|Home | About | Journals | Submit | Contact Us | Français|
O6-methylguanine-DNA methyltransferase (MGMT) is a DNA repair gene which is frequently methylated in colorectal cancer (CRC). However, it remains controversial whether methylation of specific CpG sequences within MGMT promoter leads to loss of its protein expression, and if MGMT methylation correlates with G to A transition mutations in KRAS. Two methylation sensitive regions (Mp and Eh region) of MGMT promoter were investigated in 593 specimens of colorectal tissue: 233 CRCs, 104 adenomatous polyps (AP), 220 normal colonic mucosa from CRC patients (N-C) and 36 normal colonic mucosa specimens obtained from subjects without colorectal neoplasia (N-N) by combined bisulfite restriction analysis (COBRA). The region-specific methylation data were compared to the MGMT protein expression, spectrum of KRAS mutations and other clinical features. Extensive (including both Mp and Eh) and partial (either Mp or Eh) MGMT methylation were found in 24.5% and 11.6% of CRCs, 3.8% and 27.9% of APs, 0.5% and 7.7% of C-Ns and 2.8% and 2.8% of N-Ns, respectively. Extensive methylation of MGMT promoter was primarily present in CRCs while partial methylation was common in APs. Extensive methylation of MGMT promoter was associated with loss/reduced protein expression (p < 0.0001), as well as with G to A mutations in KRAS (p = 0.0017). We herein provide first evidence that extensive methylation of MGMT promoter region is essential for methylation-induced silencing of this gene. Our data suggest that MGMT methylation may evolve and spread throughout the promoter in a stepwise manner as the colonic epithelial cells progress through the classical-adenoma-cancer multistep cascade.
O6-methylguanine-DNA methyltransferase (MGMT) is a DNA repair gene which removes promutagenic O6-methylguanine (O6-MeG) residues from DNA and is considered an important predictive factors for chemoresistance in human cancers.1–5 Loss of MGMT function permits the increased accumulation of O6-MeG in DNA, which promotes tumorigenesis through G:C to A:T transition mutations in growth regulating genes such as KRAS and p53.6–9 Mutations in MGMT have rarely been found, and it has been suggested that MGMT inactivation is primarily manifested through hypermethylation-induced silencing of its promoter in human cancers, including those of the colon and rectum.5,9–14 However, the associations between MGMT methylation and G to A transition mutations has not been consistently reproduced in different studies.15,16 Central to this controversy is the fact that there is a lack of clear understanding for precise relationships between methylation of specific MGMT promoter regions and its relationship with the loss of protein expression.
It is becoming increasingly clear that the distribution of methylated cytosines in the CpG islands of promoters is not uniform, and the regions most important for gene expression, called the “core regions,” are limited to specific sequences within these CpG islands.12,17–20 Studies done on cultured cells have revealed that the candidate core region of the MGMT promoter involves 2 methylation-sensitive regions. The first of these regions is upstream of exon 1, termed the Mp-region, and includes the minimal promoter.
Additionally, it is highly plausible that multistep carcinogenesis in the colon and rectum likely begins in morphologically normal-appearing tissues. However, no studies have thus far investigated a spectrum of colorectal tissues ranging from normal mucosa, adenomatous polyps (AP), and cancer. Accordingly, the aim of the present study was to determine the specific patterns of DNA methylation in the candidate core regions of the MGMT promoter in a complete range of colorectal neoplasia, and to compare this to MGMT protein expression, KRAS mutations, and other clinical features by analyzing methylation level in discrete regions of the MGMT promoter in a series of 593 colorectal tissue specimens.
A total of 233 CRC samples and 104 adenomatous polyps (AP) samples were obtained from Okayama University Hospital and the Chikuba Hospital in Okayama, Japan. From the 233 CRC specimens, 220 samples of adjacent normal mucosa (“normal from cancer patients”, or N-C) were available for methylation assays. We also sampled colonic biopsy specimens from 36 subjects who had no evidence of colorectal neoplasia at screening colonoscopy (“normal from non-neoplastic colons” or N-N). The tumor/node/metastasis classification system was used for cancer staging.25 APs were divided into 2 subsets: advanced polyps (i.e., polyps with high-grade dysplasia, villous architecture, and tubular adenomas ≥1 cm in diameter: 82 specimens) and minor polyps (simple tubular adenomas <1 cm in diameter: 22 specimens). Institutional review board approval and informed consent was obtained from all subjects to use their tissues.
