Inactivation of the DNA Repair Gene O⁶-Methylguanine-DNA Methyltransferase by Promoter Hypermethylation is a Common Event in Primary Human Neoplasia¹

MGMT³ is a DNA repair protein that removes mutagenic and cytotoxic adducts from O⁶-guanine in DNA (1). Alkylation of DNA at the O⁶ position of guanine is an important step in the formation of mutations in cancer, primarily due to the tendency of the O⁶-methylguanine to pair with thymine during replication, resulting in a conversion of guanine-cytosine to adenine-thymine pairs in DNA (2). Furthermore, the O⁶-alkylguanine-DNA adduct may cross-link with the opposite cytosine residues, blocking DNA replication (3). MGMT protect cells against these lesions, transferring the alkyl group from the O⁶-guanine in DNA to an active cysteine within its own sequence in a reaction that inactivates one MGMT molecule for each lesion repaired (1). Thus, the ability of a cell to withstand such damage is directly related to the number of MGMT molecules it contains and to the rate of de novo synthesis of MGMT.

The amounts of MGMT protein differ according to cellular type (4, 5) and are decreased in some tumors, with respect to their normal tissue counterpart (5, 6). A subset of tumor cell lines, termed Mer-, lack MGMT activity (7). Because loss of expression is not commonly due to deletion, mutation or rearrangement of the MGMT gene (7, 8, 9), or mRNA instability (10), other causes for loss of activity may be involved. Hypermethylation of normally unmethylated CpG islands in the promoter regions of many genes correlates with loss of transcription (11), and the human MGMT gene possesses a CpG island (12). Recent work has reported that methylation of discrete regions of the MGMT CpG island is associated with the silencing of the gene in cell lines (13, 14, 15). To study the relevance of the promoter hypermethylation of the MGMT gene in vivo, we have examined a series of more than 500 primary human tumors and corresponding normal tissues for MGMT aberrant methylation using MSP. Our results indicate that, in certain human cancers, MGMT function is lost frequently in association with hypermethylation of the promoter region and that this event may be an important step in human tumorigenesis.

DNA methylation patterns in the CpG island of the MGMT gene were determined by chemical modification of unmethylated, but not the methylated, cytosines to uracil and subsequent PCR using primers specific for either methylated or the modified unmethylated DNA (16). Primer sequences of MGMT were for the unmethylated reaction 5′-TTTGTGTTTTGATGTTTGTAGGTTTTTGT-3′ (upper primer) and 5′-AACTCCACACTCTTCCAAAAACAAAACA-3′ (lower primer) and for the methylated reaction 5′-TTTCGACGTTCGTAGGTTTTCGC-3′ (upper primer) and 5′-GCACTCTTCCGAAAACGAAACG-3′ (lower primer). The annealing temperature was 59°C. Placental DNA treated in vitro with Sss I methyltransferase (New England Biolabs) was used as positive control for methylated alleles of MGMT, and DNA from normal lymphocytes was used as negative control for methylated alleles of MGMT.

Briefly, 1 μg of DNA was denatured by NaOH and modified by sodium bisulfite. DNA samples were then purified using Wizard DNA purification resin (Promega), again treated with NaOH, precipitated with ethanol, and resuspended in water. Controls without DNA were performed for each set of PCRs. Each PCR reaction (10 μl) was directly loaded onto nondenaturing 6% polyacrylamide gels, stained with ethidium bromide, and visualized under UV illumination.

Sections of formalin-fixed, paraffin-embedded tissue (6 μm thick) were deparaffinized with xylenes for 30 min and dehydrated by using graded ethanols. Immunoperoxidase staining using diaminobenzidine as chromogen was performed with the TechMate 1000 automatic staining system (Ventana, BioTek Solutions, Tucson, AZ). Commercially available Mouse anti-MGMT monoclonal antibody (catalogue N99200; Novus Molecular Inc., San Diego, CA), previously referred to as mT3.1 (17), at 1:100 dilution was used. This antibody has previously been demonstrated to be useful for immunohistochemistry and to correlate with O⁶-alkylguanine-DNA alkyltransferase activity (18). Nuclear staining was determined by two authors (S. R. H. and P. C. B.) who did not have knowledge of the molecular analysis of those samples.

