Introduction

Medulloblastomas are the second most common malignant tumors in childhood with a frequency of about 1 in 200,000 children annually and accounting for about 20% of brain tumors in children. These tumors arise in the cerebellum, and are classified as embryonal tumors composed of small blue cells; a feature shared with other supratentorial primitive neuroectodermal tumors [1]. Insight into the underlying molecular mechanisms for the biology of medulloblastoma is beginning to emerge. In subsets of medulloblastoma, activation mutations involving important regulators of the sonic hedgehog and the Wnt pathway have been reported [2, 3]. Frequent chromosomal changes including isochromosome 17q with loss of variable regions of 17p have also been documented. Although, the immunohistochemical detection of the expression of p53 gene protein has now been shown to be a poor prognostic marker in these tumors [4, 5, 6], mutations involving p53 gene, a gene located at 17p13.3, which is a locus often deleted in medulloblastomas or amplification of mdm2 gene (protein product binds p53) is uncommon, thus implicating other mechanism/s in the deregulation of p53 expression in medulloblastomas [6, 7, 8] Genomic amplification of the MYC oncogene or overexpression of the myc mRNA identifies a poor prognostic group of medulloblastoma. The close relationship between MYC oncogene amplification and the anaplastic /large cell histologic subset with a more aggressive phenotype is well documented [9, 10, 11]. CGH studies have identified specific genomic regions of amplification and deletion, and more recent correlations of these genomic alterations with gene expression profiling identify specific genetic and expression signatures that stratify the different histologic and prognostic groups of medulloblastoma [12, 13, 14]. In addition, conventional and array based CGH studies have shown genomic gains of a number of genes including PIK3CA, PGY1, MET, ERBB2, and CSE1L [15]. ERBB2 overexpression has been associated with advanced metastatic disease and poor clinical outcome [16, 17] while trk C expression has been associated with the desmoplastic histology and better survival [18, 19]. Digital karyotyping has also been utilized for the identification of amplification of otx2; a homeobox gene located at 14q22.3 with important role in specification and regionalization of the forebrain and midbrain during early neurogenesis, and it shows a distinct correlation with the anaplastic phenotype [20, 21]. Many of the gene mutations and oncogene amplifications described above represent progression events associated with subsets of aggressive medulloblastomas. There is very little understanding of the initiating events in the development of medulloblastoma. Efforts at identifying initiating events have been focused on genetic syndromes such as nevoid basal carcinoma syndrome due to the ptch gene mutation [22], Turcot’s syndrome with APC mutation [23] and Li-Fraumeni syndrome with p53 germ line mutations [24] all of which have been associated with medulloblastoma. The observation of the development of medulloblastoma in only a subset of the ptch mutation (ptch−/−) mice, even though all mice possess the same genetic predisposition suggests that other genetic event/s that influence susceptibility to medulloblastoma exist. One would expect that such an event would be present in a large proportion of medulloblastomas and of necessity in all histologic types of medulloblastoma. Using conventional and BAC array CGH, we have identified another 14q genetic event, a 14q12 amplicon, distinctly separate from the previously reported otx2 at 14q22.3 [20, 21] which includes the FOXG1 gene. FOXG1 copy gain was first observed by aCGH and confirmed by Qrt-PCR in 9 out of 9 analyzed medulloblastomas and by Qrt-PCR in 93% (55/59) of a validating subset of medulloblastomas. We demonstrate the increased expression of FOXG1 as detected by immunohistochemistry with FOXG1 specific antibody, as well as an inverse relationship between FOXG1 expression and the expression of p21cip1 in these medulloblastomas. FOXG1 has been implicated in the repression of TGF-β induced expression of p21cip1 [25]. Our findings are consistent with such a role for FOXG1. TGF-β is produced in the cerebellum and plays an important role in the elimination of neuronal precursor cells that do not establish neuronal connections. It also cooperates with BDNF to inhibit cellular proliferation in the cerebellum. We hypothesize that the deregulation of FOXG1 by gene copy gain and amplification may represent an early event upon which all other reported medulloblastoma subset specific mutations or genetic events resulting in deregulation of critical signaling pathways are superimposed for the development and progression of medulloblastoma. This is the first report of a role for FOXG1 in the pathogenesis of medulloblastoma.