DNA was extracted from CRC, N-C and N-N tissues by standard procedures described previously.5 All DNA from APs were extracted from formalin-fixed, paraffin-embedded archival materials. For each AP, DNA was extracted using a TaKaRa DEXPAT kit (Takara Bio, Shiga, Japan). Two hundred nanogram to 1 μg of genomic DNA was subjected to sodium bisulfite modification. Bisulfite modification was carried by using a CpGenome DNA Modification Kit (Intergen, Purchase, NY).
We designed the combined bisulfite restriction analysis (COBRA) to examine both the “minimal promoter” (Mp-region) and the “enhancer” (Eh-region) of the O6-methylguanine-DNA methyltransferase (MGMT) promoter (Fig. 1a). For optimization of the COBRA, we used DNA from the hepatocellular carcinoma cell line PLC/PRF/5, which is unmethylated at its MGMT promoter. DNA was extracted from PLC/PRF/5 and methylated in vitro by treatment with SssI methylase (New England Biolabs, Ipswich, MA). Varying amounts of the artificially methylated DNA were mixed with 2 μg of PLC/PRF/5 unmodified genomic DNA before bisulfite treatment (Fig. 1b). Bisulfite PCR was carried out in a 25 μl PCR mixture containing 12.5 μl of HotStarTaq Master Mix kit (Qiagen). Primer sequences for “Mp-region” and “Eh-region” for the MGMT promoter region were: (a) Mp-F (5′-GAGGATGYGTAGATTGTTTTAG-3′) and Mp-R (5′-AAACCRAAAACCTAAAA AAAAC-3′), and (b)Eh-F (5′ GTTTTTAGAAYGTTTTGYGTTT-3′) and Eh-R (5′-CCTACAAAACCACTCRAAACTA-3′), generating fragment lengths of 163 and 145 bp, respectively. Conditions for PCR of Mp-region were as follows: 95°C for 15 min; 45 cycles of 95°C for 30 sec, 53°C for 30 sec, and 72°C for 30 sec; and finally, 7 min at 72°C. Conditions for PCR of Eh-region were as follows: 95°C for 15 min; 45 cycles of 95°C for 30 sec, 50°C for 30 sec, and 72°C for 30 sec; and finally, 7 min at 72°C. The PCR products derived from the Mp-region were digested with HhaI (New England Biolabs) at 37°C for 16 hr, and those from the Eh-region were digested with BstUI (New England Biolabs) at 60°C for 16 hr. The digested DNA was separated on 3% agarose gels in 1× TAE buffer and stained with ethidium bromide and SYBR Safe stains (Invitrogen, Carlsbad, CA). We used a Gel Logic 200 Imaging System (Eastman Kodak, Rochester, NY) to perform densitometric analyses on all gels. Band intensities were quantified using Kodak 1D analysis software (Eastman Kodak, Rochester, NY). The methylation levels (ratios of methylated to unmethylated DNA) were determined from the relative intensities of cut and uncut PCR products to quantitate methylation.
We used 5 markers (MINT1, MINT2, MINT31, p16INK4a and p14ARF) to detect CpG island methylator phenotype (CIMP).26 All 5 markers were analyzed by COBRA described previously.12 The methylation status of 5 CIMP markers was analyzed as a categorical variable (positive: methylation level ≥5%, negative: methylation level <5%). CIMP positive was defined when ≥3 of the 5 loci were methylated; CIMP negative was defined when <3 markers were methylated.
