Glioblastoma, the most common primary brain tumor in adults, is a rapidly progressive and universally fatal disease [1]. Standard therapy for these tumors has been surgery followed by radiotherapy, based predominantly on the findings of a randomised controlled trial conducted in the late 1970s which demonstrated improved survival with radiotherapy [2]. The role of chemotherapy has been controversial with no single agent or regime gaining acceptance despite a recently published metaanalysis of 12 studies demonstrating some benefit [3]. However a recent study demonstrated an improvement in median survival from 12.1 months to 14.6 months with the addition of temozolomide to the previous standard of surgery and radiotherapy. Temozolomide, an alkylating agent, causes abnormal DNA repair by causing methylation of the O⁶ position of guanine residues [4, 5].

The DNA repair enzyme O⁶-methylguanine-DNA methyltransferase (MGMT) is responsible for removal of an alkyl group from the O⁶ position of guanine, and plays an instrumental role in DNA repair [6]. MGMT protein expression occurs in normal tissues and many cancers, however the level of expression varies widely [7–10]. The MGMT gene has been well characterized, and is located on chromosome 10q26. Epigenetic silencing of MGMT expression by promoter methylation has been demonstrated to be a common event in human neoplasia including gliomas [7, 10]. Methylation of the MGMT promoter region can also result in increased formation of mutations in cancer [11]. When MGMT expression is silenced, O⁶-methylguanine is able to pair with thymine during DNA replication, resulting in conversion of guanine–cytosine to adenine–thymine pairs in DNA. Crosslinking may also occur between the O⁶-methylguanine-DNA adduct and opposite cytosine residues, thereby blocking DNA replication [10].

In the case of glioblastoma, hypermethylation of the MGMT promoter is of interest for the role it plays in effectiveness of alkylating agent chemotherapy as much as for its role in tumorigenesis. As the major mechanism of action of MGMT is in direct opposition to that of temozolomide, the expression of MGMT may be an important factor when considering sensitivity of patients with glioblastoma to temozolomide therapy. A number of studies have demonstrated the value of assessment of the methylation status of the MGMT gene in predicting the response of patients to therapy with alkylating agents [12–16]. More recently Hegi and colleagues demonstrated that methylation of the MGMT gene in patients with glioblastoma resulted in improved survival after temozolomide treatment when compared with those patients in which the MGMT promoter was unmethylated (21.7 vs. 12.7 months) [5]. This study demonstrated the potential importance of MGMT promoter methylation when considering which patients with glioblastoma should be treated with temozolomide, that is can we consider MGMT promoter methylation as a potential predictive marker for treatment benefit, thus potentially allowing to select patients for specific treatments.

A number of methods have been described to assess the methylation status of the promoter of the MGMT gene. The most commonly used method is Methylation Specific PCR (MSP). This involves the use of two sets of primers, one specific for an unmethylated promoter and one specific for a methylated promoter. The presence of a PCR product with either set of primers indicates a positive result. The number of CpG dinucleotides assessed by either set of primers is typically small, usually around 5 in number. An alternative method of assessment of the methylation status of the MGMT promoter is by promoter sequencing. This involves amplification of a region of the MGMT promoter rich in CpG dinucleotides to determine the methylation status of the promoter. Due to the use of primers that frame a larger region of the promoter, this technique allows the interrogation of a larger number of CpG dinucleotides, typically around 25. To date only one group has applied the latter technique to glioma—in oligodendroglioma [17] and more recently in high grade astrocytoma [18].

Furthermore, there are no studies examining potential alterations in the methylation status of the MGMT promoter in recurrent GBM tumor(s) when compared directly to a matched primary GBM tumor (from the same patient) to ascertain the effect of exposure to radiotherapy and chemotherapy agents on gene expression. In this study we examined MGMT promoter methylation and protein expression in 10 patients where matched primary and recurrent frozen glioblastoma tissue were available. We also measured the level of MGMT promoter methylation and protein expression in multiple sampling sites taken from large glioblastoma tumors (n = 7). Our study demonstrates that MGMT promoter methylation varies in tumor samples from the same patient before and after treatment taken at a mean interval of 207.38 days (range 19–367), and that this is unlikely to be the result of tumor heterogeneity.

DNA repair enzymes play a vital role not only in the pathogenesis of cancer, but also in the effectiveness of treatment. The methylation status of the MGMT promoter, and subsequent silencing of MGMT production and activity, may have an important role in the effectiveness of alkylating chemotherapeutic agents such as temozolomide [11]. It has been demonstrated by Hegi and colleagues that methylation of the MGMT promoter, as assessed by MSP, is predictive of patient response to temozolomide [5].

