rowitch at UCSF. FGFR1 conditional knockout (Fgfr1F/F) mice (43) were obtained from Chuxia Deng at the National Institutes of Health; FGFR2 conditional knockouts (Fgfr2F/F (44)) were provided by David Ornitz at Washington University. FGFR4 germline knockouts (Fgfr4−/− (66)) were maintained on a C57/BL6 background in our colony. NeuroD2-SmoA1 transgenic mice (32) were a gift from James Olson, Fred Hutchinson Cancer Center. FGFR triple knockout (TKO) mice were offspring of Fgfr1F/F; Fgfr2F/F; Fgfr4−/− x Fgfr1F/F; Fgfr2F/F; Fgfr4+/−; Math1-Cre+/− pairs. ptc+/− mice (24) were bred in our colony. Animals were bred and maintained in the Cancer Center Isolation Facility at Duke Medical Center.
bFGF (Peprotech, Rocky Hill, NJ) was used at 25 ng/mL except where indicated. Shh supernatant was generated by transfecting 293T cells with Shh-N expression plasmid (David Robbins, Dartmouth Medical School, Hanover, NH) and harvesting supernatant for 3 days. Supernatant was used at 20% final concentration.
GNPs were isolated from 2 to 11-day-old mice as described in (50). Tumor cells were isolated using a similar procedure from ptc+/− and NeuroD2-SmoA1 mice displaying signs of medulloblastoma. For isolation of CD15+ and CD15− populations, tumor cells were stained with anti-CD15 antibodies (clone MMA, BD Biosciences, San Jose, CA) and with phycoerythrin-conjugated goat-anti-mouse IgM secondary antibodies (Jackson ImmunoResearch, West Grove, PA), and then sorted using a BD FACSVantage flow cytometer. Cells were cultured on poly-D-lysine (PDL)-coated plates in Neurobasal medium containing 2% B27, 1 mM sodium pyruvate, 2 mM L-glutamine, and 1% Pen/Strep (all from Invitrogen, Carlsbad, CA).
RNA was isolated by using the RNAqueous kit (Ambion, Austin, TX). Lysates were treated with DNA-free DNase treatment and removal reagents (Ambion), and RNA was quantitated on a TD-700 fluorometer (Turner BioSystems, Sunnyvale, CA) by using RiboGreen (Invitrogen). RT-PCR for fgfr1-4 was performed using primers described in (34). First-strand cDNA was synthesized using equivalent amounts of total RNA (0.1-1 μg) in a 20-μl reverse transcriptase reaction mixture (Invitrogen). PCR was performed using Platinium Taq (Invitrogen) in a 25uL reaction followed by electrophoresis on a 1.5% agarose gel. Real-time PCR for gli1 was performed as described in (50). Sufu−/− MEFs were a generous gift from Matthias Lauth and Rune Toftgard at the Karolinska Institute, Huddinge, Sweden.
GNPs from P7 WT and FGFR TKO mice were cultured for 4 hours on PDL-coated plates in Neurobasal media with no additives or with 25ng/mL bFGF. Cells were washed twice with cold PBS and lysed in 200 μl RIPA buffer (0.5M Tris-HCl, pH 7.4, 1.5M NaCl, 2.5% deoxycholic acid, 10% NP-40, 10mM EDTA) with protease and phosphatase inhibitor cocktail (Thermo, Cat# 1861280) for 30 minutes on ice. Lysates were sonicated using an ultrasonicator (Misonix) at amplitude 2 for 5 seconds, and then centrifuged at 13,000 rpm for 10 min at 4°C. Protein concentrations were measured using a Bradford assay. Equal amounts of protein were separated by 8% SDS polyacylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes (Invitrogen, Cat# LC2006). Membranes were blocked with 5% BSA and then probed with anti-pERK and anti-total ERK (both from Santa Cruz), and with IRDye680-labeled Goat anti-Mouse IgG antibodies (Li-Cor Biosciences). All antibodies were diluted 1:3000 in Tris-buffered saline with 0.1% Tween-20. Proteins were visualized using Li-Cor Odyssey Imager.
For measurement of thymidine incorporation, GNPs and tumor cells were cultured in PDL-coated 96-well plates at 2 × 105 cells per well. Growth factors were added at the beginning of culture, and cells were incubated for 48 h before being pulsed with [methyl-3H]-Thymidine (PerkinElmer, Fremont, CA). After 16 h, cells were harvested by using a Mach IIIM Manual Harvester 96 (Tomtec, Hamden, CT), and incorporated radioactivity was quantitated by using a Wallac MicroBeta microplate scintillation counter (PerkinElmer).