Methods

Medulloblastoma tumors

Nineteen (19) medulloblastomas from the files of the Department of Pathology, University of Oklahoma Health Sciences Center, Oklahoma City were analyzed in this study. A validating set of 64 medulloblastomas were retrieved from the files of the Department of Pathology, Texas Children’s Hospital, Houston TX. Four of these medulloblastomas did not have enough tissue for immunohistochemical studies while 5 cases (including the four cases) did not have sufficient tissue for molecular studies, leaving a total of 60 cases for immunohistochemistry and 59 cases for molecular studies. The research protocol was approved and carried out in accordance with the guidelines of the Institutional Review Board, University of Oklahoma Health Sciences Center and Institutional Review Board, Baylor College of Medicine, Houston, Texas. Histologic sections from all tumors were reviewed and were classified according to the World Health Organization’s histologic classification (WHO) of medulloblastoma (Table 1 and 2).

Table 1 Result of genomic gains and losses by conventional CGH: [ ] denotes events observed in one hybridization with only a trend in the same direction of deletion or gain in the dye swap
Full size table
Table 2 Result of FOXG1 immunostaining in validating set of medulloblastomas
Full size table

DNA extraction, CGH, post-hybridization scanning and analysis

Genomic DNA was extracted from paraffin embedded tumor tissue using the Qaigen kit (Qiagen, Valencia, CA) and the manufacturer’s protocol. Reference DNA was obtained from paraffin embedded tonsillar tissue of normal male and female, respectively. For chromosomal CGH, the test DNA was chemically labeled in a standard nick translation reaction substituting dTTP by Spectrum Green dUTP (Vysis) while the reference (normal) DNA is labeled substituting dTTP by Spectrum Red dUTP (Vysis). The DNase 1 concentration in the labeling reaction was adjusted to give an average fragment length within 300–2000 bp. The labeled probes were separated from unincorporated nucleotides by Sephadex columns. A 1:1 mixture of test and reference DNA (120–200 ng each) was precipitated in ethanol in the presence of 10 μg of COT-1 DNA (Gibco-BRL, Gaithersburg, MD), dissolved in 10 μl of hybridization solution (50% formamide, 2× SSC, and 10% dextran sulfate), and denatured at 73°C for 5 min. The probe mixture was then applied to target normal metaphase preparations which have been denatured at 73°C for 5 min in 70% deionized formamide in 2× SSC and dehydrated through an ethanol series of 70, 90 and 100% ethanol. Hybridization occurred at 37°C for 4 days. Post-hybridization washes comprise sequential washes with 50% formamide in 2× SSC at 43°C (3× washes for 5 min each) followed by 2× SSC/0.1%NP 40 at 43°C (3× washes for 5 min each). Chromosomes were identified by means of the banding appearance produced by staining with 4,6-diamino-2-phenylindole 2-HCl (DAPI) at 80 ng/ml in an antifade (Citifluor) solution. The fluorescence signals from the hybridization were captured and analyzed using the Applied Imaging Cytogenetic Workstation image analysis equipment and Genus software.

For BAC array CGH, DNA samples for tumor and reference were fragmented by digesting with EcoR1 and concentrated using the Zymo Research’s DNA Clean and ConcentratorTM-5. An aliquot was run on 1% agarose electrophoresis gel to confirm adequate fragmentation. DNA concentration was determined using the Agilent Nanodrop™ Spectrophotometer. 500 ng of reference and tumor DNA were differentially labeled with Cy3-dCTP and Cy5-dCTP by random primer labeling method using the Bioprime labeling kit (In vitrogen). An aliquot was run on 1% agarose electrophoresis gel to ensure synthesis of DNA fragments averaging 100 bp. Test and reference DNA were denatured, mixed and hybridized to BAC array slides (Spectral Genomics). Following 16 hour incubation at 37°C, post-hybridization washes were carried out at 50°C in 2× SSC, 50% Formamide × 20 min, 2X SSC, 0.1% Igepal × 20 min, and 0.2× SSC × 10 min. Two rapid washes (5 s each) in double distilled water were carried out at room temperature followed immediately by blowing the slides dry with a stream of nitrogen gas. Slides were scanned and analyzed using ScanArray (Perkin Elmer) and SpectralWare™ software. The BAC array slide (SpectralchipTM 2600) provides a microarray-based molecular comparative genomic hybridization using 2,632 clones spanning across the genome at one megabase (1 MB) intervals on the average with clones mapped to a cytogenetic linear position.

