Epigenetic alterations in human brain tumors in a Brazilian population

Aberrant methylation of CpG islands located in promoter regions represents one of the major mechanisms for silencing cancer-related genes in tumor cells. We determined the frequency of aberrant CpG island methylation for several tumor-associated genes: DAPK, MGMT, p14, p16, TP73, RB1 and TIMP-3 in 55 brain tumors, consisting of 26 neuroepithelial tumors, 6 peripheral nerve tumors, 13 meningeal tumors and 10 metastatic brain tumors. Aberrant methylation of at least one of the seven genes studied was detected in 83.6% of the cases. The frequencies of aberrant methylation were: 40% for p14, 38.2% for MGMT, 30.9% for, p16, 14.6% for TP73 and for TIMP-3, 12.7% for DAPK and 1.8% for RB1. These data suggest that the hypermethylation observed in the genes p14, MGMT and p16 is a very important event in the formation or progression of brain tumors, since the inactivation of these genes directly interferes with the cell cycle or DNA repair. The altered methylation rate of the other genes has already been reported to be related to tumorigenesis, but the low methylation rate of RB1 found in tumors in our sample is different from that so far reported in the literature, suggesting that perhaps hypermethylation of the promoter is not the main event in the inactivation of this gene. Our results suggest that hypermethylation of the promoter region is a very common event in nervous system tumors.


Introduction
Current evidences suggest that epigenetic changes play an important role in the evolution of human cancers.CpG islands are 0.5 to 2.0 kb DNA regions rich in cytosine-guanine dinucleotides, present in the 5' region of approximately half of human genes (Baylin et al. 1998).Aberrant methylation occurs in tumor cell CpG islands as an important pathway for the repression of gene transcription of tumor-related genes, and this loss of gene expression contributes to the pathogenesis and progression of malignant neoplasms (Baylin et al. 1998).
Studies on methylation in nervous system tumors have shown that astrocytic tumors present several methylated genes (Burns et al., 1998;Nakamura et al., 1998Nakamura et al., , 2001aNakamura et al., , 2001b;;Esteller et al., 1999Esteller et al., , 2000a;;Park et al., 2000;Dong et al., 2001;Gonzalez-Gomez et al., 2003a, 2003b, 2003c).Epigenetic inactivation in non-astrocytic tumors is involved with several genes in oligodendrogliomas (Bortolotto et al., 2000;Wolter et al., 2001;Alonso et al., 2003).Watanabe et al. (2001aWatanabe et al. ( , 2001b) ) found that a simultaneous alteration in the interaction pathways RB1/CDK4/p16 INK4a / p15 INK4b and that TP53/p14 ARF /MDM2 is important in the development of the malignant phenotype of the anaplastic oligodendrogliomas.It was thus concluded that methylation is an important event in the formation of gliomas but, however, there are few studies of methylation in meningiomas (Tse et al., 1998;Boström et al., 2001), ependymomas (Alonso et al., 2003) and neuroblastomas (Gonzalez-Gomez et al., 2003d).Yin et al. (2002) analyzed nervous system tumors of various subtypes and found that 28% of the sample investigated had at least one methylated gene and that TP53 and p14 ARF presented a high frequency of genetic and epigenetic alterations, suggesting that those alterations must have an important role in the development of brain tumors.We determined the methylation profile for 55 brain tumors, studying seven genes that frequently show promoter region methylation in several neoplasms (breast, brain tumors, lung, gastric, colon cancer etc.).The genes were: deathassociated protein kinase (DAPK), O 6 -methyl-guanine-DNA methyltransferase (MGMT), p14 ARF , p16 INK4a , TP73, RB1 and tissue inhibitor of metalloproteinase 3 (TIMP-3).These genes were selected for analysis based on their localization on genomic regions involved in chromosome deletions in brain tumors or the fact that they were known to have a key role in tumor development.

