Epigenetic regulation of glial fibrillary acidic protein by DNA methylation in human malignant gliomas

Abstract

Glial fibrillary acidic protein (GFAP) is an intermediate filament expressed in glial cells that stabilizes and maintains the cytoskeleton of normal astrocytes. In glial tumors, GFAP expression is frequently lost with increasing grade of malignancy, suggesting that GFAP is important for maintaining glial cell morphology or regulating astrocytoma cell growth. Most permanent human glioma cell lines are GFAP negative by immunocytochemistry. Given that the GFAP gene is not mutated in human glioma specimens or glioma cell lines, we considered epigenetic mechanisms, such as promoter methylation, as a cause of silencing of GFAP in these tumors. In this study, we treated known GFAP-negative glioma cell lines with 5-aza-2′-deoxycytidine to examine GFAP promoter hypermethylation. Additionally, we performed bisulfite sequencing on primary glioma samples and glioma cell lines and showed an inverse relationship between GFAP promoter methylation status and GFAP expression. Using a gene reporter assay with the GFAP promoter cloned upstream of a luciferase gene, we showed that methylation of the GFAP promoter downregulates the expression of the luciferase gene. Our results suggest that epigenetic silencing of the GFAP gene through DNA methylation of its promoter region may be one mechanism by which GFAP is downregulated in human gliomas and glioma cell lines.

Human malignant gliomas are the most common primary central nervous system neoplasm accounting for 65% of all primary brain tumors. Malignant gliomas arise from a number of genetic alterations, which result in the activation of oncogenes and inactivation of tumor suppressor genes that affect primarily cell cycle and growth factor activation pathways.¹ These genetic aberrations lead to uncontrolled cellular proliferation, deregulation of apoptosis, increased invasiveness, and tumor angiogenesis—all of which promote the neoplastic phenotype.^1,2

Glial fibrillary acidic protein (GFAP) is an intermediate filament that is expressed primarily, but not exclusively, by cells of glial origin. Although its function is poorly understood, GFAP is thought to provide structural support and tensile strength to the cytoskeleton of normal astrocytes. In neuropathological preparations, GFAP is frequently used as a reliable marker of astrocytes and tumors of glial origin. Interestingly, previous studies have demonstrated that with increasing grade of astrocytoma, there is a progressive loss of GFAP expression.^3–6

The gene for GFAP is localized to human chromosome 17q21. Mutations in the GFAP gene have been identified in a few disease states, such as Alexander's disease,⁷ and recently in glioma-like tumors in some Alexander's disease patients.^8,9 The mature transcript of GFAP yields a 50-kD intracytoplasmic filamentous protein that shares considerable structural homology with other intermediate filaments in the central α-helical or rod domain.¹⁰ The unique NH2-terminal region of GFAP, when compared with other intermediate filaments, possesses 4 amino acid residues that undergo phosphorylation. The phosphorylation of GFAP by kinases such as CaM kinase II, protein kinase A, cdc2, protein kinase C, and rho kinase, leads to the disassembly of already assembled glial filaments.

The GFAP gene promoter region spans the 2-kb region upstream from the initiation start codon. In addition to the basal promoter sequences such as the TATA box, the GFAP promoter has both positive and negative regulatory elements. The regions between −250 and −80 bp and between −1980 and −1500 bp contain positive regulatory regions, whereas the sequence between −650 and −360 bp harbors a negative regulatory element.¹¹ Interestingly, a recent study has elucidated that the region conferring astrocyte-specific expression is located between −1488 and −133.¹²

Epigenetic mechanisms such as DNA methylation can lead to downregulation of gene expression and gene silencing. CpG islands are genomic regions typically present in gene promoters rich in GC dinucleotide content, which are targets for DNA methylation and subsequently inactivation of gene transcription. DNA demethylation and reactivation of transcription can be accomplished following the administration of 5-aza-2′-deoxycytidine (5-aza-dC).¹³ The exact mechanism of action of 5-aza-dC is not completely understood; however, it is believed to promote DNA demethylation and restoration of gene expression by relaxing the chromatin structure. The resulting chromatin remodeling allows transcription factors to bind to the promoter regions, assembly of the transcription machinery, and gene expression. Accordingly, in this study, we examined whether epigenetic silencing of the GFAP gene through DNA methylation could play a role in the loss of GFAP expression in human gliomas.