Quantitative real time polymerase chain reaction (Qrt-PCR):

Copy number of FOXG1 gene in the genomic DNA from each medulloblastoma tumor was determined by Qrt-PCR. PCR reactions containing genomic DNA and SYBR® GREEN PCR Master Mix (Applied Biosystems) and primers for FOXG1 were performed for 30 cycles in triplicate. Primers were designed using the Applied Biosystems primer design program for FOXG1; F-CCTGCTGGCTCAGAAATGC; R-GAGGCGA GGCACTACTTCC and PRDK1 F-TGCTGCAAGTAAAAATGAGAAAGC; R-CATTCAACTAGAA CTCAGTGATTATTG. Amplification products were verified by agarose gel electrophoresis and melting curves. DNA amplification was normalized internally to Albumin gene, relative to control normal human genomic DNA from peripheral blood lymphocytes. We chose to use the Albumin gene as reference control for Qrt-PCR instead of a reference locus on chromosome 14 because (1) we did not want to miss tumors with polysomies for chromosome 14, and (2) polysomies or copy gains/amplification involving the Albumin gene locus at 4q13.3 have not been reported in medulloblastomas.

Immunohistochemistry for FOXG1 and p21cip1

5 μm thick sections were obtained from formalin-fixed paraffin embedded sections of representative sections of the validating set of medulloblastomas. The tissue sections were deparaffinized in Xylene, followed by graded hydration in 100% and 70% ethanol to H2O. Antigen retrieval was done by boiling in 10 mM EDTA × 25 min. Endogenous peroxidase was blocked with 3% H202/Methanol × 15 min, then incubated with 20% goat serum × 20 min. Sections were incubated with 1:20 dilution of polyclonal anti-FOXG1 antibody (Abcam Inc, Cambridge, MA) × 1 h at room temperature or anti-p21cip1 antibody overnight at 4°C (DAKO). This was followed by incubation with anti-rabbit secondary antibody conjugated with HRP × 30 min. DAB was used as color reagent. The secondary antibody was conjugated with FITC for immunofluorescence staining. Fetal kidney showing nuclear positivity in the proximal and distal renal tubules served as control. The immunostained sections were graded semi-quantitatively for intensity (I) (0, negative; or 1 to 3+) as well as extent (E) (0%: 0, ≤25%: 1+, 26–50%: 2+, 51–75%: 3+, >75%: 4+) of staining.

Transfection with FOXG1 specific siRNA

DAOY, a medulloblastoma cell line was obtained from the American Type Culture Collection (ATCC, Manassas, VA) and propagated in Dulbeco’s Minimum Essential Medium (DMEM) supplemented with 10% fetal calf serum. Cells were transfected with FOXG1 specific siRNA using siPortamine and recommended protocol (Catalog # 115630, Applied biosystems). GAPDH specific siRNA and Luciferase gene specific siRNA served as positive and negative transfection controls, respectively.