Tissue samples and DNA preparation
Fresh brain-tumor tissue and blood samples were obtained from 25 women and 30 men (n = 55) Brazilian individuals whose age ranged from 10 to 77 years (mean = 50 ± 17.3 years).These individuals were treated for brain cancers at Santa Casa Hospital in Ribeirão Preto city, Brazil.This study was approved by the Medical Ethics Committee of the University of São Paulo (FMRP-USP) and National Ethics Committee of Research (CONEP) and informed patient consent was also obtained.Pathological diagnosis was performed according to the WHO classification (Kleihues and Cavenee, 2000;Kleihues et al., 2002) as: 26 neuroepithelial tumors (8 astrocytomas, 10 glioblastomas, 1 neuroblastoma, 2 ependymomas and 5 oligodendrogliomas), 6 peripheral nerve tumors (schwannomas), 13 meningiomas and 10 metastatic tumors.DNA was prepared from frozen tissues and blood samples using standard methods (Rey et al., 1992).
We used the following methodology: 2 mg of genomic DNA was denatured with NaOH, modified by treatment with sodium bisulfite, purified using the DNA clean-up kit (Promega, Madison, WI), treated again with NaOH, precipitated with ethanol, and resuspended in water.Specific PCRs were performed for the methylated and unmethylated alleles under standard conditions with variable annealing temperatures (55-66 °C).The PCR products were electrophoresed on either 2 to 3% agarose gels or non-denaturing 6% polyacrylamide gels, stained with ethidium bromide and visualized under UV illumination.Control reactions without DNA were performed for each PCR.For positive and negative controls for methylated alleles we used DNA from the lymphocytes of healthy volunteers, treated and untreated with Sss1 methyltransferase (New England Biolabs, Beverly, MA, USA) and then subjected to bisulfite treatment.Samples giving signals approximately equivalent to the positive control were designated as methylated.All experiments were repeated at least twice and the PCR for all samples demonstrating positive methylation for a specific gene was repeated at least three times to confirm these results.

Statistics
We used the methylation frequency (MF) defined as the number of individuals of the sample with methylated genes divided by the total number of individuals of the sample multiplied by 100.We also used methylation index (MI) to determine the overall methylation rate in individual samples, the MI being defined as a fraction representing the number of methylated genes divided by the number of tested genes.Statistical analysis was performed using the chi-squared (c 2 ) test and Fisher's exact test for differences between groups and the Student's t-test and Mann-Whitney test for differences between means.

Promoter methylation in all samples
The MSP results for the 7 loci are shown in Tables 1  and 2 which summarize the promoter methylation of each gene in every tumor subtype analyzed.A total of 46 of 55 samples (83.6%) displayed CpG island hypermethylation in at least one gene and the methylation frequency for each locus in this tumor series varied from 1.8 to 40%, representative examples of the MSP analysis being shown in Figure 1.The highest promoter methylation frequency for the total number of tumors was 40% for the p14 ARF locus and the lowest 1.8% for the RB1 locus , while the corresponding frequencies for gliomas were 68.8% for the MGMT locus and 0% for the RB1 locus, although it should be noted that individual tumors and gliomas showed higher methylation frequencies (Table 1).