Methods and Materials

Reagents, Antibodies, Astrocytoma Cell Lines, and Astrocytoma Samples

The permanent human astrocytoma cell lines, U251 MG, U87 MG, U118 MG, U138 MG, A172, and T98, and the HeLa cell line were cultured in high-glucose Dulbecco's minimal essential medium with 10% fetal bovine serum. All cell lines have been well characterized previously. Tissue specimens were obtained from 10 adult World Health Organization grade IV gliomas (glioblastoma multiforme [GBM]), and 1 non-neoplastic temporal lobe brain tissue specimen obtained following craniotomy for temporal lobectomy. Approval to use these materials was granted by the Research Ethics Board, the Hospital for Sick Children.

The primary GFAP antibody used for both immunocytochemistry and immunohistochemistry was rabbit anti-GFAP (DAKO). A modified phospholipase C lysis buffer used for the Dual-Glo luciferase assay system was prepared as follows: 50 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), 150 mM NaCl, 10% glycerol, 1% Triton-X 100, 1 mM EDTA, 100 mM NaF, and 10 mM NaPPi. Primers were designed with the restriction sites HindIII and BglII (Roche) at the 5′ flanking regions of the 2-kb upstream promoter of GFAP for directional cloning into the firefly luciferase vector pGL4.10[luc2] (Promega).

5-aza-dC Treatment Protocol and Real-Time PCR

In order to examine the endogenous GFAP expression in glioma cell lines, total RNA was extracted from the U251, U87, U118, U138, A172, and T98 glioma cell lines. U251 was previously shown to be GFAP positive and therefore was included in this analysis as a control.¹⁴ The HeLa cell line was used as a negative control.¹⁵

Cell lines were plated at 20%–30% confluence. Twenty-four hours later, the medium was replaced with a fresh medium containing 5 µmol/L 5-aza-dC (Sigma-Aldrich) or an equal volume of vehicle (phosphate buffered saline [PBS]). The medium and drug or vehicle were replaced every 24 hours over a 72-hour period. RNA was extracted using the RNeasy Mini Kit (Qiagen, Inc.) and quantitated spectrophotometrically by NanoDrop ND-1000 (NanoDrop). cDNA was prepared using 2 µg of total RNA, random hexamer primers, and the Omniscript RT kit (Qiagen). Reverse transcription–PCR (RT–PCR) was performed using the platinum Taq DNA polymerase (Invitrogen) in an MJ Research PTC-200 thermal cycler (Bio-Rad). Quantitative real-time RT–PCR (qRT–PCR) was performed using Platinum SYBRGreen Supermix (Invitrogen), in a StepOnePlus Real-Time PCR System (Applied Biosystems). Primer sequences for RT–PCR and qRT–PCR were as follows: GFAP-F, 5′-agaagctccaggatgaaacc-3′; GFAP-R, 5′-agcgactcaatcttcctctc-3′. Expression levels for the ACTB gene were used for normalization of relative gene expression levels.

Immunofluorescence

Following 5-aza-dC treatment for 72 hours as described previously, astrocytoma cells were grown on 10-mm glass coverslips to 70% confluency. Coverslips were washed in 1 × PBS, and cells were prepared for immunostaining by fixing in 4% paraformaldehyde for 30 minutes followed by permeabilization with 0.2% Triton-X 100 for 10 minutes. Cells were blocked using 5% bovine serum albumin (Bioshop) for 30 minutes prior to incubation with a primary antibody. Detection of the primary antibody was performed using goat antirabbit Alexa fluor-488 antibody (Invitrogen). Cell line immunofluorescence images were captured using a Zeiss Axiovert 200 inverted microscope (Carl Zeiss), equipped with a Hamamatsu Orca AG CCD camera and spinning disc confocal scan head (Quorum Technologies). Image acquisition was performed using Volocity 5.1 software (Improvision). Quantification of GFAP expression in each of the untreated and 5-aza-dC–treated cell lines was determined by an enumerating positive signal in cells in 5 random fields of view.