MTT proliferation assay

MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay is based on the ability of a mitochondrial dehydrogenase enzyme from viable cells to cleave the tetrazolium rings of the pale yellow MTT and form a dark blue formazan crystal which is largely impermeable to cell membranes, thus resulting in its accumulation within healthy cells. Solubilization of the cells by the addition of a detergent results in the liberation of the crystals which are solubilized. The number of surviving cells is directly proportional to the level of the formazan product created. The color was quantified using a simple colorimetric assay. The results were read on a multi-well scanning spectrophotometer (ELISA reader). The MTT Cell Proliferation Assay (ATCC, Manassas, VA) measures the cell proliferation rate and conversely, when metabolic events lead to apoptosis or necrosis, the reduction in cell viability. The MTT reagent yields low background absorbance values in the absence of cells. For each cell type the linear relationship between cell number and signal produced is established, thus allowing an accurate quantification of changes in the rate of cell proliferation. 10 μl of MTT Reagent is added to each well (96 well plate), including controls. Plate is returned to cell culture incubator for 2–4 h. Periodically the cells are viewed under an inverted microscope for presence of intracellular punctate purple precipitate. When the purple precipitate is clearly visible under the microscope, 100 μl of detergent reagent is added to all wells, including controls. The plate is gently swirled without shaking. Plate is left with cover in the dark for 2–4 h or overnight at room temperature. Plate cover is removed and the absorbance in each well is measured, including the blanks, at 570 nm in a micro-titer plate reader. Average values are determined from triplicate readings from which are subtracted the average value for the blank. The number of cells is determined from a previous plot of absorbance against number of cells/ml.

Results

CGH analysis

The chromosomal identification in each tumor karyotype was verified. The pattern of chromosomal gains and losses from the cCGH analysis of the initial 19 medulloblastoma tumors were automatically generated by the Applied imaging software with a cut off value of test to reference ratio of 0.8 for a deletion event and 1.2 for a copy gain event. The ratios for each chromosomal region in a tumor represents an average of about 20–30 chromosomes from 10 to 15 metaphase spreads. The events were accepted as definite when present in the initial run and in the dye swaps. Events above 1.2 or below 0.8 in the initial run but with a strong tendency towards the cut off value in the dye swap were identified as possible but not definite events. A summary of the pattern of losses and gains in the tumor set is shown in Table 1. The commonest deletion event involved 17p and was present in 5 out of 19 (26.3%) tumors. Less frequent deletions were seen at 16q, 10q, 11p, 11q and 13q. 6 out of 19 tumors (31.6%) did not show any detectable deletion. 4 out of the six tumors with no deletions were desmoplastic tumors while the other two tumors were classic tumors. A regression analysis of the frequency of deletions versus survival showed a negative correlation with P = 0.02 and r 2 = 0.3496. The commonest gain involved 17q and was seen in 9 out of 19 (47.4%) tumors. Less frequent gains were seen at 1p, 1q, 2p, 2q and 14q. 2p gain was present in 3 out of 19 tumors and the region of gain encompassed variable segments from 2p16 to 2p25 and in these three tumors included the n-myc locus.

Nine (9) out of nineteen (19) tumors with frequent genetic events by cCGH analysis were further analyzed by aCGH. aCGH is a more sensitive technique for genome wide scanning and for the detection of deletion or gain events in the genome. The scanned images of hybridized arrays were analyzed using Spectral Genomics software. This software normalizes the scanned image data using a “global” method [26] of normalization and identifies a probe locus as showing a gain or deletion if the ratio of the tumor DNA to the reference control normal DNA is greater than 1.2 or less than 0.8 in both the initial run and dye swaps for each tumor. Comparable analysis using Lowess’ method [27] of normalization showed values similar to those of the global method. Probe spots where the initial run and the dye swap show divergent values such that the initial run shows a ratio above 1.2 while the dye swap shows a ratio less than 0.8 were rejected. The log2 ratios for each probe spot/tumor were also analyzed using the Agilent CGH analytics 3.1 software. Analysis of copy gains and losses revealed multiple loci in the genome of medulloblastoma tumors with low levels of copy gain and deletions. The composite of genetic events identified from the analysis of these nine medulloblastomas by aCGH is shown in Fig. 1. The most frequent sites of genomic gains were 14q (67%), 17q (33%), 7 (33%) and 12q (33%). Less frequent gains were seen at 5q (22%) and 2p (22%). The most frequent genomic deletions were at 16q (56%), 17p (44%) and 5q (33%). Less frequent deletions were at 10q (22%), 8p (22%) and 13q (22%).