DNA obtained from 524 primary human tumors, cancer cell lines, and normal tissues was subjected to MGMT promoter methylation study using MSP. The region chosen for MGMT spans the area of greatest CpG density immediately 5′ to the transcription start site, in an area previously found to be hypermethylated in Mer- cell lines (14, 15). To determine the specificity of our approach, the colon carcinoma cell lines SW48 and HT29, reported previously as a Mer-deficient (Mer-) line and proficient (Mer+) line (19), respectively, were first studied. The SW48 cell line, which did not express MGMT, was fully methylated in this region, whereas the MGMT-expressing cell line HT29 was completely unmethylated (Fig. 1,B). The MGMT promoter methylation status was investigated further in a set of normal human tissues, including brain, colonic mucosa, breast, pancreas, lung, and peripheral blood lymphocytes. Among these normal tissues tested, all were completely unmethylated at the MGMT gene (Fig. 1 A). Thus, the methylation of MGMT identified in cell lines must have originated during tumor progression or in cell culture.

To address the relevance of the promoter hypermethylation of the MGMT gene in vivo, we examined a large series of primary human tumors for aberrant MGMT methylation using MSP. The results for the primary tumors with MGMT promoter hypermethylation are summarized in Table 1, and examples are shown in Fig. 1. Hypermethylation of MGMT was detected in many tumors. Aberrant MGMT hypermethylation was most frequent in primary gliomas and colorectal carcinomas, showing aberrant methylation in ∼40% of the cases. Among the gliomas and colorectal carcinomas tested, 54 of 140 (38%) and 14 of 36 (38%), respectively, had MGMT promoter hypermethylation. Aberrant methylation of MGMT in gliomas was found in similar frequencies across the spectrum of grades: astrocytoma/oligodendroglioma (grade II) had hypermethylation in 8 of 26 (31%), anaplastic astrocytoma (grade III) in 10 of 20 (50%), and glioblastoma multiforme (grade IV) in 36 of 87 (41%). In contrast to the subtypes described above, the promoter of MGMT was unmethylated in meningiomas (n = 25), pilocytic astrocytomas (n = 4) and ependymomas (n = 3). A single medulloblastoma examined was hypermethylated at MGMT. We also examined cell lines of these tumor types and generally found a similar incidence of MGMT promoter methylation as seen in primary tumors, with two of four (50%) gliomas and three of seven (40%) colon cancer cell lines having aberrant methylation at the MGMT CpG island.

A second group of neoplasms with frequent MGMT promoter methylation included lymphomas, NSCLCs, and head and neck carcinomas, where this alteration was found in ∼25% of the cases (Table 1). In the non-Hodgkin’s lymphomas studied, 15 of 62 (24%) were hypermethylated at MGMT. The presence of aberrant promoter methylation was a common feature in aggressive/high-grade lymphomas where 11 of 14 (79%) were hypermethylated, including all 9 diffuse large cell lymphomas analyzed. In contrast, this MGMT epigenetic alteration was uncommon in indolent/low-grade lymphomas, where only 2 of 22 (9%) cases had MGMT methylation.

Hypermethylation of the MGMT promoter was not a universal finding in cancers in general, however. In contrast to the tumor types described above, other tumor types had infrequent MGMT promoter methylation: 2 of 18 (11%) pancreatic carcinomas, 2 of 18 (11%) melanomas, 1 of 12 (8%) renal carcinomas, 2 of 31 (6%) acute leukemias (1 AML and 1 ALL), and 2 of 44 (4%) bladder carcinomas. Finally, no carcinomas of the breast (n = 36), endometrium (n = 17), ovary (n = 23), or SCLCs (n = 7) analyzed had aberrant MGMT methylation.

In primary tumors, unlike cancer cell lines that were often either completely methylated or completely unmethylated in this region of the MGMT CpG island (Fig. 1,B), hypermethylation of MGMT was always accompanied by amplification in the unmethylated reaction as well. The presence of this unmethylated MGMT region could indicate the presence of normal tissues in these nonmicrodissected samples, with unmethylated MGMT alleles, as observed in Fig. 1 A. However, heterogeneity in the patterns of methylation in the tumor itself might also be present. To address this issue, we sought to determine the consequences of hypermethylation of the CpG island of MGMT on expression of this enzyme.