The methylation index of multiple loci in a tumor series ranged from 0 to 0.71 (Table 2), with an overall mean of 0.22 and this table also shows that the neuroepithelial tumors displayed an overall mean of 0.25, the tumors of peripheral nerve groups 0.14, meningiomas 0.16 and the other brain tumors groups 0.25.No statistical differences were found in the methylation rate when DAPK, p16 INK4a , TP73 and RB1 loci were analyzed in the groups of tumors studied (p > 0.05) but the MGMT locus showed highly significant results (p < 0.05) when the epithelial tumors were compared to meningiomas (Fisher's test = 0.00004, significant at p = 0.05) and significant when epithelial tumors were compared to other tumor types (Fisher's test = 0.01754, significant at p = 0.05).The p14 ARF locus showed no significant differences when the meningiomas and epithelial tumors were compared to others tumors types.However a highly significant difference for this locus was observed after comparison between peripheral nerve tumors and meningiomas (Fisher's test = 0.00052, significant at p = 0.05) or peripheral nerve tumors and others tumors types (Fisher's test = 0.00478, significant at p = 0.05).The TIMP-3 locus showed significant results when epithelial tumors were compared to meningiomas (Fisher's test = 0.00668, significant at p = 0.05) and the meningiomas were compared with other tumor types (Fisher's test = 0.03112, significant at p = 0.05).No significant association could be determined between the methylation status of each locus and data on the age and sex of the patients, the overall MI value for patients aged 50 or less was 0.26 and over 50 years 0.19 (not significant by the Student's t-test at p > 0.05) while the overall MI values grouped according to sex were 0.19 for males and 0.26 females (not significant by the Student's t-test at p > 0.05) (Tables 1 and 2).

Promoter methylation in the glioma sample
Regarding promoter methylation in the gliomas investigated (excluding ependymomas and neuroblastomas), we found that a total of 18 of the 23 samples (83.6%) displayed CpG island hypermethylation in at least one gene (Tables 1 and 2).The frequency of methylation among the genes ranged from 0 for the RB1 locus to 60.8% for the MGMT locus, although again it should be noted that individual types of gliomas showed higher methylation frequencies (Table 1).There was a significant difference (Fisher's test = 0.028, significant at p = 0.05) for the TP73 locus when oligodendrogliomas were compared to glioblastomas but all the other gliomas and genes showed no significant differences (p < 0.05), although when astrocytomas and oligodendrogliomas where compared regarding the TIMP-3 gene the probability value was very close to the  11/M 0.14 significance limit (Fisher's test = 0.535, no significant at p = 0.05).
The MI ranged from 0 to 0.57, with a mean of 0.25.When the three types of gliomas were considered the means were 0.27 for astrocytomas, 0.25 for glioblastomas and 0.23 for oligodendrogliomas (Table 2).

Discussion
This study of 55 brain tumors and gliomas investigated the methylation status of a set of loci which have been identified as frequently methylated in other cancers or cell lines.Moreover, the chosen loci are located at genomic regions involved in chromosomal deletions in brain tumors (TP73 at 1p; p14 ARF and p16 INK4a at 9p) or they play key tumor-related functions (RB1, MGMT etc.).Transcriptional silencing by hypermethylation of CpG islands in the promoter regions is accepted as a primary mechanism involved in the inactivation of tumor-related genes (Esteller et al., 2001a;Esteller and Herman, 2002).
Our study of seven genes shows that hypermethylation is variable according to the gene.The highest methylation frequencies were detected for the p14 ARF , MGMT and p16 INK4a genes and intermediate frequencies for the TP73, TIMP-3 and DAPK genes and low frequencies for the RB1 gene.