Fig. 1
figure 1

Summary composite ideogram for BAC array CGH analysis: Nine [9] test medulloblastomas showing the frequency of distribution of genetic changes including regions of genomic deletions (red) and gains (blue). Frequent deletions for 17p and 16q, frequent gains for 14q12, and occasional trisomies for chromosomes 5, 7, 14 and 17q are present

Full size image

Quantitative real time PCR (Qrt-PCR)

For validation of observed gains and deletions from aCGH, the genomic DNA of the tumors was subjected to Qrt-PCR. To determine which genes are potentially affected at 14q12 (the locus with the most frequent gain by aCGH), the BAC probe at the 14q12 locus was identified as RP11-125A5. Potential involved genes localized to the 14q12 chromosomal region with overlap or proximity to the BAC—RP11-125A5 were identified using the UCSC browser (www.genome.ucsc.edu). The two genes closest to the gained BAC probe (Fig. 2A) which mapped to this region, FOXG1 and PRKD1 (PRKCM) were subjected to Qrt-PCR. Copy gain of FOXG1, a gene mapping centromeric to the gained BAC probe was evident in 9 out of 9 tumors (Fig. 2B). On the other hand, allelic deletion for PRKD1, a gene mapping telomeric to the gained BAC probe was seen in 7 out of the 9 tumors. Values represent triplicate runs for each tumor and were confirmed in repeat runs of the Qrt-PCR.

Fig. 2
figure 2

(A) Ideogram of the 14q12 locus (RP11–125A5): It shows high frequency of gain (6/9) at this locus. Chromosome 14 linear position 26.5 M to 30.27 M, approximately [enclosed in the blue rectangle (broken lines)] is enlarged to the right of the ideogram and shows the localization of the RP11–125A5 BAC (arrow). The genes closest to this BAC in this “zoomed in” image based on the human genome assembly 17 are centromeric FOXG1 and telomeric PRKD1 (PRKCM) (arrow heads) (B) Qrt-PCR on genomic DNA for FOXG1: Nine test medulloblastoma tumors show 2 to 7 fold gain in FOXG1 gene copy number in all tumors

Full size image

To validate the frequency of copy gain involving FOXG1 in medulloblastoma, a total of 59 out of a validating set of 60 medulloblastomas had sufficient genomic DNA available to be used as a validating subset of tumors. They were subjected to Qrt-PCR under the same conditions as described. 93% (55/59) of these medulloblastomas were found to have multiple copies (2–21 fold) of the FOXG1 gene. 64.4% (38 out of 59) of the tumors with multiple copies had ≥ 5 copies consistent with gene amplification (Fig. 3), thus representing a frequent genetic event in this subset of medulloblastoma.

Fig. 3
figure 3

Qrt-PCR on genomic DNA of validating medulloblastomas: 59 out of 60 tumors in the validating set of medulloblastomas show a 2–21 fold gain in FOXG1 gene copy number in 93% (55/59) of tumors. 64.4% (38/59) of tumors have ≥ 5 fold gain in gene copy number. Human normal peripheral blood lymphocyte genomic DNA (YN) served as control. Values represent the mean of triplicate runs for each tumor. Repeat triplicate runs confirm result

