Introduction

Gliomas are the most common primary brain tumors that are classified into four different malignancy grades according to the World Health Organization (WHO) classification [1]. Glioblastoma multiforme (GBM) belongs to WHO grade IV and is the most common and malignant human brain tumor. The median survival for GBM is only 12 to 18 months [2]. Anaplastic astrocytoma (WHO grade III) and diffuse astrocytoma (WHO grade II) have relatively low degree of malignancy and an inherent tendency for recurrence and malignant progression. WHO grade I tumors are rare. Currently, the most effective treatment method for glioma is surgery combined with radiation and/or chemotherapy [3]. Temozolomide (TMZ), an alkylating anticancer drug, is currently the most effective chemotherapy for treatment of glioma. However, it is known that not all patients benefit to the same extent for TMZ treatment.

MGMT is a highly efficient DNA methyltransferase that repair DNA alkylation damage caused by a variety of alkylating agents, especially TMZ [4]. Loss of MGMT gene expression frequently occurs in various human malignancies [5]. Studies have shown that transcriptional silencing caused by MGMT gene promoter methylation plays a crucial role for gliomas response to alkylating agents [6, 7]. For MGMT promoter methylation, most trials revealed a positive relationship between promoter methylation and therapeutic response of glioblastoma patients treated with alkylating agents like TMZ [6, 8–11]. While some other studies failed to show this correlation [12], this might be due to the current CpG islands detected are of different importance for the silencing of the MGMT gene in some certain populations, which is supported by the finding that the position of the methylated CpG islands is highly variable in MGMT deficient tumor cell lines [13]. The question, which CpG island in the MGMT promoter is most reliable and efficient in silencing needs to be addressed.

In this study, we screened the whole CpG dinucleotides methylation located in MGMT promoter and first exon region using pyrosequencing to obtain quantitative results for all CpG sites in 42 Chinese glioma patients. Then, we statistically analyzed the relationship between location-specific CpG methylation and MGMT gene expression levels. Besides, correlation between IDH1 mutation and MGMT promoter methylation was also analyzed.

Materials and methods

Samples

Tissue samples were collected from 42 patients with astrocytoma, oligodendroglioma, and GBM (WHO grades I, II, III, and IV) treated at the department of Neurosurgery, Tangdu Hospital, affiliated with The Fourth Military Medical University in Xi’an city, China. None of the patients had any other forms of cancer. Histological typing of the tissues was performed according to the WHO grading system of brain tumors using standard histological and immunohistological methods [1]. The study was approved by the Tangdu Hospital, and informed consent was obtained from all patients.

Bisulfite treatment

DNA from frozen tumors was extracted using E.Z.N.A tissue DNA kit (OMEGA Inc.). Two micrograms of DNA was converted using the EpiTect bisulfite kit (QIAGEN Inc.), according to the manufacturer’s instructions. Cytosine and its counterpart 5-methylcytosine behave differently when single-stranded DNA is treated with sodium bisulfite. While cytosine residues react with this reagent and are converted to uracil, 5-methylcytosine remains inert under the same conditions. In a subsequent polymerase chain reaction (PCR), the uracil residues are transcribed to thymine and 5-methylcytosine to cytosine [14].

PCR

Genomic DNA sequences of MGMT were downloaded from the UCSC Genome Browser database. PCR primers were designed using Methprimer. PCR was performed using a Mastercycler Gradient (Eppendorf Inc.). We used 25 μl volume reactions based on the number of amplicons required for analysis in the PyroMark Q24 Instrument (QIAGEN Inc.). Each reaction contained 400 nM of each primer, 30 ng bisulfite-treated human genomic DNA, 10 × HS reaction buffer containing 1.5 mM MgCl2, 1 U Hotstart enzyme (Genscript), and 250 μM dNTP mixture. The PCR program consisted of an initial polymerase activation step at 95 °C for 15 min followed by 50 cycles of denaturation at 95 °C for 30 s, primer annealing at 58–60 °C for 30 s, extension at 72 °C for 40 s, and a final extension step at 72 °C for 5 min. The PCR primers and annealing temperatures are shown in Table S1.

Pyrosequencing

PCR products from bisulfite-treated genomic DNA samples were analyzed by pyrosequencing to quantify site-specific methylation. The sequence primers are summarized in Table S1. Ten microliters of the amplification product was incubated with 1 μl streptavidin sepharose high performance beads (GE Healthcare Inc.) in 40 μl binding buffer. The template strands were purified and rendered single­stranded using a PyroMark Q24 vacuum workstation (QIAGEN Inc.). Beads were released into 20 μl annealing buffer containing 12.5 pmol of the respective sequencing primer, and sequencing primers were annealed to the target by incubation at 80 °C for 2 min. Quantitative DNA methylation analysis was performed on a PyroMark Q24 system with the Qiagen kit, and the results were analyzed using the PyroMark Q24 software (version 2.0.6, Qiagen Inc.). We are presenting a pyrosequencing diagram in Fig. 1a.