Bisulfite Genomic Sequencing

Genomic DNA was subjected to bisulfite conversion using the MethylEasy Exceed DNA Bisulfite Modification Kit (Human Genetic Signatures). After bisulfite conversion, the GFAP promoter region was amplified using bisulfite-PCR oligonucleotides. Primer sequences for GFAP bisulfite PCR were as follows: Fwd 5′- TTTATTTTTTTAGGTTGATGTGTGG-3′, Rev 5′-CCCTTCCTTATCTAACCTCCCTATA-3′. PCR products were gel purified, TA cloned into the pCR2.1 vector (Topo TA Cloning Kit, Invitrogen), transformed into TOP10 chemically competent cells, and plated under antibiotic selection. Plasmid DNA from isolated colonies was extracted by miniprep (Qiagen) and sequenced to determine GFAP methylation status. Multiple clones were sequenced, providing a consensus of the promoter-region methylation status. Bisulfite sequencing data were analyzed using BiQ Analyzer software.¹⁶

Luciferase Assay

Luciferase assays were performed using a Dual-Glo luciferase assay system according to the manufacturer's instructions (Promega). Briefly, the 2-kb GFAP promoter was directionally cloned into pGL4.10 (Promega) using the PCR primers Fwd 5′-GCGAGATCTGAGGCTGATGTGGAAGGATC-3′ and Rev 5′- GCGAAGCTTCCAGACGGCGGCCAGGAG-3′. The GFAP promoter was then cut from the GFAP-pGL4.10 vector using restriction enzymes Bgl2 and HindII (Roche). The 2-kb insert was then treated with either M.SssI (NEB) for methylation of CpG sites or no enzyme for a positive control. Insert and backbone vectors were re-ligated using T4 ligase (NEB) overnight at room temperature. Plasmids were run on a 1% agarose gel and gel was extracted using the QIAEX II Gel Extraction Kit (Qiagen). Two micrograms of both methylated and unmethylated GFAP-pGL4.10 plasmids were transfected into U251 MG cells. These constructs were cotransfected with the Renilla luciferase pRL-SV40 vector (Promega) as a normalization and transfection efficiency control. As a positive control for promoter activity, we used the luciferase vector (pGL4) that was under the control of the cytomegalovirus (CMV) promoter. Cells were then lysed 24 hours post-transfection; firefly luciferase levels were detected by adding Dual-Glo luciferase substrate and measured using a Lumat LB 9507 luminometer (Berthold Technologies). Renilla luciferase activities were then assayed by adding an equal volume of Dual-Glo Stop & Glo substrate (comprising the stop solution for firefly luciferase and substrate for renilla luciferase) and remeasuring in the luminometer.

Gene Databank Query

The Catalogue Of Somatic Mutations in Cancer database at the Sanger Institute (http://www.sanger.ac.uk/genetics/CGP/cosmic/) and the Cancer Genome Atlas database (http://cancergenome.nih.gov) were queried for mutations or deletions on chromosome 17q21 for the GFAP gene locus in human malignant glioma.

Results

GFAP Expression in Glioma Cell Lines

Immunocytochemistry and real-time PCR expression analysis revealed very low expression of GFAP in all glioma cell lines, except U251 MG, which was comparable to the expression found in HeLa, the negative control cell line (Fig. 1). As expected, GFAP expression in the GFAP-positive U251 MG cell line was 3 orders of magnitude greater than that observed in the other glioma cell lines.

5-aza-dC Treatment of GFAP-Negative Glioma Cells Restores GFAP Expression

Given that there are no reported genomic aberrations of the GFAP gene locus in human gliomas, the loss of GFAP expression in glioma specimens and glioma cell lines could be the result of epigenetic silencing through DNA promoter methylation. 5-aza-dC has been used extensively to pharmacologically unmask epigenetically silenced genes.¹³ Therefore, we treated the GFAP-negative glioma cell lines with 5-aza-dC, extracted total RNA as described above, and analyzed GFAP expression by a real-time PCR. Using the real-time PCR, we detected an increase in GFAP transcript expression following 5-aza-dC treatment in all GFAP-negative cell lines, with the U87 and the T98 lines showing the highest levels of expression (Fig. 2).