Full size image

Immunohistochemical analysis for FOXG1

Expression of FOXG1 was assessed by immunohistochemistry in normal cerebellum and medulloblastoma. In normal cerebellum, FOXG1 expression was absent in the nuclei of the internal granular neurons and Purkinje cells (Fig. 4F). Fetal kidney was noted to have a robust expression of FOXG1 in the developing proximal and distal renal tubules and it served as the immunohistochemistry positive control tissue. In medulloblastomas, FOXG1 protein was detected in the nuclei of 59 of 60 validating tumors. Since FOXG1 is not expressed in normal cerebellum, its expression in medulloblastoma represents aberrant expression. FOXG1 protein expression was demonstrated as follows: mild (+) in 20/60 (33%); moderate to marked (++/+++) in 39/60 (65%); 1 tumor (2%) was negative. The positively staining nuclei were widely distributed in the tumors (Fig. 4A–C and E). FOXG1 protein was detectable in desmoplastic (nodular), classic, as well as, anaplastic / large cell histologic subtypes (Table 2). An inverse relationship was noted between the expression of FOXG1 and p21cip1 (Fig. 4C and D) consistent with a negative regulation of p21cip1 expression in most of these tumors. FOXG1 copy gain (>2 to 21 folds) showed a significant positive correlation with protein expression (Spearman’s rank order correlation coefficient = 0.276; P = 0.038).

Fig. 4
figure 4

Immunostaining for FOXG1 and p21cip1 in medullblastomas: Medulloblastomas show variable diffuse and intense nuclear positivity for FOXG1 (A, C, E), a rare perinuclear dot-like pattern of staining for FOXG1 (B) and negative nuclear staining for p21cip1 (D). Figures C and D are from the same tumor. Mature cerebellum shows no demonstrable nuclear positivity for FOXG1 in internal granular neurons or Purkinje cells (F) [A–D and F × 200; E × 400].

Full size image

In vitro modulation of FOXG1 effect using FOXG1 specific siRNA

The medulloblastoma cell line, DAOY was used for in vitro experiments. Copy gain of FOXG1 gene and expression were determined by Qrt-PCR and Qrt-RT-PCR respectively. A five fold copy of the FOXG1 gene and 2 × 103 fold upregulation of the expression of the FOXG1 mRNA was observed (Fig. 5A). A two-fold copy gain of the FOXG1 gene and a 4 × 103 fold upregulation of the expression of the FOXG1 mRNA was also observed in D283-Med. Following treatment of DAOY with FOXG1 specific siRNA (Catalog # 115630, Applied biosystems), a 65–70% “knockdown” of FOXG1 mRNA expression (Fig. 5B) was accompanied by a 23% reduction in proliferation rate (Fig. 5C). Other FOXG1 specific siRNA (Catalog # 6439 and 145145, Applied biosystems) showed similar FOXG1 knockdown efficiency (data not shown). The presence of 0.1 ng/ml of TGF-β augmented the reduction in proliferation index to 39%. Immunofluorescence staining with p21cip1 specific antibody shows upregulation of p21cip1 protein in DAOY cells treated with FOXG1 specific siRNA (Fig. 5D). The upregulation of p21cip1 seen in the DAOY cells examined by immunofluorescence is variable and reflects variable efficiency of transfection of individual cells with the siRNA. Using Qrt-RT-PCR, a 2-fold upregulation of p21cip1 mRNA following siRNA transfection was demonstrated (Fig. 5B).

Fig. 5
figure 5

(A) Qrt-RT-PCR for FOXG1 in medulloblastoma cell lines: High levels of expression of FOXG1 are seen in DAOY (∼2 × 103 fold) and D283Med (∼4 × 103 fold) when compared with normal cerebellum. Levels are comparable to those seen in Fetal Brain (FB) and Normal adult cerebral cortex (NB). (B) Transfection of DAOY cells with FOXG1 specific siRNA: It shows a 65–70% inhibition of FOXG1 mRNA and a 2 fold upregulation of p21cip1 expression. Cells transfected with nonspecific Luciferase gene siRNA served as negative controls and do not show any evidence of “knockdown”of FOXG1 mRNA or GAPDH mRNA. Cells transfected with siRNA to GAPDH served as positive controls for the transfection protocol and show a more efficient “knockdown ‘ of GAPDH mRNA when compared with FOXG1 mRNA “knockdown” by FOXG1 specific siRNA. This is consistent with variable efficiency of transfection of siRNA for different genes. (C) MTT assay for cell proliferation: There was a 23% decrease in proliferation index (P < 0.05) which further decreased by 39% in the presence 0.1 ng/ml of TGF-β. Note paradoxical increase in proliferation index in the absence of FOXG1 inhibition (P < 0.05). (D) Immunofluorescence for p21cip1 following siRNA treatment of DAOY: Cells transfected with (A) Luciferase gene specific siRNA as negative control, do not show demonstrable increase in expression of p21cip1, while in (B) FOXG1 specific siRNA transfection shows upregulation of expression of p21cip1. The variable expression between cells is consistent with variable transduction in individual cells [all × 400]