The high methylation frequency of genes p14 ARF , MGMT and p16 INK4a agree with several previous studies (Costello et al., 1996;Dong et al., 2001;Gonzalez-Gomez et al., 2003b, 2003c, 2003d) and it is known that the p14 ARF and p16 INK4a genes are located very close to each other at 9p21 and share an exon, albeit in another reading frame (Sherr, 1996).The transcriptional activity of the p14 ARF gene is regulated independently of the p16 INK4a gene, the p14 ARF gene participating in an autoregulatory feedback loop involving p53 and MDM2 (Robertson and Jones, 1998).Methylation studies of these genes have produced variable results.Yin et al. (2002) found methylation of p14 ARF in 18% of the brain tumors studied by them but a significantly low methylation of this gene in gliomas, p16 INK4a being methylated in only 7% of the total and 6% of gliomas.Alonso et al. (2003) studied a series of oligodendrogliomas and found that 44% of p14 ARF genes and 22% of p16 INK4a genes were methylated.Dong et al. (2001) found that 2% of p14 ARF genes and 12% of p16 INK4a genes were methylated, while Wolter et al. (2001) reported that 21% of p14 ARF genes were methylated in grade II oligodendrogliomas and 15% of the same genes were methylated in anaplastic oligodendrogliomas while only one sample of their series showed aberrant methylation of p16 INK4a .In our sample 40% of p14 ARF genes and 30.9% of p16 INK4a genes were methylated when considering tumors and gliomas together but when considering gliomas on their own only 34.8% of p14 ARF genes and 26.1% of p16 INK4a were methylated.We also observed that both genes were methylated in 26.9% of the neuroepithelial tumors.Although our sample of oligodendroglial tumors was small both genes were methylated in 50% of the oligodendrogliomas and 33.3% of the anaplastic oligodendrogliomas, showing a frequency pattern similar to that described by other authors.We also found that in meningiomas 61.5% of p14 ARF genes and 23% of p16 INK4a were methylated but in schwannomas only the p16 INK4a gene was methylated.These data are similar to the results reported by Yin et al. (2002), with other data showing a moderate to high methylation frequency in neuroepithelial tumors (Merlo et al., 1995;Nishikawa et al., 1995;Costello et al., 1996Costello et al., , 2000;;Dong et al., 2001;Wolter et al., 2001;Yin et al., 2002;Alonso et al., 2003Gonzalez-Gomez et al., 2003b, 2003c, 2003d) but a low frequency in meningeal tumors (Park et al., 2000;Boström et al., 2001), although methylation studies in nervous system tumors are still rare.Yin et al. (2002) found a higher frequency of methylation in p14 ARF than in p16 INK4a in the meningiomas and that p14 ARF was not methylated in the schwannomas.Our data showed the same results, indicating that in these tumors there is preferential methylation of one gene as opposed to the other.Theses results support the idea that inactivation by hypermethylation of p14 ARF and/or p16 INK4a is an important mechanism in the development of brain tumors because the inactivation of p14 ARF by epigenetic mechanisms may interfere with the network of TP53 interactions (Zhang et al., 1998;Esteller et al., 2000bEsteller et al., , 2001aEsteller et al., , 2001b;;Amatya et al., 2005).
Our data show that the MGMT gene has a high rate of methylation (38.2%), agreeing with other reports (Dong et al., 2001;Esteller et al., 2001a;Alonso et al., 2003).It is known that MGMT is a DNA repair gene, and it has been proposed that silencing the gene through promoter hypermethylation confers on the cancer cells an additional mutational capacity.It was proposed that the methylation of this gene is a factor in the transitory mutation of TP53 (Esteller et al., 1999).Brell et al. (2005) suggested that the absence of MGMT expression is a positive predictive marker in patients with anaplastic glioma.Epigenetic silencing of this gene by promoter methylation compromises DNA repair and has been associated with longer survival in patients with glioblastoma who receive alkylating agents (Hegi et al., 2005;Esteller et al., 2000a).Nakamura et al. (2001c) found that MGMT methylation was more frequent (75%) in secondary glioblastomas, which progressed from low degree astrocytomas, than in primary glioblastomas (36% MGMT methylation) and also verified that there was a possible association between methylation of this gene and alterations in the TP53 gene.Esteller (2000) observed that 40% of gliomas but none of the meningiomas studied had the promoter region of the MGMT gene hypermethylated, suggesting that MGMT might play a key role in tumorigenesis, as its inactivation causes several nucleotide changes in the genomic DNA.In our sample we found that the MGMT gene was methylated in 57.7% of gliomas and 7.7% of meningiomas.The differences between the me-thylation frequencies observed for gliomas and meningiomas in the present investigation are highly significant (Fisher's test = 0.0004, significant at p = 0.05), supporting the data previously reported.Even so, more studies are necessary to try to elucidate the role of this gene in the formation of these tumors.