GFAP Expression Is Induced in 5-aza-dC–Treated GFAP-Negative Glioma Cell Lines

Although we were able to detect increased expression of the GFAP transcript in the GFAP-negative glioma cell lines treated by 5-aza-dC, we set out to confirm that GFAP was also expressed in these cells immunocytochemically following the same treatment. As expected, untreated U251 control cells demonstrated abundant intracytoplasmic GFAP filaments (Fig. 3A). In comparison, there was no detectable signal of GFAP in any of the untreated GFAP-negative glioma cell lines. However, following 5-aza-dC treatment, all glioma cell lines became positive for GFAP expression (Fig. 3B). In order to quantify the relative re-expression of GFAP in each cell line, we enumerated the expression of GFAP in 5 random fields of view under lower magnification. The expression of GFAP was calculated to be 90% in the GFAP-positive U251 cell line. In comparison, the GFAP-negative cell lines went from 0% in untreated samples to 65%–85% GFAP re-expression in 5-aza-dC–treated cell lines. A typical low-magnification field of view used for enumeration is shown in Supplementary Material, Fig. S1.

GFAP Promoter Is Differentially Methylated in Glioma Cell Lines and Primary Patient Samples

Given that GFAP expression was restored in 5-aza-dC–treated GFAP-negative glioma cells, this suggested that the inactivation of GFAP in glioma cell lines was potentially occurring on the basis of DNA promoter hypermethylation. Examination of the upstream 2.2 kb of the GFAP promoter does not reveal CpG islands when analyzed with the MethPrimer prediction algorithm. A CpG island is present within the GFAP gene within intron 5, although the methylation profile for this region in the GFAP-negative U87 cell line indicated this region to be unmethylated (data not shown) and therefore unlikely to play a role in gene silencing by methylation. There are clusters of CpG dinucleotides throughout the GFAP promoter region that conform to putative methylation sites. We designed primers to several of these regions, including the region that flanked −1062 and −870, which contained 6 potential CpG sites as identified by the MethPrimer analysis software. We therefore used bisulfite sequencing to identify methylated CpG sites in this region.¹⁷ Of the 6 CpG sites in this region, CG2, CG5, and CG6 show the greatest differences between U251 MG and the GFAP-negative cell lines, U87 MG and T98. These sites are methylated 40% of the time in U251 cells compared with 49% and 87% in T98 and U87, respectively (Fig. 4A). We also analyzed the methylation status of the CpG sites in primary glioma samples from 10 patients diagnosed with GBM and observed variable promoter methylation (Fig. 4B). Bisulfite sequencing of these primary samples revealed that the 6 CpG sites ranged from 7% GBM C to 88% (GBM A) methylation. Previously, Takizawa et al.¹⁸ identified a region that contains a CpG dinucleotide within a STAT3 binding element that is differentially regulated by methylation during development. We evaluated this region by bisulfite sequencing and determined that the CpG site within the STAT3 binding element was 100% unmethylated in both U251 and U87 cell lines (data not shown).

GFAP Promoter Methylation Status Correlates with GFAP Expression in Primary Tumors

The methylation status of the primary tumors varied between each of the GBM samples assayed. Given that a limitation of this technique involves isolation of DNA from a tumor that may comprise heterogeneous cell types, including astrocytes, endothelial cells, smooth muscle cells, and other normal stromal cells, it is reasonable to conclude that the methylation analysis might be influenced by the presence of DNA from normal cell types trapped within the primary tumor mass. Therefore, the primary samples were analyzed for GFAP expression to determine if there was a correlation between methylation status and GFAP expression. Following immunohistochemistry with an anti-GFAP antibody, 3 of the GBM samples were found to be GFAP negative and 6 GFAP positive, as was a GFAP-positive sample of non-neoplastic brain (Fig. 5). When the immunohistochemistry results were compared with the GFAP promoter methylation status, a strong inverse correlation was found for 9 of the 10 primary specimens. The primary samples that retained GFAP expression had little GFAP promoter methylation (7%, 12%, 9%, 22%, 23%, and 9% for GBM C, E, F, G, H, and J, respectively), whereas the GFAP-negative GBMs had elevated methylation (88%, 32%, and 58% for GBM A, B, and D, respectively). Interestingly, the immunohistochemical staining of GBM I was negative for GFAP, despite a very low level of GFAP promoter methylation (8%), indicating that additional regulatory mechanisms of GFAP expression may exist in this sample or possibly a mutation within the GFAP gene that disrupts protein expression.