Bisulfite-treated genomic DNA samples were amplified with primers specific for the promoter region of the MGMT gene. Primer sequences for the first PCR were: MGMT-F (5′-AGCGGTTTTAGGAGGGGAGAGAT-3′) and MGMT-R (5′-AAACICCTACAAAACCACTCGAAA-3′), and primer sequences for “methylated” and “unmethylated” for the nested PCR were: (a) M-F (5′-TTTAGGCGGAAGTTGGGAAGGCGTC-3′) and MGMT-2R (5′-CCTACAAAACCACTCGAAACTACC-3′), and (b) U-F (5′ TTTAGGCGGAAGTTGGGAAGGTGTT-3′) and MGMT-2R (5′-CCTACAAAACCACTCGAAACTACC-3′), generating fragment lengths of 722 and 511 bp, respectively. Conditions of the first PCR and the nested PCR were as follows: 95°C for 15 min; 45 cycles of 95°C for 30 sec, 58°C for 30 sec, and 72°C for 30 sec; and finally, 7 min at 72°C. The nested PCR products were purified using a QIAquick PCR purification kit (Qiagen) and directly sequenced using a Thermo Sequenase sequencing kit (Amersham, Corp., Piscataway, NJ) and a SQ-5500E Hitachi Automated DNA sequencer.
Microsatellite instability (MSI) analysis and KRAS mutations at codons 12 and 13 were determined using standard protocols as described previously.12
CRCs from 85 patients and 29 APs were examined using immunohistochemical (IHC) analysis for MGMT protein expression analysis. Staining was carried out manually with formalin-fixed paraffin-embedded tissues. Thin (5 μm) sections of representative blocks were deparaffinized and dehydrated using gradient solvents. After antigen retrieval in the citrate buffer (pH 6.0), endogenous peroxidase was blocked with 3% H2O2. Thereafter slides were incubated overnight in the presence of an anti-MGMT monoclonal antibody (clone MT3.1, PharMingen, Freemont, CA; dilution 1:100). A further incubation was carried out with a secondary antibody and the avidin-biotin-peroxidase complex (Vector Laboratories, Burlingame, CA), and then incubated with biotinyltyramide followed by streptavidin-peroxidase. Diaminobenzidine was used as a chromogen and hematoxylin as a nuclear counterstain. Sections with obvious nuclear staining were deemed positive. The only foci of neoplasia that were scored as negative were those for which there was definite evidence of positively staining admixed or surrounding nonneoplastic cells such as normal colonic mucosal cells, lymphocytes or stromal cells.
The methylation status of both regions in the MGMT promoter as determined by COBRA was analyzed as a categorical variable (positive: methylation level >1%, negative: methylation level <1% by using Kodak 1D analysis software). Methylation levels of specimens classified as methylation positive were analyzed as a continuous variable. To determine the association of methylation levels between the Mp and the Eh regions, we calculated the Z-scores based on quantitative COBRA (Z score = [methylation level of each sample – mean methylation level of 593 samples analyzed]/[SD of methylation level of total analyzed samples], Mp-region; mean methylation level = 4.23%; SD = 12.9%, Eh-region; mean methylation level = 3.88%; SD = 11.6%), and expressed these data on a log scale by calculating both Pearson's correlation coefficients (r) and Spearman's rank correlation coefficients (ρ). A correlation between subjects was considered strong when |r| or |ρ| >0.7, but a weaker correlation when |r| or |ρ| were between 0.5 and 0.7. Differences in frequency were evaluated by a χ2-test. All reported p values were 2-sided and a p value <0.05 was considered statistically significant.
COBRA for MGMT was designed to examine the methylation status in 2 discrete regions of the MGMT promoter (the Mp and Eh regions) as shown in Figure 1a. The methylation levels were measured in a quantitative manner, and the lower limit of measurable methylation was >1% which correlated well with the loss of MGMT expression (Fig. 1b). We investigated MGMT methylation in the 593 colonic tissue specimens, including 233 CRCs, 220 N-Cs, 104 APs and 36 N-N samples. Table I shows the results of MGMT promoter methylation analysis in which the results are analyzed as a categorical variable. Using this technique, we found methylation in 36.1% (84/233) of CRCs, 31.7% (33/104) of APs, 8.2% (18/220) of N-C and 5.6% (2/36) of N-N.
We next analyzed the relationships between levels of methylation in both regions of the MGMT promoter in each subset as a continuous variable using Z scores. As shown in Figure 1d, there was lack of correlation between the levels of methylation of the Mp region and that of the Eh region by Z-score in methylated N-Cs and N-Ns (r = 0.1112; p = 0.64, ρ = −0.4422; p = 0.05), as well as in methylated APs (r = −0.2312; p 5 0.2, ρ = −20.5353; p = 0.001). However, there was a significant positive correlation for methylation at Mp and Eh regions in CRCs (r = 0.7028; p < 0.0001, ρ = 0.6669; p < 0.0001). Figure 1e represents the methylation profiles of both Mp and Eh region in 593 specimens analyzed by a categorical variable. As shown in Figure 1e, the level of MGMT methylation was under 5% in almost all of normal colonic mucosal tissues (both N-N and N-C).