A low TP53 mutation rate was found in our sample (Anselmo et al., unpublished data) leading us to suppose that the methylation of the MGMT and p14 ARF genes might be one of the alternative mechanisms involved in the inactivation of TP53, one of the tumor suppressor genes most involved in carcinogenesis (Zhang et al., 1998;Yin et al., 2002;Amatya et al., 2005).The TP73 gene was highly conserved in our sample, with only one conformational alteration being found (Anselmo et al., unpublished data), agreeing with previously published reports (Alonso et al., 2001a(Alonso et al., , 2001b;;Lomas et al., 2001).These findings suggest that epigenetic mechanisms are involved in the inactivation of the TP73 gene (Corn et al., 1999;Banelli et al., 2000;Alonso et al., 2001aAlonso et al., , 2001b;;Dong et al., 2001;Gonzalez-Gomez et al., 2003b, 2003c, 2003d;Lomas et al., 2004).The methylation rate of the sample analyzed by us agrees with other results, which described a variable rate of TP73 methylation (Dong et al., 2001;Esteller et al., 2001a;Alonso et al., 2003;Gonzalez-Gomez et al., 2003b, 2003c, 2003d).However, additional studies are necessary to elucidate the role of this gene in brain tumor carcinogenesis.
We found that in our sample the TIMP-3 gene was moderately methylated (14.6%), while in the literature a high degree of methylation has been reported for neuroepithelial tumors (Bachman et al., 1999;Alonso et al., 2001aAlonso et al., , 2001b;;Gonzalez-Gomez et al., 2003b, 2003c, 2003d).This gene is necessary to suppress tissue remodeling, angiogenesis, invasion and metastasis (Bachman et al., 1999).However, we did not find any methylation of the TIMP-3s gene in meningiomas.Bello et al. (2004) showed aberrant methylation of TIMP-3 with statistical significance in 18% of grade I tumors versus 40% of grade II -III meningiomas.It has been suggested that the aberrantly methylated TIMP-3 occurring in the low-grade forms of some histological subtypes of brain tumors may indicate that this epigenetic change occurs early during brain tumor progression (Bello et al., 2004) however no methylation of this gene was observed in the meningiomas studied by us.With regard to the gliomas our results were weakly significant (p = 0.049) as compared with pilocytic astrocytomas and schwannomas.However, our sample was relatively small and there are few studies regarding these neoplasms, so more data are needed to confirm if the inactivation of this gene by methylation is an important event in the tumorigenesis of these tumors.In fact, 18% of TIMP-3 aberrant methylation has previously been described in schwannomas (Gonzalez-Gomez et al., 2003e).
The protein coded by DAPK is possibly a mediator of apoptosis induced by IFNg and is considered as a potential metastasis inhibitor gene (Sanchez-Cespedes et al., 2000).A significant increase in the rate of methylation of the DAPK gene has been reported in oligodendrogliomas, when grade II and grade III tumors were compared, suggesting that this gene may be involved in the progression of gliomas (Dong et al., 2001;Alonso et al., 2001a).We found 12.7% of the cases methylated, in agreement with the literature (Alonso et al., 2001a(Alonso et al., , 2001b;;Dong et al., 2001;Gonzalez-Gomez et al., 2003b, 2003c, 2003d), indicating that a certain degree of methylation of this gene occurs in neuroepithelial tumors.Schwannomas did not present methylation in our sample and only 4.5% (2 of 44 cases) in a previous tumor study (Gonzalez-Gomez et al., 2003e).