Fig. 1
figure 1figure 1

a The pyrosequencing diagrams present quality control for completion of bisulfate treatment. b Hierarchical clustering of 40 glioma patients with 87 CpG sites were analyzed by pyrosequencing. Green, black, and red represent methylation levels of CpG sites. By hierarchical clustering, the samples are classified into two distinct patient groups: high methylation and low methylation. The expression level of MGMT gene for each patient is shown at the bottom. b Box plot of MGMT gene methylation levels and expression. The box plot represents quantiles, medians, and extreme points. The squares and dots represent the average methylation levels of 87 CpG sites for each patient. The MGMT gene expression and methylation levels had significant differences, p < 0.001

Full size image

MGMT mRNA expression analysis

Total RNA was isolated using the RNeasy mini kit (Qiagen Inc.), according to the manufacturer’s instructions. RNA quality was assessed using the NanoDrop 2000 (Thermo Fisher Scientific Inc.). mRNA expression levels of MGMT were determined with the one-step PrimeScript RT-PCR kit (TaKaRa Inc.) by qRT-PCR using LightCycler 480 (Roche Inc.). The qRT-PCR program consisted of a reverse transcription step at 42 °C for 5 min, RT enzyme deactivation at 95 °C for 10 s followed by 50 cycles of denaturation at 95 °C for 5 s, and primer annealing at 57 °C for 20 s. The expression level of β-actin was used for normalization. All assays were run in triplicate. Relative quantification analyses were performed using the formula: 2− (Ct value of MGMT − Ct value of β-actin) [15]. The primers and probes are as follows: MGMT forward primer, 5′-CCTGGCTAA- TGCCTATTTC-3′; MGMT reverse primer, 5′-TGAACGACTCTTGCTGGAAAAC-3′; MGMT probe, 5′-FAM-ACCAGCCCGAGGCTATCGAAGAGTTC-BHQ1–3′; β-actin forward primer, 5′-CAGAGCCTCGCCTTTGCC-3′; β-actin reverse primer, 5′-CACATGCCGGAGCCGTT-3′; β-actin probe, 5′-FAM-ACCAGCCCGAGGCTATCGAAGAGTTC-BHQ1–3′.

Detection of IDH1 mutations

Pyrosequencing was used to determine IDH1 mutation status. Sequencing for mutations within the IDH1 gene was performed using the primers specified by Setty and colleagues [16].

Statistical analysis

Statistical analysis was performed using IBM SPSS statistics 19. Hierarchical clustering was based on the degree of methylation. Average linkage was used with squared Euclidean distance as an interval measure. Mann­Whitney test was used to correlate IDH1 mutation with MGMT methylation and determine the relationship between MGMT methylation, as a quantitative variable, and expression, as a qualitative variable (low versus high expression). Fractional methylation of a promoter region was calculated by averaging methylation of all CpG sites in that region. The correlation of individual CpG site methylation with MGMT mRNA expression level was calculated using Spearman’s rank correlation. Chi-square test was used to determine the repartition between IDH1 mutation and glioma classification.

Results

Determination of MGMT gene promoter methylation by pyrosequencing

To identify the DNA methylation region or sites in MGMT gene that could represent MGMT overall methylation status and further define the region that significantly correlate with MGMT mRNA expression, the entire sequence of MGMT gene was examined. A total of 98 CpG dinucleotides in the promoter and the first exon region were identified, which span 762 bp in length and start from −452 bp to +308 bp relative to the transcription start site (TSS), of which 57 CpG dinucleotides were upstream and 30 were downstream of the TSS. Firstly, we designed pyrosequencing primers for the 98 CpG dinucleotides and tested these primers using SW480 cell line DNA as the template, which was reported to be MGMT gene highly methylated cell line. To further validate the feasibility of these primers on human samples, DNA from glioma patients was used in this established system. The results showed that this system works well for human tissues, and the representative spectrum is shown in Fig. 1a and the sequencing primers are shown in Table S1.

The methylation status of 42 glioma patients was then determined by the established pyrosequencing system. In total, 98 CpG sites were analyzed, 11 CpG sites located in −300, −63, +5, +11, +16, +38, +47, +75, +209, +242, and +271 showed uniformly methylated in all samples, and, therefore, the data mining in this study has excluded these 11 sites. Hierarchical clustering of the methylation profiles of the 87 CpG sites showed strong stratification of two distinct patient groups. The first group included 13 patients (32.5 %) with considerable methylation at almost all of the CpG sites tested (average range, 25–53 %). The second group included the 27 other patients (67.5 %) that had a lower global methylation level (average range, 7–20 %) (Fig. 1b).

