Introduction
Since the introduction of temozolomide (TMZ) chemotherapy in the standard care protocol for glioblastoma (GBM) patients, the analysis of O
6-methylguanine DNA methyltransferase (MGMT) status has become a key biological marker. MGMT is a DNA repair protein which removes alkyl adducts on the O
6 position of guanine, inducing resistance against alkylating agents such as TMZ. MGMT status is currently used to stratify patients in clinical trials, such as in the RTOG 0525 randomized phase III trial that compared standard adjuvant TMZ with a dose-dense schedule in newly diagnosed GBM patients [
1]. It was also used to select patients in the CENTRIC phase III trial that assesses the usefulness of adding cilengitide to the standard treatment in newly diagnosed GBM patients [
2]. As MGMT status is a strong predictive factor of response to treatment with TMZ [
3], it is determined in most on-going clinical trials using this drug. The recently published results of the NOA-08 trial on elderly malignant astrocytoma patients and the Nordic trial on elderly GBM patients showed that elderly malignant astrocytoma patients with methylated
MGMT promoter may receive as much benefit from TMZ as from radiotherapy alone. This suggests that testing MGMT methylation status may help treatment decision making in these patients, which might increase the demand for MGMT methylation test in clinical practice [
4].
Despite the increasing needs for MGMT methylation testing, there is no consensus concerning the best technique for its assessment.
MGMT is mainly regulated at the epigenetic level: the methylation of the MGMT CpG island silences the gene and therefore is associated with a lack of MGMT protein expression. Most studies reporting a link between MGMT status and survival in GBM patients have used techniques based on DNA methylation [
5]. These techniques are designed to detect methylated (or unmethylated) CpGs located in exon 1 and immediately downstream, under the assumption that methylation of these CpGs reflects protein expression and therefore can predict response to TMZ. The CpG island of
MGMT includes 98 CpG sites [
6] and it has been shown that the patterns of methylation are rather heterogeneous. Some studies investigated to determine which CpG sites are critical for MGMT expression. Everhard et al. studied methylation at 52 CpG sites by pyrosequencing (PSQ) in GBM and compared the results with mRNA expression. These authors found that methylations of the whole 52 CpGs (CpGs 12–46 and CpGs 71–97), as well as CpG 27, 32, 32–33, 72–83, 73, 75, 79 and 80 were significantly correlated with expression. Shah et al. analyzed the methylation profile of 97 CpGs by bisulfite sequencing of GBM tissues and correlated the results with mRNA and protein expressions. 39 CpGs and 25 CpGs were significantly correlated with mRNA and protein expression, respectively [
7]. Malley et al. studied the methylation status of the entire CpG island of
MGMT using PSQ and compared it with
MGMT mRNA expression in GBM cell lines and xenografts. They identified two separate regions (spanning CpG 25–50 and CpG 73–90) where methylation was significantly correlated with expression. Furthermore, using a luciferase reporter assay they showed that individual CpGs (in particular CpG 89) can play a significant role in
MGMT promoter activity [
6]. The primers commonly used for the methylation-specific PCR technique (MSP) bind to sequences encompassing CpGs 76–80 (forward) and CpGs 84–87 (reverse) [
8]. As a derived method, a real-time-quantitative PCR-based MSP, developed by MDxHealth (Liège, Belgium), which has been applied in several international clinical trials and is used for MGMT testing by some clinical laboratories, such as LabCorp in north America, utilizes primers that include CpGs 76–80 and CpGs 88–90. This technique generally detects MGMT methylation in about 30 % of GBM [
1,
9]. These MSP-based techniques have the potential drawback of failing to detect heterogeneous methylation because primers are designed to amplify sequences where all CpGs are fully methylated. Another drawback of using a commercial service is a high cost and the long turnover time, which is not always suitable in a day-to-day practice.
In our recent study in which we compared five methods (MS-PCR, MethyLight, PSQ, MS-HRM and IHC) to analyze MGMT status in a series of 100 GBM patients who had received standard care treatment (Radiotherapy plus concomitant adjuvant TMZ chemotherapy), we found that the best prediction of survival was obtained with PSQ [
10]. PSQ allows quantification of methylation at each individual CpG and therefore can detect heterogeneous methylation. The PSQ assay used in this previous study examined 5 CpG sites (CpGs 74–78, PyroMark Q96 CpG MGMT kit, Qiagen). However, some of the critical CpGs for
MGMT promoter were not included. In an attempt to determine the clinically most relevant CpGs for MGMT methylation assessment, we extended our PSQ analysis to cover CpG 74 through CpG 89 in one subset of patients and tested the impact of methylation at each CpG site as well as the average methylation values of selected consecutive CpGs on predicting patient survival.
