Prognostic Significance of O⁶-Methylguanine-DNA Methyltransferase Determined by Promoter Hypermethylation and Immunohistochemical Expression in Anaplastic Gliomas

Anaplastic gliomas (WHO grade 3) show a wide variability of clinical outcome. Despite optimal treatment, mainly consisting of gross total resection followed by radiotherapy and chemotherapy with alkylating agents (1), therapeutic response and survival times vary considerably. This fact suggests that a large number of factors, including patient, tumor, and treatment characteristics, may influence the outcome (2–4).

Alkylating agents cause cell death by forming cross-links between adjacent strands of DNA due to alkylation of the O⁶ position of guanine. The cellular DNA repair protein O⁶-methylguanine-DNA methyltransferase (MGMT) inhibits the cross-linking of double-stranded DNA by removing alkylating lesions (5–7). A direct relationship between MGMT activity and resistance to alkylating nitrosoureas and methylating agents (i.e., ionizing radiations) has been well documented in cell lines and xenografts derived from a variety of human tumors, including gliomas (8). Moreover, depletion of MGMT activity with the substrate analogue inhibitor O⁶-benzylguanine increases the rate of cell death by alkylating nitrosoureas in human glioma-derived cell lines (9).

MGMT is ubiquitously expressed in normal human tissues, although its levels vary considerably between organs and individuals. Remarkably, MGMT activity is usually higher in malignant tissues than in their normal counterparts (6, 10). However, tumors have long been noted to be heterogeneous with respect to MGMT expression; interestingly, in a subset of cancer cells, its expression is silenced mostly due to abnormal promoter methylation (7, 10–12). Among human neoplasms, gliomas present a high frequency of MGMT inactivation by promoter hypermethylation, occurring in almost 30% of them (6, 13, 14).

Some clinical studies suggest that MGMT inactivation by aberrant promoter methylation correlates with sensitivity to methylating agents in gliomas (15) and, eventually, to an improved prognosis (16–19). However, other clinical studies failed to show any relationship between MGMT methylation status and overall survival or response to alkylating agents in patients with gliomas (20–23). These contradictory results could be due to the fact that most previous studies evaluated heterogeneous groups of patients, including both those with anaplastic glioma and glioblastomas multiforme (7, 12, 16, 17, 20, 22–28). Because of the different biological behavior of these two types of tumors and the lower incidence of the former, it is plausible that the real benefit of the absence of MGMT in anaplastic gliomas was masked. In that sense, it is important to point out that although malignant, anaplastic gliomas have a better prognosis and a higher likelihood of response to chemotherapy than glioblastomas. Accordingly, only patients with histologic diagnosis of anaplastic glioma were included in the present study.

There are very few data concerning the correlation between MGMT promoter methylation and MGMT protein expression in gliomas, and contradictory results have also been reported (6, 29). Similarly, there are a limited number of reports using immunohistochemistry for the detection of MGMT in gliomas (23, 27, 30). Immunohistochemistry is a simple method for the detection of MGMT protein expression and it can be easily done even on archival paraffin-embedded specimens without using sophisticated equipment. In addition, it allows for the evaluation not only of the degree of staining but also the localization of the target factor in individual cells.

Taking into account all the above-mentioned considerations, the aim of this study was to evaluate the prognostic significance of MGMT in patients with anaplastic glioma. For this purpose, both MGMT promoter methylation and protein expression were analyzed in a series of patients with newly diagnosed WHO grade 3 glioma and managed according to a common diagnostic and therapeutic protocol.

Patients and tissue collection. Ninety-three patients admitted to the Neurosurgical Departments of Bellvitge Hospital and Hospital Clinic of Barcelona, between January 1993 and December 2001, were included. Seventy-two of these patients were reported in a previous study in which clinical prognostic factors were evaluated (4). Tumor samples were obtained during surgical treatment, formalin-fixed and paraffin-embedded for histologic studies and, in some cases, frozen and stored at −80°C until processing. Sixty patients were males (64.5%) and the median age was 48 years (range, 16-74). Sixty-two patients (66.7%) underwent resective surgery (34 gross total resection and 28 partial resection) and 31 patients (33.3%) biopsy. All tumors were histologically verified by two independent neuropathologists (IF, TR) and classified according to WHO (2). Seventy-five patients (80.6%) had anaplastic astrocytoma, whereas 18 (19.4%) had tumors with an oligodendroglial component (4 anaplastic oligodendroglioma and 14 anaplastic oligoastrocytoma). Informed written consent was obtained from all patients, and tissue collection was approved by each Institutional Review Board.