Correlation between overall methylation and mRNA expression of MGMT gene

We selected ACTB as the reference gene and determined the MGMT mRNA expression in 37 glioma samples. According to the median value of all tested samples, the patients were divided into two groups: high expression group and low expression group. MGMT expression and its overall methylation (mean value) were considered as qualitative and quantitative variables, respectively. The two variables were analyzed using the Mann-Whitney test, p value of one-tailed test <0.001, suggesting that the overall MGMT methylation significantly correlated with its mRNA expression (Fig. 1c).

Correlation between specific CpG island methylation and gene expression

To further identify the specific CpG dinucleotides that correlated best with MGMT mRNA expression, the association of methylation status of each CpG dinucleotides and MGMT expression was separately analyzed. By using Spearman’s rank correlation, a total of 28 individual CpG dinucleotides were identified that significantly correlated with MGMT gene expression (p < 0.01), (Fig. 2a, b). Meanwhile, we also analyzed the association of the average methylation of these 28 CpG dinucleotides and MGMT expression by Mann-Whitney test. As shown in Fig. 2c, these 28 dinucleotides correlated significantly with MGMT gene expression (p < 0.0001). (Fig. 2c).

Fig. 2
figure 2

CpG methylation of the MGMT promoter region and MGMT gene expression in gliomas. a Spearman’s rank correlations between gene expression and methylation of the 87 CpG methylation in the promoter region spanning TSS. b p values of associations between 87 CpG sites methylation and MGMT gene expression. The dotted lines in b correspond to the threshold of 0.01. The graph at the bottom indicates the three CpG regions. c The squares and dots represent the average methylation levels of 28 individual CpG sites for each patient. The MGMT gene expression and methylation levels had extremely significant differences, p < 0.0001

Full size image

To define a region that could facilitate the future clinical detection of MGMT methylation, we further examined all the CpG sites in three regions (CpGs at −452 to −374, −216 to −119, and 106 to 153) and correlated the methylation and gene expression using Spearman’s rank. All three CpG regions significantly correlated with mRNA expression (p < 0.01).

Association between MGMT methylation and glioma grades

Among the 42 glioma patients, 13 were GBM (WHO IV) and 27 were other grades gliomas (WHO II, III). In order to find potential biomarkers that will be useful in glioma classification, we focused on the CpG sites methylation related to GBM tumor grade and low-grade gliomas. The data was therefore divided into two groups: GBM patients and low-grade glioma patients, and analyzed with the t test. Twenty-five methylation sites had significant differences between the two groups (p < 0.05). The average level of methylation for relatively low-grade glioma patients was 23.1 %, but for GBM patients, it was 14.4 % (Fig. 3a). Eleven of these sites were downstream and 14 were upstream of the TSS.

Fig. 3
figure 3

a 87 CpG sites between WHO II, III grade gliomas and WHO IV patients were analyzed by the t test. In the clustering figure, 25 methylation sites had significant differences, p < 0.05. CpG methylation of the MGMT promoter region and MGMT gene expression in gliomas. b Spearman’s rank correlations between gene expression and methylation of the 25 CpG sites methylation in the promoter region spanning TSS. c p values of associations between 25 CpG sites methylation and MGMT gene expression. The dotted line in c corresponds to the threshold of 0.01

Full size image

Using Spearman’s rank correlation, a total of 10 individual CpG sites spanning the promoter were significantly correlated with MGMT gene expression (p < 0.01). The 10 individual CpG sites correlated with MGMT gene expression and glioma classification (Fig. 3b, c).

IDH1 mutation significantly correlated with MGMT hypermethylation

A total of 40 glioma patients were retrospectively analyzed by pyrosequencing to reveal IDH1 gene mutation status. Of these, 20 samples had IDH1 mutation, and 20 had intact IDH1 (Table 1). Mann-Whitney test showed a significant correlation between IDH1 mutation and MGMT hypermethylation (p value of one-tailed test = 0.012) (Fig. 4a). There was no significant correlation between IDH1 mutation and MGMT gene expression (p = 0.957) (Fig. 4b).