Full size image

Discussion

This report documents for the first time, the dysregulation of FOXG1 in medulloblastoma with frequent gene copy gain and aberrant protein expression in all histologic types of medulloblastoma and in 93% of our cases of medulloblastoma. It suggests that dysregulation of FOXG1 is a significant and possibly early event in the development of medulloblastoma. We hypothesize that overexpression or aberrant expression of FOXG1 in the precursor stem cells of medulloblastoma would maintain these cells in the undifferentiated state and also prevent their response to normal programmed cell death, thereby setting the stage for additional progression related genetic events associated with neoplastic transformation. FOXG1 is known for its role in neurodevelopment [28–30] and has been shown to reduce the responsiveness of cells in vitro to growth inhibition by TGFβ. Members of the TGF-β superfamily including TGF-β1, β2 and BMP2, 4 and 7 also play significant roles during the development of the CNS. TGF-β2 and BDNF have been implicated in the maintenance of an antiproliferative microenvironment in the normal cerebellum [31]. TGF-β1 has been shown to selectively increase the expression of p21cip1 in the ventricular zone (VZ). A decrease in the fraction of ventricular zone (VZ)-cycling cells by 21% was found to be associated with an increase in the number of VZ cells exiting the cell cycle by 24%. In addition, high p21cip1 expression levels were observed in VZ cells as they exited the cell cycle [32]. Beta catenin expression is associated with sequential expression of BMPs 2, 4, 7 resulting in neuronal development as well as gliogenesis [33]. A recent study has reported better survival in medulloblastomas with activated (mutated) β catenin [34]. This observation was unexpected and is counterintuitive. Whether this observation in medulloblastomas is due to the antiproliferative effects of BMPs induced by β catenin remains to be determined. Paracrine TGF-β2 mediated activation of the Smads 2 and 4 also have anti-proliferative effects on the mature cerebellum by upregulating the expression of p21cip1 and p27kip1 in the cerebellum. In addition, cerebellum-derived BDNF via ERK1/2 signaling augments TGF-β2 synthesis / secretion into the cerebellar microenvironment, thereby enhancing the anti-proliferative effects of TGF-β2 in the cerebellum [31].

FOXG1 expression has been shown to be sufficient to overcome the ability of TGF-β to block cells released from contact inhibition from re-entry into the cell cycle. These effects are mediated through the binding of FOXG1 with smad-FOXO transcriptional complexes [35, 36]. In vivo studies suggest that FOXG1 activity may not be primarily directed at inducing cell proliferation but may play a role in maintaining the undifferentiated state in order to properly time neurogenesis and allow the progenitor population sufficient time to expand, thereby delaying early cortical cell fate [37]. In essence, FOXG1 is a transcriptional repressor that protects neuroepithelial progenitor cells from cytostatic and differentiation inducing signals [38, 39]. FOXG1 expression is restricted to the neuroepithelial progenitor population that comprises the telencephalon in the developing mouse [40], and it is essential for forebrain formation [38]. Furthermore, excess FOXG1 expression in vivo is associated with neural progenitor cell overgrowth; an effect requiring DNA-binding and repressor activity [41]. FOXG1 is expressed at very low levels in normal cerebellum. We are not able to detect FOXG1 in the differentiated cerebellum. A comparison of medulloblastoma cell lines DAOY and D283-Med with differentiated normal cerebellum shows upregulation of FOXG1 expression to about 2 × 103 and 4 × 103 folds, respectively. These high levels (compared with normal cerebellum) in these medulloblastoma cell lines suggest that FOXG1 expression in cerebellum derived medulloblastoma tumors and cell lines is an aberrant expression with possible role in neoplastic transformation. Available reports suggest that rather than directly affecting cell proliferation, mechanisms for FOXG1 mediated overgrowth involve counteracting signaling induced by cytostatic factors of the TGF-β superfamily, including TGF-β and BMP. This is achieved through the repression of the transcription of cyclin dependent kinase inhibitors p15Ink4b/p21cip1 and by reducing the frequency of normal programmed cell death or apoptosis [25, 35, 41]. An alternative mechanism for opposing TGF-β induced signaling involves hyperactive PI3K pathway that may also inhibit TGF-β induced cytostasis by activating an akt-mediated phosphorylation of the FOXO1, 3 and 4 proteins with an exclusion of the FOXO proteins from the nucleus, thereby disrupting the Smad-FOXO transcriptional complex formation (Fig. 6) [39].