The RB1 gene is one of the genes which play key roles in cell cycle control or apoptosis (Sherr, 1996).Our data show a very low rate of methylation of RB1 (1.8%), agreeing with the data of Yin et al. (2002) who suggested that this gene probably does not play an important role in the development of most brain tumors.The state of methylation of this gene is very variable in the diverse tumor types, being hypermethylated in retinoblastomas, glioblastomas, pituitary adenomas and oligodendrogliomas but unmethylated in some cases of leukemia (Sakai et al., 1991;Greger et al., 1994;Ohtani-Fujita et al., 1997;Stirzaker et al., 1997;Melki et al., 1999;Simpson et al., 2000;Nakamura et al., 2001a;Gonzalez-Gomez et al., 2003a).One of the main checkpoints in the control of cell proliferation occurs between the G1 ® S phases of the cell cycle and involves the RB and p16 proteins, loss of activity of which may cause disturbance in cell growth regulation which leads to altered cell growth.In this study, we observed that the p16 genes were hypermethylated suggesting that the signaling pathway controlled by this protein could be altered.The alteration in this control pathway has already been observed in many different tumor types (Debbas and White, 1993;Morgenbesser et al., 1994;Tse et al., 1998;Dong et al., 2001).
For almost all the genes analyzed the majority of our results are similar to those reported in the literature.The conservation of the TP53 and TP73 genes found in our samples (Anselmo et al., unpublished data) suggests that the regulation of these genes occurs through another pathway, since we observed a significant degree of methylation in the MGMT, p14 ARF , p16 INK4a genes which may have a key role in the regulation of both TP53 and RB1, which were extremely conserved in our material.
There is a possible involvement of the p16 INK4a gene in the formation of schwannomas and of p14 ARF in the formation of meningiomas, but more detailed studies are necessary.
The processes of gene inactivation by methylation do not occur in only one gene, but may occur simultaneously in a great number of genes which are often located on several chromosomes (Dong et al., 2001).We may conclude that the frequency of methylation varies according to the gene and/or the type of tumor.
Possible insights into the origins of brain tumors, as well as the possibility of improving treatment are coming to light from correlative studies of several parameters that have examined the relationships between specific types of tumors and their genetic signatures.For example, some studies have shown that oligodendrogliomas harboring allelic loss of 1p and 19q were significantly more likely to be located in the frontal lobes or bilaterally than tumors with intact alleles, which predominated in the temporal lobes and diencephalon, the `older' and deeper region of the brain (Zlatescu et al., 2001;Mueller et al., 2002), suggesting a link between the origins of oligodendrogliomas and the molecular regulation of brain development.Other studies have described an association between imaging features and tumor genotype (Cairncross, 2004, Walker et al., 2004).Some benefits from agents used in chemotherapy are related to methylation of some genes in gliomas (Brell et al., 2005, Hegi et al., 2005).Molecular genetic markers introduce important new considerations into the design and interpretation of clinical trials for some tumors and are beginning to influence treatment strategies for individual patients stricken with neoplasias.A detailed knowledge of the alterations occurring in specific tumor types allied to the use of molecular testing and other clinical and pathological criteria for refining diagnoses and perhaps individualize therapy, associated with interdisciplinary studies may shed new light on the importance of molecular markers as decisive therapeutic factors.
In conclusion, we have examined CpG island methylation in brain tumors by analyzing the methylation status of 7 genes using MSP.We found that CpG island methylation is a relatively frequent event in these neoplasms and also that it sometimes occurred during the early stages of carcinogenesis and may be already present in the low-grade forms of these neoplasms.The frequency and timing of hypermethylation varied according to the gene and concordant inactivation of some genes appears to contribute to the carcinogenesis of some of these tumors.

Methylation in human brain tumors 417 Figure 1 -
Figure 1 -Methylation analysis by MSP in brain tumors.The studied gene is given on the left of each panel.Lane U = amplified product with primers recognizing unmethylated sequences; Lane M = amplified product with primers recognizing methylated sequences.PC = positive control (DNA treated with methylase); NC = negative control (DNA of non malignant tissue); L = ladder.

Table 1 -
Methylation of brain tumors, relationships between their histological characteristics and the degree of methylation presented by the different genes studied.

Table 2 -
Summary of methylation of all seven genes in brain tumors.Filled boxes represent methylated loci and open boxes unmethylated loci.M = Male Sex; F = Female Sex.