In Vitro Methylation of GFAP Promoter Abolishes Transcriptional Activity

In order to directly assess if methylation of the GFAP promoter was sufficient to block gene expression, an in vitro methylation assay was performed. The upstream 2-kb GFAP promoter was purified and either untreated or treated in vitro with M.Sss1 CpG methyltransferase. The fragments were then directionally cloned into a luciferase reporter construct and co-transfected with a renilla control construct into the U251 MG glioma cell line. In addition, we transfected cells with a negative control that lacked an upstream promoter sequence (empty vector) and a positive control that had a CMV promoter driving luciferase expression. The cell extracts were analyzed for luciferase expression after normalizing for renilla expression. We observed high expression of luciferase in cells transfected with the expression construct that was not treated in vitro with the methyltransferase and the CMV-positive control. In contrast, little luciferase expression was detected in the empty vector control and the luciferase expression was abolished in the cell lysates obtained from cells transfected with the expression construct that was treated with the in vitro methyltransferase (Fig. 6). Taken together, these data suggest that the key CpG sites within the promoter region are capable of regulating transcription from the GFAP promoter.

Gene Databank Query

From a review of over 200 primary gliomas in the databases, the 17q21 locus and the gene for GFAP are not involved in copy number aberrations or mutations.^19,20

Discussion

In this study, we have demonstrated the absence of GFAP expression in many glioma cell lines and primary patient tumor samples. Interestingly, following treatment of GFAP-negative glioma cell lines with 5-aza-dC, we show that these cell lines can re-express GFAP. In addition, we performed bisulfite sequencing on primary glioma samples and glioma cell lines and showed an inverse relationship between GFAP promoter methylation status and GFAP expression. Using a gene reporter assay with a 2-kb GFAP promoter cloned upstream of a luciferase gene, we showed that methylation of the GFAP promoter downregulates the expression of the luciferase gene. Our data suggest that epigenetic silencing of the GFAP gene by DNA methylation of its promoter region may be one mechanism by which GFAP is downregulated in human gliomas and glioma cell lines.

Several studies have documented that GFAP expression is frequently diminished in human glioma specimens with increasing degree of anaplasia.^5,6,21 We have previously shown that primary malignant gliomas that are initially positive for GFAP when placed into tissue culture in vitro frequently lose expression of GFAP after serial passaging.²² We have also shown that the GFAP gene locus in GFAP-negative glioma cell lines does not harbor any large DNA rearrangements or deletions to account for the loss of gene expression.²³ Recent genome-wide sequencing of GBMs has revealed no somatic mutations of GFAP, although reduced GFAP expression has been observed in primary GBMs, xenograft specimens, and GBM cell lines.^19,20 Although each GFAP exon was resequenced by Parsons et al.¹⁹ for the presence of somatic mutations, a potential limitation of the exon resequencing studies is contamination of normal cells within the patient samples that may limit the ability to detect homozygous deletions and somatic mutations. To us, this suggested that GFAP expression in human gliomas may be regulated by other mechanisms at the transcriptional, post-translational, or epigenetic level. Several years ago, we employed methyl-sensitive restriction enzyme digests to human gliomas and showed putative hypermethylation of the GFAP gene in GFAP-negative human glioma cell lines.²³ Using deletional constructs of the GFAP promoter region that were methylated with Msp-1 and HpaII methylases in chloramphenicol acetyltransferase reporter assays, we showed that a 2-kb segment of the mouse GFAP promoter was sufficient to inactivate GFAP transcription.²³

Recently, we used an epigenetic genome-wide screen to identify genes aberrantly silenced by promoter methylation in human medulloblastoma.²⁴ In this study, we used bisulfite sequencing to accurately identify the regions of promoter methylation of SPINT2 in medulloblastoma cell lines, and methylation-specific PCR to examine the methylation state of SPINT2 in medulloblastoma tumor specimens.²⁴ With the advent of these new techniques to study DNA methylation, we chose to apply them to the analysis of human gliomas and to the study of the GFAP gene in particular.