Discussion
Methylation status of
MGMT is currently recognized as a strong prognostic and predictive factor for newly diagnosed GBM patients treated by TMZ in an adjuvant setting [
3,
10,
14‐
16,
18‐
21]. However, there is a wide choice of techniques to assess methylation and depending on the method, the percentage of patients classified as potential responders to alkylating agents can vary greatly, as we have recently pointed out [
10]. PSQ has been shown to be a robust technique, with good clinical performances in predicting TMZ response, according to the results of different studies. With a cut-off between 8 and 10 % (average of all CpGs tested), from 42 to 53 % of patients are considered as methylated [
10,
14‐
16,
18]. However, other CpGs, apart from the 5 CpGs (74–78) analyzed in most of these studies can play a critical role in the transcriptional control of
MGMT, and could therefore be useful biomarkers to predict the outcome of GBM patients treated with TMZ. In our study, we analyzed 16 CpGs by PSQ with a custom-designed test and sought to determine which individual CpG, or combination of CpGs is best at predicting therapeutic response in a cohort of newly diagnosed GBM patients that were treated with the Stupp regimen.
Among the topmost ten ranking GpGs or means of CpGs associated with outcome, we found CpGs 89, means of CpGs 84–88, 85–89 and 74–89. Substitution of CpGs 89, CpGs 84–87 and CpGs 76–87 has been shown to significantly attenuate promoter activity of
MGMT in a luciferase reporter assay [
6]. This firmly supports the hypothesis that
MGMT methylation impacts the survival of patients through a decreased expression of MGMT that would reduce resistance against alkylating agents. A similar conclusion was drawn from the study of Bady et al. These authors compared the MGMT CpG methylation levels obtained by the HumanMethylation 450 BeadChips (Illumina) to MGMT expression and the patients’ outcome. Among the 18 probes of interest located in or near the promoter region, the two CpGs showing the strongest correlation with expression (CpG 31 and CpG 84 in our study) were also those best correlated with outcome [
22]. It is of note that the methylation levels of CpG 84, which is the only CpG interrogated by their BeadChip among the 16 CpGs we tested, is also well correlated with the patients’ outcome in our study.
A major issue for quantitative techniques such as PSQ is the determination of a cut-off to dichotomize patients into methylated and unmethylated status. To allow comparisons among the tested CpGs, we calculated an optimal outcome-based cut-off for each CpG, as carried out in a previous study [
10]. In this previous study using the Qiagen PSQ test (CpGs 74–78), the optimized cut-offs were very similar for OS (4, 11, 6, 6, 5 and 8) and PFS (4,4,8,6,4 and 8) to the values obtained in the present study for OS (8, 11, 5, 7, 4 and 9) and for PFS (8, 5, 8, 6, 4 and 7), concerning CpG 74, CpG 75, CpG 76, CpG 77, CpG 78 and mean CpG 74–78, respectively. In the present study we also validated the cut-off of 9 % (mean values CpG 74–78) in an independent cohort of 50 GBM patients. As frozen tumor tissue is not always available in daily practice, for this cohort of patients, we worked with FFPE samples. Recently, Reifenberger et al. [
15] found a good degree of concordance between PSQ and MS-PCR: at a cut-off of 8 % (mean values CpG 74–78) 153/166 (92 %) of patients were identically classified. Furthermore, for patients treated with chemotherapy, PSQ and MS-PCR looked similar to predict outcome in this series of elderly patients (>70 years). In their study, Bady et al. [
22] used an external data-set of 50 GBM patients that had been pyrosequenced by our group. Using an iterative procedure based on segmented regression, these authors estimated the cut-off at 7.28 % average methylation. This shows that PSQ is a robust technique and several reports are now available that agree on the best cut-off for the most commonly used PSQ test (the mean of CpGs 74–78) being around 9 %.
In conclusion, the methylation levels at several individual CpGs sites or combinations of CpGs in the MGMT CpG island determined by PSQ—some of which were previously found to be correlated with MGMT expression—are highly significant predictive markers for GBM patients treated with the current standard care treatment. CpGs 84, 89 and mean CpGs 84–88 appear particularly useful. The mean of CpGs 74–78 was also among the CpGs or combinations of CpGs most strongly associated with the outcome of patients. Because (1) a commercial kit is available for determining the level of methylation of these 5 CpGs by PSQ, which makes it easy to standardize the test (2) this kit is currently successfully used by different groups (3) we can now be confident about the best cut-off allowing stratification of patients into good and poor responders to TMZ, we recommend to test CpGs 74–78 for PSQ with the PyroMark CpG MGMT kit and use the mean methylation of all 5 CpGs to determine the MGMT methylation status.
Acknowledgments
Samples in Rennes were collected and stored by the Centre de Ressources Biologiques (CRB). The specimens provided by the Marseille’s team were stored in the AP-HM tumor bank (authorization number 2008/70). M.S.N. post-edited the English style. Funding was provided by the French Ministry of Health (Support for Costly Cancer Technical Evaluation–STIC–Gov-0478).