Methylation of O–6-Methylguanine DNA Methyltransferase Characterizes a Subset of Colorectal Cancer with Low-level DNA Microsatellite Instability¹

Classifying colorectal cancer on the basis of the presence or absence of mutational bandshifts in microsatellite loci has highlighted two major molecular pathways. The MSI³-positive pathway proceeds with little evidence of loss of heterozygosity, whereas MSS or MSI-negative cancers show a loss of classical tumor suppressor loci. These have been described as“mutator” and “suppressor” pathways, respectively(1).

When a series of colorectal cancers is studied with a panel of markers for the detection of MSI, MSI tumors are distributed bimodally with a breakpoint at around 30% (2, 3). A National Cancer Institute workshop recommended the use of five microsatellite markers(BAT25, BAT26, D5S346, D2S123, and D17S250) to distinguish MSI-H cancers with instability at two or more loci from MSI-L cancers with instability at one locus (4). It was accepted that additional markers would be required to separate MSS and cancers with a low frequency of bandshifts, but it was also noted that MSS and MSI-L cancers were phenotypically similar (4).

There is evidence that MSI-L status has biological significance in colorectal neoplasia, but also that MSI-L lesions may be heterogeneous. A low level of MSI is associated with cancers from individuals with germline mutation of hMSH6 (5). Instability at mononucleotide markers including BAT25 and BAT26is relatively common in cancers from subjects with hMSH6germline mutation (6). By contrast, instability in BAT25, BAT26 and BAT40 was not observed in sporadic MSI-L cancers, which instead implicated markers with di-, tri-, tetra-, and pentanucleotide repeats (2). Similar patterns of microsatellite marker sensitivities have been described in MSI-L polyps including hyperplastic polyps, mixed polyps and serrated adenomas (serrated polyps; Refs. 7, 8, 9), and in MSI-L adenomas from subjects with hereditary nonpolyposis colorectal cancer (10). In serrated polyps, MSI-L is not associated with the loss of expression of hMLH1 or hMSH2 (8), whereas MSI-L adenomas in hereditary nonpolyposis colorectal cancer show a loss of expression congruent with the germline mutation (10).

A feature distinguishing MSI-L and MSI-H colorectal cancers is the high frequency of K-ras mutation in the former (11). Interestingly, the frequency of K-ras mutation appears to be higher in MSI-L than MSS cancers (11). K-rasmutation has also been documented in microscopic foci of hyperplasia known as ACF (12). Hyperplastic ACF may enlarge to become hyperplastic polyps, and a small number may convert to adenoma when still of microscopic size (13). ACF may also show MSI(14) and aberrant gene methylation (15). The demonstration of K-ras mutation and MSI within a particular class of microscopic lesion suggests that the changes are related and may represent early steps in the evolution of MSI-L cancer. This suggestion fits with descriptions of a serrated neoplastic pathway(7).

The DNA repair gene MGMT removes mutagenic adducts from the 0⁶ position of guanine (16). Inactivation of MGMT is associated with genetic mutation in colorectal neoplasms, particularly G to A transitions in K-ras (17). MGMT inactivation occurs through promoter hypermethylation and has been demonstrated in small colorectal adenomas (18). This provides a mechanism to explain both K-ras mutation and low-level MSI. The latter would arise not by inactivation of a DNA mismatch repair gene but by overloading the DNA mismatch repair system. If these premises were correct, one would expect to observe a high frequency of MGMT methylation in the MSI-L subset of colorectal cancers and in precursor lesions.

Ninety paired normal and tumor samples were obtained in a fresh state from colorectal cancers resected at the Royal Brisbane Hospital between 1989 and 2000. All patients provided informed consent. Cancers were classified according to the level of MSI using a panel of markers recommended by a National Cancer Institute workshop (BAT25, BAT26, D5S346, D2S123, and D17S250; Ref. 4) using methods described previously (11). Two or more positive loci indicated a MSI-H tumor, one indicated MSI-L, and no positive markers indicated MSS(4). The compound marker MYCL was also included, but cancers remained classified as MSI-L if bandshifts occurred in MYCL, a single dinucleotide marker, and no mononucleotide markers. The MSI-enriched series included 23 MSI-H, 44 MSI-L, and 23 MSS cancers. No additional criteria were applied to select tumors within MSI subclasses. Six MSI-L cancers were additional to an earlier reported series (11).

Establishment of the methylation status of the MGMT gene used six CpG sites in the promoter region and used methylation-specific PCR (18). This sensitive technique is based on the premise that unmethylated cytosines in bisulfite-modified genomic DNA are converted to uracil bases, whereas methylated cytosines are preserved. The region of interest is then amplified with primers specific to either the methylated DNA or to the modified and unmethylated DNA. The PCR product (10 μl) was visualized on a 10% nondenaturing polyacrylamide gel stained with ethidium bromide. Additional assay details have been published previously (18, 19).

Mutation of the K-ras oncogene was assayed at the two most common “hot spots” for mutation: codons 12 and 13. A modified RFLP approach was adopted (20). Briefly, mismatched primers were used to amplify a 157-bp PCR product that resisted restriction enzyme digestion with BstN1 (New England Biolabs) if a mutation was present in codon 12 or with BglI if a mutation was present in codon 13. Putative mutations were confirmed by manual sequencing using an Ampli-Cycle cycle sequencing kit (Perkin-Elmer).