Patients were managed according to a previously established common diagnostic and therapeutic protocol, including adjuvant radiotherapy and chemotherapy after surgical resection. Radiotherapy consisted of focal cranial irradiation with a margin on the order of 2 to 3 cm surrounding the tumor volume. A high-energy and rigid immobilization system was used, and a dose of 60 Gy was given in standard daily fractions of 2 Gy. Chemotherapy consisted of carmustine [1,3-bis(2-chloroethyl)-1-nitrosourea] for patients with anaplastic astrocytoma, or the PCV regimen (procarbazine, lomustine, and vincristine) for those with tumors with oligodendroglial component. All patients received radiotherapy, but 21 of them (22.6%) did not receive chemotherapy because of disease progression and severe clinical deterioration during radiotherapy (n = 19), or rejection of treatment (n = 2). Tumor response was evaluated by cranial magnetic resonance imaging (MRI) and classified according to Macdonald's criteria (31). Patient follow-up was obtained using clinical chart review and was approved by the various institutional review boards.

Immunohistochemistry procedures. Immunohistochemistry was carried out following the streptavidin-biotin-peroxidase (LSAB) method (DAKO LSAB2 System, Peroxidase; DAKO, Carpinteria, CA) according to a previous published protocol (4). The mouse monoclonal antibody (MGMT Ab-1; clone MT 3.1, Neomarker, Westinghouse, Fremont, CA) was diluted 1:20 and the mouse Ki-67 antibody (DAKO) was used at a dilution of 1:25. Sections were slightly counterstained with hematoxylin. For MGMT immunohistochemistry, human liver was used as positive control and tonsil tissue served as positive control for Ki-67 antibody. Negative controls were done omitting the primary antibody.

Each slide was individually reviewed and scored by two observers (MB, AT). Fifteen to 20 fields at ×400 magnification were analyzed per specimen. The immunoreactivity of MGMT protein was evaluated semiquantitatively by estimating the fraction of positive cells and a level <5% was regarded as negative, 5% to 25% as low reactivity, 25% to 50% as moderate reactivity, and >50% as high reactivity. Only nuclear staining was considered for grading. Ki-67 scoring was accomplished by determining the percentage of positive nuclei from regions of maximal nuclear staining after counting 1,000 tumor cells, or as many cells as possible in the case of small specimens at ×400 magnification. Cells were counted as Ki-67 positive if diffuse nuclear staining was present. All immunohistochemical analysis was carried out blind to the clinical information.

Methylation-specific PCR. Genomic DNA was isolated from paraffin embedded samples and, when possible, from frozen tumor. DNA methylation status of CpG islands at the MGMT promoter was determined by methylation-specific PCR (MSP) as previously described (6), with some modifications. For PCR amplification, previous reported specific primer sequences were used: 5′-TTTGTGTTTTGATGTTTGTAGGTTTTTGT-3′ (forward primer) and 5′-AACTCCACACTCTTCCAAAAACAAAACA-3′ (reverse primer) for the unmethylated reaction and 5′-TTTCGACGTTCGTAGGTTTTCGC-3′ (forward primer) and 5′-CGACTCTTCCGAAAACGAAACG-3′ (reverse primer) for the methylated reaction (17). The annealing temperature was 59°C. All reactions were done twice to exclude unspecific PCR amplifications. Low-quality DNA yielding uncertain PCR results was discarded. Normal human lymphocyte DNA was used as negative control for methylated alleles of MGMT, and Placental DNA treated in vitro with SssI methyltransferase (New England Biolabs, Beverly, MA) was used as positive control. Controls without DNA were used for each set of MSP assay. PCR products were separated on 8% polyacrylamide gels, stained with ethidium bromide, and examined under UV illumination. Investigators doing these assays were blinded to clinical information.