Fig. 6
figure 6

Pathway interaction for FOXG1: Schematic illustrating the reported role of FOXG1 in inhibiting the transcriptional activity of TGFβ induced FOXO/Smad3/4 complex, resulting in the down regulation of the transcription of p21cip1 which among other functions inhibits G1 to S phase transit. FOXG1 also forms a transcriptional repressor complex with Hes1 (transcription upregulated by activated Notch signaling) resulting in the inhibition of the transcription of proneural genes thus suggesting that aberrant upregulation of FOXG1 may partially account for persistence of an undifferentiated component in medulloblastomas

Full size image

An immunohistochemical study of medulloblastomas shows increased expression of insulin growth factor-1 (IGF-1) and downstream proteins of the PI3K-akt pathway in medulloblastoma [42]. Overexpression of IGF-2 has also been shown to enhance tumor development in the ptch mutation mice. In this study, Shh-induced tumor formation was enhanced by coexpression with IGF2 and akt with neither IGF2 nor Akt causing tumors when expressed independently. Interestingly, the induced tumors showed upregulated expression of insulin receptor substrate 1 and phosphorylated forms of IGF1 receptor and Akt, a finding consistent with the activation of IGF signaling [43]. This accumulating evidence would indicate that FOXG1 expression and PI3K pathway activation may independently or in combination mediate resistance to TGF-β induced cytostasis. It is reasonable to speculate that combined IGF signaling induced phosphorylation of FOXO via akt and aberrant FOXG1 upregulation may have additive effects in the opposition of TGF-β induced cytostasis in medulloblastoma cell lines and tumors.

Current studies implicate Notch2-mediated signaling in the growth of medulloblastoma [44]. In addition, shh signaling implicated in the growth of medulloblastoma has been demonstrated to mediate some of its biologic effects through the activation of Notch2-mediated signaling [45, 46]. It is also noteworthy that FOXG1 has been implicated in the repression of neuronal differentiation by forming transcriptional repressor complexes with Groucho/TLE proteins, histone deacetylases and Hes1 to effect the Notch activation-mediated negative regulation of proneural differentiation genes [47]. These observations suggest a major regulatory role for FOXG1 at the critical points of transcriptional regulation by the PI3K-akt, Notch, Shh via Notch2 and TGF-β signaling pathways.

We hypothesize that the role of FOXG1 in medulloblastoma pathogenesis is (i) to maintain cerebellar neural stem-like cells in an undifferentiated state by forming transcriptional repressor complexes with effector proteins of the Notch signaling pathway such as Hes1 that negatively regulate the expression of proneural genes, and (ii) to repress TGF-β induced cytostasis, thereby allowing the expansion of the progenitor population as occurs during neurogenesis [30]. These possible interactions between FOXG1 and the Notch pathway in medulloblastoma are currently being explored to determine if they do occur in medulloblastomas as have been observed during neurogenesis. If such interactions are found in medulloblastoma, they will be consistent with a role for excess FOXG1 expression secondary to gene copy gain/amplification in favoring progenitor cell overgrowth and providing the milieu for the accumulation of additional transforming genetic events and mutations. Thereafter, one may wonder if FOXG1 fits the definition of a “neoplasia susceptibility gene”. This determination awaits future studies.