Levels of MGMT expression often vary in primary tumors (5, 6). MGMT activity in primary tumors may be affected by contamination of tumor tissue with normal cells, such as endothelial cells, reactive astrocytes, or infiltrating lymphocytes, which express MGMT. Previous studies comparing MGMT activity, Western blot analysis, and immunohistochemistry have demonstrated only a moderate association between these methods of determining MGMT expression in colon carcinoma and normal colon (20). For these reasons, we analyzed MGMT expression by immunohistochemistry so that the expression in the neoplastic cells could be directly measured, using a previously characterized antibody (18). We first examined cell lines with known MGMT expression and methylation patterns to validate this approach. Expression of MGMT was absent in SW48 (Mer- and hypermethylated at the MGMT promoter), but was present in the Mer+ colon cancer cell lines HT29 and LoVo (data not shown).

Paraffin sections from 31 human primary tumors (including 8 brain tumors, 9 colon carcinomas, and 14 lymphomas) were then studied for MGMT expression. The immunohistochemical data related to MGMT promoter hypermethylation are shown in Table 2, and examples of the results obtained are shown in Fig. 2. All sections examined had nuclear staining of MGMT in normal cells adjacent to or within tumors, although in some cases this staining was weak. This provided an internal positive control. Among the 31 total tumors analyzed, 13 were judged to have loss of MGMT expression, whereas 18 tumors expressed MGMT. The level of expression observed in tumors varied greatly, and tumors that expressed MGMT often demonstrated higher expression than adjacent normal tissue. In the majority of these high MGMT expressing tumors, the expression was homogeneously increased. However, some cell to cell variation within in the neoplastic cells was observed. This finding is consistent with previous studies in these and other tissue types that tumor formation may be associated with increased levels of MGMT (20, 21, 22, 23). However, a subset of tumors had diminished or absent MGMT expression, and in these tumors, protein was absent in the majority of tumor cells, with only occasional cells expressing detectable MGMT protein. The loss of staining was most uniform in the lymphomas, with greater cellular heterogeneity in the colon and brain tumors. Similar heterogeneity has been reported in colon (20) and lung (23) carcinomas demonstrating decreased expression of MGMT measured by immunohistochemistry.

When the expression data were matched to previous studies of the methylation of MGMT, 12 of 13 tumors that lack MGMT expression (92%) showed MGMT promoter hypermethylation, whereas 17 of 18 (94%) tumors that retained expression of MGMT were unmethylated at the MGMT CpG island (Fisher’s exact test, <0.0001). This association between loss of MGMT expression and promoter hypermethylation was seen in each of the tumor types examined (Table 2). Among eight gliomas (five glioblastomas multiforme, two anaplastic astrocytomas, and one pilocytic astrocytoma), three had loss of protein expression and MGMT promoter hypermethylation, whereas five retained MGMT expression and were unmethylated at MGMT. Among nine primary colorectal carcinomas, three had lack of MGMT expression associated with abnormal MGMT methylation, whereas six of seven (85%) expressing tumors were unmethylated at MGMT. In all cases, adjacent normal colon epithelia expressed the MGMT protein. In two colonic carcinoma cases, we were able to examine MGMT expression in both primary and metastatic lesions. A hepatic metastasis from a primary tumor hypermethylated at MGMT, which lacked expression of the protein, also demonstrated loss of MGMT expression, whereas a lymph node metastasis from an unmethylated primary tumor, which expressed MGMT, retained MGMT expression in the metastasis. Finally, the analysis of 14 lymphomas (8 diffuse lymphomas, 5 follicular lymphomas, and 1 marginal zone lymphoma) revealed that among seven tumors showing loss of MGMT expression, six had MGMT promoter methylation, whereas all seven expressing tumors were unmethylated at the MGMT loci.

In cell culture, a causal relationship between methylation of the MGMT CpG island and transcriptional silencing of the gene has been demonstrated (13, 14, 15). Our data now demonstrate that MGMT promoter hypermethylation associated with loss of expression is frequent in primary human neoplasia as well. The tight correlation between MGMT CpG island methylation and loss of expression in these tumors provides an explanation for the loss of MGMT activity previously reported in a subset of many tumor types (5, 6, 20, 21, 23, 24). However, in tumor types where we did not detect MGMT hypermethylation, such as breast, ovarian carcinoma, or meningioma, decreased MGMT activity is infrequent (5, 6, 24).