To further support and confirm the methylation data obtained by COBRA, we next performed direct sequencing of the nested MSP products from a subset of 33 CRCs and the corresponding N-Cs. This approach allowed us to determine the methylation status of the 65 CpG sites that are located in the −232 bp to +214 bp region of the MGMT promoter, that includes the COBRA region we investigated for methylation (in which the transcription initiation site is +1). The primers we designed allowed distinction between sequences containing methylated and unmethylated CpG doublets in these regions (Fig. 2a). Similar to previous in vitro reports, each unmethylated CpG site in the MGMT promoters of these methylated samples was located between the Mp and Eh regions, i.e., between +28 bp and +70 bp (Fig. 2b). On rare occasions, we noticed that our bisulfite sequencing data were not in complete agreement with the COBRA results. We believe this discrepancy was primarily because of the much higher degree of sensitivity of nested PCR-bisulfite sequencing (0.002%) compared to the less sensitive COBRA (0.5%). Sequencing results demonstrated that methylated alleles did not have a continuum of methylated cytosines in the promoter regions, even when these regions were otherwise highly methylated.
We investigated MGMT protein expression in 85 CRC tissue samples and 29 AP samples. Representative examples of IHC staining results are shown in Figures 3a–d. We categorized tumors into 3 groups based on the IHC results analyzed as a categorical variable: complete loss of MGMT expression, focal loss and positive MGMT protein expression. Using these criteria, we found that among the 17 tumors with complete loss of MGMT expression, 76% of the tumors demonstrated extensive methylation, while 18% showed partial methylation, and 6% were unmethylated. By contrast, among the 90 tumors with positive MGMT staining, only 9% showed extensive methylation, 23% had partial methylation and 68% were not methylated. No significant differences in frequencies of MGMT protein expression were observed in tumors depicting focal loss in the MGMT expression (14–43%; Table II).
Table III presents the correlations between the methylation status of MGMT and the clinical and pathological features in CRCs and APs. In CRCs, extensive methylation was detected more frequently in elderly patients, compared to younger patients. We also examined whether any associations existed between the degrees of methylation in the discrete regions of MGMT promoter in normal mucosa and the age of the patients. However, no such correlations were detected with these analyses (data not shown).
We observed that the frequencies of MGMT methylation (partial or extensive) were significantly higher in stage T1 and T2 tumors when data were compared to T3 and T4 tumors (P < .0001). Similar observation was made when such associations were investigated in stage I–II versus stage III–IV tumors (p = 0.04). Other than this, there were no significant differences between MGMT methylation status and other clinical factors.
Among APs, there were no significant differences between MGMT methylation status and clinical factors. However, in comparison with CRCs, extensive methylation was rarely observed, but whenever present, it was primarily present in advanced APs.
We next compared the MGMT methylation status to MSI and CIMP status. We observed that MGMT methylation frequencies were relatively higher in microsatellite instability-high (MSI-H) CRCs (35% with extensive methylation and 20% with partial methylation), whereas no such differences were present in MSI-low (MSI-L) and microsatellite stable (MSS) CRCs. Additionally, both extensive and partial methylation in the MGMT promoter was frequently present in CIMP positive CRCs (39% and 21%, respectively) compared to CIMP negative CRCs (22% and 10%, respectively; p = 0.01).
We found KRAS mutations in 76 CRCs (33%) and 21 APs (20%). When we compared the complete spectrum of KRAS mutation to MGMT methylation, a statistically significant correlation was observed only for CRCs (χ2 test; p = 0.0001), but not in APs (p = 0.32). Our results suggest that the reduction or loss of MGMT expression was related to extensive methylation in the MGMT promoter. For this reason, we next investigated the association between extensive methylation of MGMT promoter and G to A mutations in the KRAS. In all tumors, 39% (24/61) of tumors with extensive MGMT methylation showed G to A mutations, while these alterations were present in only 20% (11/56) of tumors with partial methylation and 15% (33/220) of tumors with unmethylated MGMT promoter (Table IV, p = 0.0001; extensive versus partial or unmethylated MGMT promoter). We observed that G to A mutations in KRAS significantly correlated with extensive methylation in the MGMT promoter in colorectal tumors.