Statistical methods. The relationship between tumor MGMT expression and MSP results, as well as their correlation with clinicopathologic variables, were evaluated by the χ² and the Fisher's exact test. The length of follow-up was described as the median and range. The effect of single variables on overall survival and progression-free survival was determined by both univariate and multivariate analyses. Overall survival was calculated from surgery to the date of death or last follow-up visit. Progression-free survival was counted from surgery to the date of tumor relapse or progression. Probabilities of overall survival and progression-free survival were calculated according to the Kaplan-Meier method and compared with the log-rank test. Tumor MGMT expression and MGMT promoter methylation status, along with demographic (age and gender), clinical (postoperative Karnofsky Performance Status [KPS]), radiological (ring enhancement), pathologic (presence of oligodendroglial component and proliferative index), and therapeutic (extent of resection) variables achieving a P < 0.2 in the univariate analysis, were subsequently introduced in a backward stepwise proportional hazard analysis (Cox model) to identify independent predictors of survival. For continuous variables, the cutoff level chosen was their median value. MGMT immunohistochemistry was reclassified as negative and positive (including low, moderate, and high staining) for statistical purposes. To minimize any potential bias, patients who did not adhere to the established treatment protocol were not excluded from the analysis (32). Nevertheless, a second analysis including those patients receiving adjuvant chemotherapy was also done. All statistical analysis was done at a significance level of P = 0.05, using the statistical package SPSS 11.0 (SPSS, Inc., Chicago, IL).

MGMT protein expression and MGMT promoter methylation status. MGMT positive immunostaining was detected in 51 tumors (54.8%). Nineteen tumors showed low, 20 moderate, and 12 high MGMT immunoreactivity. In all positive tumor samples, heterogenous immunostaining was observed; areas with complete loss of MGMT expression alternated with areas of scattered or clustered cells with strong immunoreactivity.

After specific MSP, DNA was correctly amplified in 40 out of all analyzed samples. These 40 cases did not statistically differ from the rest of the series in any of the evaluated variables (data not shown). MGMT promoter hypermethylation was detected in 20 of these 40 patients (50%). In all methylated samples, signals of unmethylated DNA were also present (Fig. 1).

In four patients, tumors showed lack of MGMT protein expression and promoter methylation, and 13 tumors with unmethylated profile showed MGMT immunoreactivity. However, in 16 cases in which MGMT promoter was aberrantly methylated, some degree of MGMT immunoreactivity was detected. In addition, seven cases with complete loss of MGMT expression lacked MGMT promoter hypermethylation. Accordingly, no correlation between MGMT protein expression and MGMT promoter methylation was observed (P = 0.29).

Neither MGMT protein expression nor MGMT promoter methylation status were related to any clinicopathologic variables, except for gender, for which the incidence of unmethylated tumors was found significantly associated to male gender (P = 0.004). Importantly, no differences were observed in either MGMT protein expression or promoter methylation status with respect to adjuvant chemotherapy (Table 1).

Overall survival. After a median follow-up of 89.7 months (range, 28-131), 28 patients (30.1%) remained alive. Median survival was 18 months (range, 1.6-131). In the univariate analysis, factors influencing overall survival were age, postoperative KPS, proliferative index, contrast enhancement on MRI, and extent of resection. There was a trend toward a longer overall survival for those patients whose tumors had an oligodendroglial component. The Cox's regression model revealed that postoperative KPS of ≥80, proliferative index of ≤4.8%, absence of contrast enhancement, and gross total resection were independently associated with a longer survival. Lack of MGMT immunoreactivity showed a tendency to an increased overall survival (Table 2). Finally, when a second analysis including only those patients that were actually treated with adjuvant chemotherapy (n = 72) was done, the absence of MGMT immunostaining was identified as an independent prognostic factor, along with postoperative KPS of ≥80, proliferative index of ≤4.8%, and gross total resection (Table 3; Fig. 2).

Progression-free survival. At the end of follow-up, 34 patients (36.6%) remained free of progression. Statistically significant prognostic factors on the univariate analysis were age, postoperative KPS, proliferative index, presence of an oligodendroglial component, absence of contrast enhancement on MRI, and extent of resection. As shown in Table 2, the multivariate analysis showed that postoperative KPS of ≥80, presence of an oligodendroglial component, proliferative index of ≤4.8%, absence of contrast enhancement on MRI, and gross total resection were independently associated with longer progression-free survival. When a secondary analysis of patients receiving adjuvant chemotherapy was done, MGMT expression was not independently associated with progression-free survival (Table 3).

