Skip to main content

Identification of the genetic and/or epigenetic events that lead to mesenchymal transformation in glioblastoma

Final Report Summary - GAMNG (Identification of the genetic and/or epigenetic events that lead to mesenchymal transformation in glioblastoma)

Final publishable summary

Dr. Carro Maria Stella
Group leader
University of Freiburg, Department of Neurosurgery
Breisacherstaße 64 79106 Freiburg, Germany
Phone: 0049-0761-27054400
Fax: 0049-0761-27054470
Email: maria.carro@uniklinik-freiburg.de


Glioblastoma multiforme (GBM) is the most common and malignant type of brain tumor in adults and is nearly uniformly fatal, with a median survival of one year (Ohgaki and Kleihues, 2005). Previously, we identified two transcription factors, Stat3 and CEBPB, as master regulators of the mesenchymal gene signature (MGS) in gliomas which was previously shown to co-segregate with the poor-prognosis group (Carro et al., 2010) (Phillips et al., 2006). Recently, loss of NF1 has been linked with the mesenchymal tumor subtype (Verhaak et al., 2010). Moreover, Phillips et al (2006) highlighted EGFR and PTEN as potential oncogene and tumor suppressor genes associated with MGS respectively.
Our objective for this project was to specifically test the causal effect of NF1, EGFR, and PTEN on the MGS expression, as well as on tumor formation and invasion. In addition, we proposed to search for mutations in the gene promoter of STAT3 and CEBPB genes by performing sequencing analysis in normal and tumor samples. Moreover, we planned to measure changes in the methylation status of STAT3 and CEBPB promoters by pyrosequencing in a set of normal and tumor samples.
To get an insight into the possible effect of EGFR, PTEN, and NF1 copy number change on the expression of Stat3 and CEBPB, we performed linear regression analysis of Stat3 or CEBPB expression on EGFR, PTEN, and NF1 expression in a set of 188 GBM data available through the TCGA data portal (http://cancergenome.nih.gov/). We observed no relationship between Stat3/ CEBPB and EGFR/PTEN (data not shown). However, we observed that loss of NF1 expression is associated with increased CEBPB expression (p value = 4.3 x 10-14) (Fig.1). Western blot analysis in primary cancer stem cells (CSCs) confirmed that NF1 expression was inversely correlated to CEBPB (data not shown). This suggests that CEBPB but not Stat3 expression in GBM could be regulated by NF1 signaling.
In order to characterize the role of NF1 in CEBPB regulation and to understand whether deregulation of CEBPB expression is sufficient to affect the expression of mesenchymal genes, we performed silencing of NF1 in two CSCs (145s and 66s). Silencing of NF1 induced CEBPB but not Stat3 expression (Fig.2). Overexpression of the NF1 GTPase-activating domain (NF1-GRD, aminoacids 1131-1534) spanning the whole predicted Ras-GAP domain (McCormick, 1990) in a mesenchymal cell line (233s) led to a decrease in CEBPB mRNA and protein expression (data not shown). qPCR and IF analysis also revealed a significant decrease of the mesenchymal markers YKL40 and CD44 upon NF1-GRD overexpression (data not shown).
We then investigated whether the perturbation of NF1 expression was sufficient to induce changes in the mesenchymal gene signature. Microarray analysis followed by Gene Set Enrichment Analysis (GSEA) of eight brain tumor samples with NF1 loss or NF1 wild type status revealed a strong enrichment of MGS in NF1 deleted samples (FDR q-value = 0.04 data not shown). Interestingly, overexpression of NF1-GRD in mesenchymal cells CSC233s led to a reduction of the mesenchymal signature (enrichment score = -0.5; FDR q-value = 0.1 Fig.3). Conversely, we observed an enrichment of the MGS upon NF1 knockdown in proneural cells CSC 3021, (enrichment score = 0.5; FDR q-value = 0.4 Fig.3).
Taken together, our data suggested that NF1 modification is able to alter the MGS expression in GBM. NF1-led gene expression changes might be driven by its effect on CEBPB alone; however, as our earlier data demonstrated that complete reprogramming towards a mesenchymal phenotype requires both CEBPB and STAT3 (Carro et al, 2010), an additional transcription factor (TF) might be needed to induce a mesenchymal phenotype. We therefore screened the microarray data for TFs altered in the NF1 deletion, overexpression, and knockdown experiments. TFs that were regulated in at least two matching experiments were further validated by qPCR. Interestingly, the ASCL1 TF, which is known to be expressed in gliomas WHO°II and III and °IV of proneural subtype and secondary GBMs (Somasundaram et al., 2005), (Castro et al., 2011), was downregulated upon NF1 silencing. Therefore, ASCL1 could mediate NF1 signaling in lower grade and proneural GBMs. In the future, we plan to further investigate this mechanism and the possible interaction with CEBPB function.
It was previously shown that mesenchymal glioblastoma cells are able to differentiate into osteocytes (Ricci-Vitiani et al., 2008), a feature they share with mesenchymal stem cells. Following this, we observed a decrease in osteogenic differentiation measured by Alizarin Red staining upon overexpression of NF1-GRD in mesenchymal 233 cells (Fig. 4). NF1-GRD overexpression in CSCs233 also led to decreased cell invasion when tested in a matrigel invasion assay (data not shown).
Overall, our data indicates that decreased expression of NF1, most likely as a consequence of point mutations and deletion (Verhaak et al., 2010), is associated with increased expression of CEBPB in GBMs. Since NF1 functions as a negative regulator of the Ras pathway, CEBPB expression might occur through activation of Ras signaling. In addition, the identification of ASCL1 as an additional TF regulated by NF1 suggests that other TFs could play a role in the regulation of the mesenchymal signature. A better understanding of the mechanisms by which NF1 regulates CEBPB and other TFs will help to identify specific targets for GBM tumors characterized by low NF1 and high CEBPB expression.