Our study demonstrates the biological significance of methylation observed in the discrete regions of the MGMT promoter in a large cohort of specimens obtained from the large intestine. Our data demonstrates that increased frequency of MGMT methylation is present in APs (32%) and CRCs (36%), although modest frequency of MGMT methylation may be detected in normal colonic mucosa. Of particular interest, we observed a gradual increase in the MGMT methylation density in the 2 discrete promoter regions from normal-adenoma-carcinoma sequence which has not been described previously. We also noticed that extensive methylation of the MGMT promoter significantly associated with loss or reduced MGMT expression, consequent to appearance of G to A transition mutations in the KRAS gene. Perhaps, another interesting observation from our study was that the pattern of MGMT methylation was dependent on the tumor size/depth of invasion. Partial methylation was frequently observed in the earlier stages of colorectal tumors (APs, T1, and T2 CRCs) but rare in T3 and T4 CRCs, while extensive methylation was primarily abundant in CRCs but rare in APs (Fig. 4). These data suggest that MGMT methylation is not a secondary event, but is central to tumor development in the advancing stages, since 43% of the T3 and T4 CRCs lose MGMT expression due to extensive promoter methylation.
Our data emphasize the relationship between the pattern of promoter methylation and the MGMT protein expression. This has been a limitation of the existing literature and has not been adequately addressed or consistently reproduced because various studies have used one or the other MGMT promoter regions to correlate methylation data with the loss of MGMT protein.9,14,27,28 A previous report raised the controversy whether partial methylation (involving either Mp or Eh region) of the MGMT promoter may result in reduced protein expression in vitro.13 However, our results clearly demonstrate that extensive methylation of the MGMT promoter is required for the resultant loss/reduction of its expression.
It has previously been suggested that the reduction of MGMT protein is necessary for G to A mutations in KRAS.6,8 Our results provide new data to address this somewhat controversial issue for the relationship between KRAS mutations and MGMT methylation.7,9,11,15,16 We confirmed that MGMT methylation was associated with the KRAS mutations in CRCs but not APs. KRAS mutations were rarely found in small APs, but were increasingly frequent in more advanced APs and CRCs.29 Similar phenomenon was observed for extensive methylation in the MGMT promoter. Therefore, KRAS mutations in early stage APs and in approximately 70% of CRCs with KRAS mutations were not the result from MGMT inactivation due to its promoter methylation. However, in CRCs, 40% of G to A mutations in KRAS was the result from MGMT inactivation by extensive methylation. As a result, we propose at least 2 causes of the KRAS mutation; first a mechanism that is independent from MGMT inactivation, and might be an early and common phenomenon in colorectal tumorigenesis; second, it is possible that KRAS mutations may be a cause of MGMT inactivation, but this might be a later event in colorectal tumorigenesis. The biological significance of the simultaneous presence of KRAS mutations and MGMT methylation is difficult to resolve from our present study. These mutations were predominantly present in MGMT methylated CRCs, which suggests that KRAS mutations have a pathogenetic role in CRC tumorigenesis.12 In support of this, a mechanism for RAS mediated epigenetic silencing has been proposed recently that involves a specific but complex pathway necessary to maintain a transformed phenotype in fibroblasts.30
In conclusion, our findings provide a first demonstration that extensive methylation in the core regions of MGMT promoter is necessary for methylation-induced silencing of this gene and consequence to G to A mutations in KRAS and also suggest that MGMT methylation evolves and spreads throughout the promoter in a stepwise normal adenoma carcinoma cascade.
We are grateful to Ms. Naoko Hoshijima for excellent technical assistance and Dr. Minoru Koi for critical reading of this manuscript.
Grant sponsor: The Japanese Ministry of Education, Science, Sports and Culture of Japan; Grant numbers: 12671227, 11671237, 11671240, 14031227; Grant sponsor: National Institutes of Health; Grant numbers: R01-CA72851, R01-CA98572; Grant sponsor: Baylor Research Institute.