An alternative method of assessing the methylation status of the MGMT promoter is sequencing of bisulfite treated genomic DNA. Bisulfite treatment results in the conversion of unmethylated cytosine residues to uracil, while methylated cytosines are protected. By amplification with PCR and sequencing, it is possible to assess the methylation status of a far wider region of the promoter than MSP. In our case, we could assess 25 CpG dinucleotides whereas primers for MSP generally contain around 5 CpG dinucleotides. Promoter sequencing of the MGMT gene has previously been described in gliomas, but only by one group on two occasions [17, 18]. The downregulation of MGMT expression by promoter methylation appears to be gradual and related to the degree of methylation as assessed by promoter sequencing based on a recent in vitro investigation [19].

Our study has demonstrated that the extent of methylation of the MGMT promoter may either increase or decrease between different specimens taken from the same tumor before and after adjuvant treatment. There are a number of potential explanations for these changes, including regional variation within the tumor, direct influence on methylation by treatment, selection of unmethylated cell populations by treatment and further dedifferentiation of tumor.

Glioblastomas are known to be heterogenous at not only the macroscopic and microscopic levels but also at the molecular level [1, 20, 21]. The most well recognized implication of this heterogeneity is sampling error that can occur at the time of histological diagnosis. Additionally, the application of immunohistochemical markers such as proliferative markers may be affected by heterogeneity as has been described [22]. More recently, heterogeneity of molecular markers (specifically loss of heterozygosity of chromosomes 1p and 19q) has been demonstrated [23]. In melanoma, a similarly heterogenous tumour, variability in MGMT promoter methylation has been demonstrated [24].

By obtaining tissue from different regions of a number of large glioblastomas we were able to demonstrate that for the most part there was little heterogeneity in MGMT promoter status across each tumor. Similarly, there was little variability in MGMT protein expression across each tumor. This suggests that the variation seen between samples from consecutive surgeries is unlikely to represent variation within the one persistent or recurrent tumor. This confirms the findings of a recent study [18].

By contrast, Patient B in our series can be considered an example of regional variation within the same tumor. This patient had evidence of multifocal disease at presentation, and the largest region of tumor was debulked. When another region expanded rapidly less than 3 weeks later, a second operation was performed. This appeared to be rapid progression rather than tumor recurrence. While the MGMT promoter was completely unmethylated in the first tumor sample, the second demonstrated 44% methylation.

Pharmacological manipulation of methylation in cancer may have important implications for treatment. Indeed specific demethylating agents such as 5-aza-2′-deoxycitidine have been developed [25]. What is less clear is the impact of other treatments including radiotherapy. Widespread changes in DNA methylation have been demonstrated in some in vitro studies, however in one such study the MGMT promoter was not demonstrated to be affected by radiotherapy [26, 27]. On the other hand, microscopic changes after radiotherapy in glioblastoma have been well recognized. So while there is as yet no evidence of alteration of MGMT promoter methylation after radiotherapy, it is possible that some of the changes observed in our study are attributable to radiotherapy.

As demonstrated in Table 1, the majority of patients were treated with alkylating agent chemotherapy in the interval between surgeries. It would be expected that tumor cells within the greater tumor population with a methylated MGMT promoter, and thus silencing of MGMT expression, would be more sensitive to this form of treatment. These cells may be killed by the treatment, while those relatively resistant cells with an unmethylated MGMT promoter may survive. Analysis of the MGMT promoter in patient E where the tumor initially showed some methylation (12%) but on a later sample was entirely unmethylated may reflect this. Additionally, the MGMT protein acts in a suicide manner, that is it may only remove an alkyl group from the O⁶ position of guanine on one occasion prior to degradation [28, 29]. MGMT protein may be depleted by alkylating agent treatment, and as reflected by immunohistochemistry, without alteration in methylation status. Indeed this property has recently been exploited in clinical trials utilising metronomic temozolomide dosing regimes [30].

As tumors develop, regions of both hypermethylation and hypomethylation are known to develop [31, 32]. As a tumor progresses, irrespective of treatment, the inevitable changes in methylation state that occur may have been reflected in the results demonstrated above. For example, the progression in patient A from glioblastoma to gliosarcoma may have been associated with the development of a generalised hypermethylated state, with methylation of tumor suppressor genes leading to increased tumor aggressiveness despite a potential increased sensitivity to temozolomide which methylation of the MGMT promoter could confer. This may explain why a tumor may become more methylated after radiotherapy and alkylating agent chemotherapy.

The presence of normal tissue in a tumor sample may have some impact on methylation assessment. As most normal astrocytes, microglia, lymphocytes and endothelial cells generally have an unmethylated MGMT promoter this may result in an underestimation of methylation in a tumor. This could manifest as an apparent change in methylation status if there were varying degrees of “normal tissue contamination” in two or more samples. The extent of inflammation varies widely between various glioblastomas and the presence of DNA from inflammatory cells is difficult to control for in methylation analysis.

This study confirms that MGMT promoter methylation is largely homogenous across glioblastomas. While the value of MGMT promoter methylation as a predictive marker has been demonstrated at initial biopsy, the findings of our study suggest that after treatment it may be necessary to repeat this analysis if further treatment decisions are to be based on methylation analysis.