For cell cycle analysis, cells were cultured in PDL-coated plates as described above. After 48 hours, cells were fixed and permeabilized for 20 min in 100 μl of BD Cytofix/Cytoperm Solution (BD Biosciences, San Jose, CA). Cells were then washed with 1 ml of BD Perm/Wash Buffer (BD Biosciences) and resuspended in 20 μl of 7-Aminoactinomycin D (7-AAD). 300 μl of FACS buffer (5% fetal bovine serum in PBS) was added to the cells, and samples were analyzed on a FACScan Flow Cytometer (Becton Dickinson, Franklin Lakes, NJ). Cell cycle profiles were analyzed using FlowJo software (TreeStar Inc., Ashland, OR).
For staining of frozen tissue, brains were fixed in 4% paraformaldehyde, cryoprotected in 30% sucrose, embedded in Tissue-Tek OCT (Sakura Finetek, Torrance, CA) and sectioned sagittally at a thickness of 12 μm. Sections were permeabilized with PBST (PBS + 0.1% Triton X-100), blocked with PBST + 10% normal goat serum, and incubated with primary antibodies specific for Ki67 (Abcam, Cambridge, MA) and NeuN (Millipore, Temecula, CA) overnight at 4°C followed by Alexa Fluor 594-anti-rabbit and Alexa Fluor 488-anti-mouse secondary antibodies (Molecular Probes, Eugene, OR) for 1 hr at room temperature. Sections were counterstained with DAPI (Molecular Probes) and mounted in Fluoromount-G (Southern Biotechnology Associates, Birmingham, AL). Tiled images were taken using a Zeiss LSM 510 inverted confocal microscope and related software at the Duke Light Microscopy Core Facility.
To measure apoptosis, GNPs or tumor cells were fixed in 4% PFA and stained with antibodies against cleaved caspase-3 (Cell Signaling Technology) and goat anti-rabbit IgG secondary antibodies (Jackson Immunoresearch). Following staining, immunofluorescent pictures of six randomly-selected fields were taken from each well of cultured cells using a Zeiss LSM 510 inverted confocal microscope and related software at the Duke Light Microscopy Core Facility. Cells were counted for positive cleaved caspase-3 and DAPI respectively. The ratio of cleaved caspase-3 to DAPI staining was determined for each culture condition and compared.
Tumor cells from Math1-GFP; ptc+/− mice were transplanted into SCID-beige mice as described in (49). For bFGF pretreatment studies, animals were followed for 6 months and sacrificed when they developed symptoms of medulloblastoma (ataxia, lethargy, hydrocephaly). For in vivo bFGF treatment studies, PBS or bFGF (80ng/ml, 2.5 μl) was injected into the transplant site 8-12 days after transplantation, and again 3-5 days later. When any of the mice in a cohort exhibited symptoms, the entire cohort was sacrificed. Animals were perfused with 4% paraformaldehyde and brains were imaged by whole mount microscopy using a Leica MC 16 FA microscope, Micropublisher 5.0 RTV camera and QCapture software (QImaging, Surray, BC). Brains were then frozen in OCT, sectioned from end to end, and stained with hematoxylin and eosin or with DAPI as described above. Sections were examined for the presence of GFP+ tumor cells, and lateral tumor extension was calculated by counting the number of sections containing such cells and multiplying by the thickness of these sections.
Supplementary Figure 1. bFGF inhibits cell cycle progression. GNPs were isolated from control or FGFR1 KO mice at P7 and cultured in Neurobasal media with no supplements (Control), Shh or Shh + bFGF. Cell cycle analysis was performed after 48 hours. Shh promoted cell cycle progression in both WT and KO cells; but bFGF inhibited cell cycle progression only in WT cells.
Supplementary Figure 2. bFGF does not promote apoptosis of GNPs. Freshly isolated WT (A) and FGFR1 KO (B) GNPs were cultured for 24 hours in the presence or absence of 25 ng/mL bFGF, then fixed in 4% PFA and stained with antibodies against cleaved caspase-3 and DAPI. Confocal images of six randomly-selected fields were taken for each well. The percentage of DAPI+ cells positive for cleaved caspase-3 was determined. There was no significant change in the percentage of cleaved caspase-3 positive cells after bFGF treatment in either WT or FGFR1 KO GNPs (based on unpaired Student’s t-tests).