DNA methylation is one mechanism by which genes are regulated during embryogenesis, leading to timely cell differentiation and appropriate tissue patterning (reviewed in Kiefer²⁵). Interestingly, Takizawa et al.¹⁸ showed that DNA methylation of GFAP is a critical cell-intrinsic determinant of astrocyte differentiation in embryogenesis. In this study, the authors identified a CpG dinucleotide within a STAT3 binding element of the GFAP promoter. This CpG dinucleotide was highly methylated in E11.5 neuroepithelial cells, but demethylated in cells at a later developmental stage, at which time these cells become responsive to STAT3 and express GFAP.¹⁸ Furthermore, Shimozaki et al.²⁶ determined that the STAT3 CpG site is methylated in undifferentiated embryonic stem cells and is maintained in neuroectoderm-like derived embryonic stem cells but becomes unmethylated when the cells become competent to differentiate into GFAP-positive astrocytes.

Another mechanism of control of gene expression in development occurs through genomic imprinting, a process whereby 1 of 2 alleles that is inherited by the offspring is partially or completely silenced, usually through DNA methylation. These genetic loci are known as imprinted genes, and although small in number (less than 1% of all genes in the mammalian genome), many are expressed in the brain. Interestingly, there is increasing evidence that imprinted genes influence brain function and behavior by affecting neurodevelopmental processes.²⁷ That said, there is no evidence at this time that the gene for GFAP is an imprinted gene in the human brain.

How does the loss of expression of GFAP influence the phenotype of human malignant gliomas? As our understanding of the function of intermediate filaments in a variety of cell types is still incomplete, this is a difficult question to answer. Previous studies in our lab have demonstrated that when GFAP expression is restored in GFAP-negative SF-126 glioma cells, glioma cell proliferation and tumorigenicity are dramatically reduced.²⁸ Conversely, when GFAP expression is eliminated from GFAP-positive U251 MG glioma cell lines by antisense treatment, glioma cells had a growth advantage and enhanced invasiveness.²⁹ These and other studies³⁰ suggest that intermediate filaments such as GFAP have cellular functions, which may extend beyond their roles as known cytoskeletal support structures within the cytoplasm of glial cells. These functions are likely mediated through the interaction of GFAP and other intermediate filament-associated proteins, an example of which is plectin.³¹

In attempts to further characterize additional proteins that interact with GFAP, we recently identified that fascin, an actin bundling protein, interacts primarily with the NH2-terminal region of GFAP.¹⁴ This interaction is of interest because it provides a potential link between the intermediate filamentous and the actin microfilamentous systems within glial cells. As actin microfilaments are the primary effector molecules found within filopodia, stress fibers, microspikes, and microvilli^32–34—cellular structures known to be important in cell migration—it is tempting to speculate that the reduction in GFAP content found in the higher grades of human gliomas may exert its influence on the migration and invasiveness of glioma cells through the modulation of the GFAP:fascin interaction.

In summary, our data provide a possible explanation for the loss of GFAP expression in human glioma cell lines and tumor specimens. Precisely how and why DNA methylation occurs on the promoter region of the GFAP gene in human gliomas, whether by the overall process of genome instability associated with cancers, by increased activity of DNA methyltransferase, or by de novo or clonally inherited methylation mechanisms, remains an important question worthy of further investigation.

Supplementary Material

Supplementary material is available at Neuro-Oncology Journal online.

Conflicts of interest statement. None declared.

Funding

This work was supported by grants from the Canadian Institutes of Health Research (MOP 74610), b.r.a.i.n.child, the Wiley Fund, and the Laurie Berman Fund for Brain Tumour Research at the Hospital for Sick Children. P.N.K. was supported by a fellowship from the Canadian Cancer Society. J.T.R. is a scientist of the CIHR.

References