Because MGMT plays a major role in the repair of the O⁶-methylguanine DNA adducts, which are formed after exposure to methylating agents [including the monofunctional alkylating agents MNU or MNNG (N-methyl-N’-nitro-N-nitrosoguanidine)] and to clinical drugs such as dacarbazine, procarbazine, temozolomide, lomustine [CCNU, 1-(2-chloroethyl)-3-cyclohesyl-1-nitrosourea], nimustine [ACNU, 1-(4-amino-2-methyl-5-pyrimidinyl)methyl-3–2(2-chloroethyl)-3-nitrousourea], and carmustine [BCNU, 1,3-bis(2-chloroethyl)-1-nitrousourea], our results identify potentially critical events in carcinogenesis and may define a set of chemosensitive tumors. During DNA replication, O⁶-methylguanine can pair with thymine, resulting in conversion of guanine-cytosine to adenine-thymine pairs in DNA (2). Such mutations are often present in tumors induced by alkylating agents (25). To counteract these effects, MGMT protein can remove such adducts before mutations result. Thus, MGMT gene knockout mice have a higher susceptibility to MNU-induced tumorigenesis, specifically manifesting more frequent lung adenomas and thymic lymphomas when compared with wild-type mice (26). Our findings that primary NSCLCs and lymphomas are tumor types with frequent methylation-associated inactivation of the MGMT gene suggest that this epigenetic alteration may confer an increased risk of alkylating agent-induced carcinogenesis in the systems of these organs.

The importance of MGMT in carcinogenesis is also evident in the MGMT transgenic mice, who possess protection against skin carcinogenesis, thymic lymphomas, and aberrant crypt foci of the colon induced by alkylating agents (27, 28, 29). Interestingly, simple methylating agents, such as MNU, ACNU (a CNU derivative), and NNK [4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone], a nitrosamine present in tobacco and tobacco smoke, have been shown to induce ras mutations (30, 31, 32). Tumor types with silencing of the MGMT by aberrant promoter hypermethylation in this study include those that have a frequent rate of K-ras mutations described in the literature, such as colon, lung, and head and neck carcinomas (30). This suggests that one potential consequence of loss of MGMT expression could be an increase in susceptibility to K-ras mutation. The MGMT transgenic mouse, in fact, is not only protected against aberrant crypt foci formation, but also has a lower incidence of crypt foci with G to A mutations in K-ras (29). However, other tumors associated with carcinogen exposure, such as bladder carcinoma, and other tumors with a high rate of K-ras mutations, such as pancreatic carcinoma, were not frequent targets of MGMT promoter hypermethylation, suggesting that inactivation of MGMT is not required for the acquisition of K-ras mutations.

From a clinical standpoint, MGMT promoter hypermethylation with subsequent loss of MGMT expression might play a role in modulating chemosensitivity to alkylating agents. CNUs and methylating agents, such as streptozotocin, are commonly used in the treatment of neoplastic diseases such as brain tumors, colon carcinoma, malignant melanoma, Hodgkin’s disease, lymphoma, myeloma, and skin cancer. Among the tumors studied in the present study, gliomas, colon carcinomas, and lymphomas had the highest frequency of MGMT promoter hypermethylation. These tumor types, in addition to NSCLC, have all been reported to have decreased MGMT expression and activity (5, 6, 20, 21, 23, 24, 33) in a subset of tumors. Thus, the decreased MGMT activity observed in a significant fraction of these tumors and the response of the former tumor types to CNUs may be due, in part, to the lack of O⁶-chloroethylguanine repair by inactivation at MGMT by aberrant promoter methylation. The tight correlation between loss of expression at the protein level and hypermethylation of the MGMT 5′ CpG island we report in our study directly links these two processes in vivo and suggests an important role for aberrant promoter hypermethylation of MGMT in primary human cancer.

We acknowledge numerous colleagues who provided primary tumor samples and cell lines for this study, most notably Dr. Adrian Merlo. We also thank Dr. Anthony Pegg for helpful discussion.