Supplementary Figure 3. (A) bFGF inhibits proliferation of tumor cells expressing activated Smoothened. Tumor cells from NeuroD2:SmoA1 transgenic mice were cultured on PDL-coated plates in control media or media containing 25ng/mL bFGF. After 48 hours, cells were pulsed with tritiated thymidine (3H-Td) and cultured for 16hr before being assayed for thymidine incorporation. Data represent means of triplicate samples ± SEM. Differences between control and bFGF-treated cells were statistically significant (p < 0.0001 based on unpaired Student’s t-test). (B) bFGF inhibits gli expression in WT and Sufu−/− MEFs. WT and Sufu−/− MEFs were cultured in the presence or absence of 25 ng/mL bFGF. After 24 hours, RNA was isolated and gli1 and actin were measured by RT-PCR. Data represent the ratio of gli1:actin, normalized to levels in Control WT MEFs. bFGF caused a three-fold reduction in expression of gli1 in Sufu−/− MEFs.
Supplementary Figure 4: FGFR1 is required for bFGF-mediated suppression of gli expression. GNPs were isolated from Math1-Cre; FGFR1F/F mice at P7 and cultured in control media or media containing Shh, Shh + 25 ng/ml bFGF, or Shh + 1 μM cyclopamine. After 24 hours, RNA was isolated and gli1 and actin levels were determined by real time RT-PCR. Data represent the ratio of gli1:actin, normalized to levels in Control WT GNPs.
Supplementary Figure 5. Impaired ERK phosphorylation in cells lacking FGF receptors. GNPs from FGFR TKO (Math1-Cre; Fgfr1F/F; Fgfr2F/F; Fgfr4−/−) mice and wild type littermates were treated with 25 ng/mL bFGF for 4 hours. Lysates were collected and analyzed by Western blotting for phospho-ERK (pERK) and total ERK (tERK). Increased ERK phosphorylation was observed in WT but not KO cells.
Supplementary Figure 6. Cells lacking FGFRs 1, 2 and 4 are not inhibited by bFGF. GNPs from FGFR TKO (Math1-Cre; Fgfr1F/F; Fgfr2F/F; Fgfr4−/−) mice and wild type littermates were cultured in either control (20% Mock supernatant) or Shh (20% Shh supernatant) media with or without 25ng/mL bFGF or 100ng/mL BMP2 for 48 hr. Cells were pulsed with 3H-Td and cultured for 16 hours before being assayed for thymidine incorporation. Data represent means of triplicate samples ± SEM. FGFR TKO GNPs proliferated in response to Shh; this proliferative response was not inhibited by bFGF, but was inhibited by BMP2. Differences in proliferation between Control and Shh, Shh and bFGF, and Shh and Shh+BMP were statistically significant (p < 0.0005 based unpaired Student’s t-tests) for both WT and TKO cells. Differences between Shh and Shh+bFGF were significant for WT but not for TKO cells.
Supplementary Figure 7. bFGF does not induce apoptosis of ptc+/− tumor cells. Tumor cells were cultured in the presence or absence of 25 ng/mL bFGF for 24 hr, then fixed in 4% PFA and stained for cleaved caspase-3 and DAPI. Confocal images of six randomly-selected fields were taken, and the percentage of DAPI+ cells positive for cleaved caspase-3 was determined. bFGF did not increase the ratio of cells positive for cleaved caspase-3. There was no significant change in the percentage of cleaved caspase-3 positive cells after bFGF treatment (based on unpaired Student’s t-test).
Supplementary Figure 8. H&E staining of PBS and bFGF-injected tumors. Cerebella from mice injected with ptc+/− medulloblastoma cells treated with PBS (A) or bFGF (B) were fixed, sectioned and stained with H&E. Tumor is denoted by asterisk in (A).
The authors thank Marcie Kritzik for help with the manuscript, Kerry Dorr for contributions to the early stages of this work, Albert Basson for helpful discussions, Chuxia Deng, David Ornitz and James Olson for providing mice, Matthias Lauth and Rune Toftgard for Sufu−/− MEFs, Sam Johnson for help with microscopy, Amanda Conway for help with Western blotting and Mike Cook and Beth Harvat for help with flow cytometry. This work was funded by grant number MH67916 from the National Institute of Mental Health, by funds from the Pediatric Brain Tumor Foundation of the US, and by a Leadership Award (LA1-01747) from the California Institute for Regenerative Medicine.