Table 1 Sample characteristics of 42 glioma patients included in this study and frequency of MGMT methylation and IDH1 mutation
Full size table
Fig. 4
figure 4

Box plot of MGMT gene methylation levels and IDH1 mutation. a The squares and dots represent the average methylation levels of 87 CpG sites for each patient. The IDH1 mutation between MGMT gene hypermethylation and hypomethylation had significant differences, p < 0.05. b The squares and dots represent the expression level for each patient. The IDH1 mutation and MGMT gene expression levels had no significant correlation

Full size image

Discussion

Epigenetic silencing of MGMT promoter methylation has been associated with decreased risk and longer survival in patients with glioma who receive alkylating agent. A meta-analysis including 573 patients in eight cohort studies showed that the aberrant methylation of MGMT was associated with a longer overall survival (OS) in glioma patients. Further, subgroup analysis based on ethnicity indicated that there was an association between the aberrant methylation of MGMT and disease free survival (DFS) in glioma patients among both Asian and Caucasian populations (Asians, HR = 1.40, 95 % CI = 0.03 ~ 2.78, p = 0.046; Caucasians, HR = 1.89, 95 % CI = 0.28 ~ 3.50, p = 0.021, respectively). However, MGMT methylation and OS were found to not be correlated in Asian populations [17, 18], suggesting that the predictive value of MGMT methylation might be varied in different genetic background; another possibility for this contradictory result may lie in the position of CpG islands detected are of different importance for the silencing of the MGMT gene in different populations. Thus, identifying the most reliable CpGs dinucleotides of MGMT gene methylation in Chinese population is very crucial for personalized tumor therapy.

In this study, we identified three specific regions covering 28 CpG sites of MGMT gene that have significant correlation with its expression and further proved that hypermethylation of CpG sites in MGMT gene promoter is associated with IDH1 mutation in Chinese glioma patients. Everhard et al. examined 52 CpG sites of MGMT gene in 54 patients with GBM and 24 non-tumoral brain tissues. These patients were then divided into three groups: unmethylated group, intermediate group, and methylated group. They found that six isolated CpG sites (CpGs −228, −186, 95, 113, 135, 137) and two CpG regions (−186 to −172 and 93 to 153) were significantly correlated with MGMT gene expression [19]. Malley et al. examined 98 CpG sites in 22 patients with GBM, 13 glioma cell lines, and six normal brain tissue and identified two differentially methylated regions (DMR): DMR1 (CpG25-50) and DMR2 (CpG73-90). These regions were highly correlated with MGMT promoter activity [5]. Shah et al. used bisulfite sequencing to analyze methylation levels of 97 CpG sites in 70 patients. All the CpG sites were clustered into three regions: R1 and R2 were upstream, while R3 was downstream of TSS. They also correlated individual CpG site methylation patterns to mRNA expression, protein expression, and progression-free survival (PFS) and found that seven CpG sites were significantly correlated with protein expression and PFS [20]. In our study, the 28 CpG sites fall under either DMR1 or DMR2 and are similar to the Everhard and Shah studies. Our results showed that the methylation sites of MGMT gene in Chinese glioma patients were similar to European and American glioma patients.

IDH mutation is the cause of CpG island methylator phenotype (CIMP) and leads to the CIMP phenotype by stably reshaping the epigenome. This remodeling involves modulating patterns of methylation on a genome-wide scale, changing transcriptional programs, and altering the differentiation state [21–25]. In our study, there was a significant correlation between IDH1 mutation and MGMT hypermethylation, which was consistent with previous reports [26]. However, we did not find any correlation between IDH1 mutation and MGMT gene expression. IDH1 mutation is believed to often occur in low-grade gliomas [27, 28], which was seen in our study as well. However, the data is statistically not significant, and we believe that with a larger sample size, this data will be statistically significant. Methylation of the MGMT gene promoter and IDH1 mutation were shown to be independent prognostic markers for survival of GBM patients [6, 29, 30]. Recently, several studies have demonstrated that the combination of IDH1 mutation and MGMT methylation status was significantly better than IDH1 or MGMT alone in predicting survival of glioma patients [31–33].

McDonald et al. found that the c.-56C>T (rs16906252) SNP in the promoter was associated with the presence of MGMT methylation in de novo GBM. Furthermore, carriers of the variant T allele demonstrated a clear survival benefit [34]. However, in our research, this SNP was not found in Chinese glioma patients.

There are many methods to detect MGMT gene methylation. Methylation-specific PCR (MSP) is a simple and widely used method. However, it is a qualitative method, which sometimes produces false-positive results. Bisulfite genomic sequencing (BSP) is an accurate method. The greatest disadvantage of this method is its cumbersome operation. Pyrosequencing is a new and reliable technique to detect MGMT methylation levels and can provide quantitative information at each individual CpG [35, 36]. The principle of this method is “Sequencing by Synthesis.” Pyrosequencing is a reliable technique to detect MGMT methylation levels and was the primary technique to analyze individual CpGs of MGMT gene in our study.

A significant limitation in this study is in the absence of a validation cohort of these identified CpG islands. The survival information for these patients was not available, which greatly restricted us to correlate the methylation of the newly identified CpGs sites and patient outcome. The second limitation of this study is the small sample size of clinical cohort, especially the even small number of the subtype of gliomas; thus, we have difficultly to further analyze the association of the methylation and specific subgroup tumors. A larger cohort validation will greatly strengthen the current result.