This study represents, to our knowledge, the first investigation aimed at evaluating the prognostic significance of MGMT in a series of patients with newly diagnosed WHO grade 3 gliomas uniformly treated, in which both tumor MGMT expression and MGMT promoter methylation status were simultaneously evaluated. Our results show that absence of tumor MGMT expression is independently associated with a longer overall survival in patients with anaplastic gliomas who received adjuvant chemotherapy, whereas aberrant MGMT promoter methylation has no prognostic implications. The strength of these results relies mainly on the fact that MGMT predictive value has been evaluated after adjusting this variable for well-recognized clinicopathologic prognostic factors.

The proportion of tumors exhibiting either lack of MGMT protein immunoreactivity or MGMT promoter hypermethylation did not differ from previously reported studies (6, 8, 16–19, 22, 27, 33–39). However, in the present study, there was an inconsistent correlation between aberrant promoter methylation and loss of protein expression. Very few studies have investigated the relationship between MGMT promoter hypermethylation and protein expression in gliomas and contradictory results were observed (6, 38, 40). Moreover, this inconsistency was also observed in other neoplasms (39, 41) and was not limited to the MGMT gene (42–44). Whereas methylation is clearly involved in the inactivation of MGMT gene in numerous tumors and cancer cell lines (5, 6, 8, 11, 36), regulation of MGMT expression is a more complex phenomenon in which abnormal methylation of the promoter is not the only determining factor (5, 38, 39, 41, 45, 46). Indeed, several studies indicate that grade of methylation both in the promoter region and in neighboring sequences may regulate gene expression (40, 43, 46, 47). Furthermore, MSP is a highly sensitive qualitative technique in which a methylated band may be observed even if cells that carry MGMT promoter hypermethylation represent only a minor portion of the tumor (5).

A heterogeneous pattern of MGMT immunoreactivity was observed among different regions of the same tumor, ranging from no immunostaining to strong immunoexpression. Differences in the methylation status of the MGMT promoter in tumor cell subpopulations may explain this heterogeneity. In fact, most of the methylated tumors in the present study also exhibited a profile of unmethylation. Whereas the presence of contaminating normal cells may not be ruled out (6), other potential explanations for this variability include monoallelic promoter methylation, methylation of a small proportion of malignant cells or loss of heterozygosity in 10q26 (33, 36, 38, 44).

The influence of MGMT promoter methylation status and MGMT protein expression on the prognosis of patients with gliomas has been evaluated in several studies with contradictory results (Table 4). Most previously published data were obtained from heterogeneous groups of patients with different grades and histologies, each with its own natural course and treatment response. In addition, many of these studies included distinct treatment protocols. Some studies have shown that MGMT promoter methylation is associated with improved time to progression or overall survival, either in the whole series or in subset analyses (16–19). Conversely, other investigators have found that median progression-free survival was longer in patients with unmethylated gliomas (34), whereas no association between MGMT methylation and prognosis was reported (22, 33). In the present study, we also failed to show this correlation, although we could not definitively ruled out the absence of prognostic value of aberrant MGMT promoter because of the relative low number of specimens analyzed. With respect to tumor MGMT protein expression determined by immunohistochemistry or quantitative immunofluorescence, results are also contradictory (7, 23, 24, 27). Although differences in the study designs could explain, at least in part, these contradictory results, other possibilities should be considered. In that sense, the observation of a longer survival in patients receiving chemotherapy whose tumors did not express MGMT may reflect a difficulty of cancer cells in correcting DNA damage induced by chemotherapy, thus leading them to death. On the other hand, loss of MGMT expression could also be a negative prognostic factor because of an increased susceptibility to acquiring other mutations (48–50). Finally, it is important to keep in mind that hypermethylation can simultaneously occur in other genes with a CpG island in their promoter regions (51), making it more difficult to predict the final behavior of tumor cells.

The prognostic significance of MGMT protein expression or MGMT promoter methylation in glioma patients will remain controversial. The marked heterogeneity in survival among patients with anaplastic glioma is probably not related to the function of a single gene. In that sense, consideration of MGMT as the only relevant molecular factor in tumor chemosensitivity may constitute an oversimplification, because it would imply that two groups of tumors (MGMT positive and negative) are otherwise very similar. Further clinical studies are needed to clarify whether MGMT discriminates among biologically distinct groups of tumors each with a different natural history and response to treatment.

We thank T. Yohannan for editorial assistance.