Formalin fixed, paraffin-embedded cancer tissues (that corresponded with the fresh tissue samples used for assessing methylation of MGMT) were obtained from the files of the Royal Brisbane Hospital Pathology Department. Sections were also prepared from 31 serrated polyps (hyperplastic polyps, mixed polyps, and serrated adenomas) of known DNA microsatellite status. These were either sporadic lesions (7) or had presented in subjects with hyperplastic polyposis (8) or hereditary nonpolyposis colorectal cancer (10). Paraffin sections were affixed to Superfrost Plus adhesive slides (Menzel-Gläser, Braunschweig,Germany). After dewaxing and rehydration to distilled water, the sections were subject to heat antigen retrieval in 0.001 m EDTA (pH 8). Endogenous peroxidase activity sections were blocked using 1.0%H₂O₂, 0.1%NaN₃ in TBS [0.05 m Tris and 0.15 m NaCl (pH 7.2–7.4)]. After transfer to a humidified chamber, the sections were incubated with 10% normal(nonimmune) goat serum (Zymed Corp., San Francisco, CA). The sections were incubated overnight with mouse anti-MGMT monoclonal antibody (clone MT3.1; NeoMarkers, Fremont, CA) diluted 1:125 in TBS. After washing in TBS, biotin-like activity was blocked using the Biotin Blocking Kit (Dako Corp., Carpinteria, CA). The sections were subsequently incubated with biotinylated goat antimouse immunoglobulins (Jackson ImmunoResearch, West Grove, PA) at 1:400 dilution and then with streptavidin-horseradish peroxidase conjugate(Jackson ImmunoResearch) diluted 1:600. Color was developed in 3,3′-diaminobenzidine (Sigma Chemical Co., St. Louis, MO) with H₂O₂ as substrate. Normal epithelium and stromal cells provided a positive internal control. Statistical significance was identified using the two-sided Fisher’s exact test or Pearson’s χ² when one or more sample number(s) in any one comparison fell below a sample size of n = 5.

Methylation of MGMT was observed in 3 of 23 (13%)MSI-H, in 28 of 44 (64%) MSI-L, and in 6 of 23 (26%) MSS colorectal carcinomas (P < 0.0001; Fig. 1). K-ras was mutated in 1 of 23 MSI-H, in 21 of 44 MSI-L, and in 7 of 23 MSS cancers (P = 0.003). Within the MSI-L subset, K-ras mutation correlated with MGMT methylation (P < 0.0002;Table 1). Manual sequencing identified nucleotide substitutions in 24 of 29 cancer samples with a K-ras mutation: G to A in 14 (58%); G to T in 8 (33%); and G to C in 2 (8%). The frequency of mutational bandshifts in MYCL and D2S123 (the most commonly mutated microsatellite loci in MSI-L cancers) is shown for MSI-L cancers stratified by MGMT methylation and K-rasmutation status (Table 1). Bandshifts in MYCL were present in 28 of 42 informative MSI-L cancers (67%), and showed no correlation with MGMT methylation or K-ras mutational status(Table 1). By contrast, bandshifts in D2S123 occurred in 10 of 42 informative MSI-L (24%) cancers but were significantly correlated with MGMT methylation (P < 0.05) and positively associated with K-ras mutation,although this fell short of statistical significance(P = 0.07). K-ras was mutated in 3 of 6 (50%) MSS cancers with MGMT methylation and in 4 of 16(25%) unmethylated MSS cancers, the trend falling short of significance (P = 0.33).

Immunohistochemically demonstrated MGMT protein expression was scored as “none” or “trace” in 23 of 31 (74%) methylated cancers and in 10 of 49 (20%) unmethylated cancers (P < 0.0001). Stromal expression was variable but maintained in all cases(Fig. 2, A and B). In normal mucosa, MGMTexpression was limited to nuclei within cycling cells of the lower crypt and nuclei of stromal cells within the lamina propria. Expression of MGMT was lost in 2 of 5 MSI-H, in 5 of 20 MSI-L, and in 2 of 6 MSS serrated polyps. Three patterns of loss of MGMT expression in serrated polyps were evident: (a) affecting the entire polyp or a large component; (b) limited to small groups of nondysplastic crypts; and (c) limited to dysplastic subclones (Fig. 2 C).

The classification of sporadic colorectal cancer on the basis of DNA microsatellite status has generated groups showing biological,pathological, and clinical differences (21, 22). In general, genetic alterations in MSI-L cancers are similar to those of the suppressor or MSS pathway, and MSS and MSI-L cancers are not readily distinguished either clinically or morphologically (11, 23). Nevertheless, MSI-L cancers do show patterns of 5q loss of heterozygosity (11), β-catenin immunostaining(11), BCL-2 immunostaining (24),apoptotic activity (25), and lymphocytic infiltration(25) that are intermediate between MSS and MSI-H cancers. It could be argued that misdiagnosed MSI-H cases would account for these observations, but given the sharply defined bimodal distribution of MSI cancers (2, 3), misclassification of significant numbers of MSI-H cases as MSI-L is unlikely. In fact, the likelihood of miscalling a MSI-L cancer as MSI-H is greater than the converse. This is because the National Cancer Institute-NIH panel includes three dinucleotide markers that are sensitive for MSI-L status (2, 4). Cancers with instability in two of these dinucleotide markers and no instability at mononucleotide markers are classified as MSI-H, despite the fact that sporadic cancers with instability limited to dinucleotide markers do not show immunohistochemical loss of a DNA mismatch repair protein and are unlike MSI-H cancers in their clinical,pathological, and molecular characteristics (10, 23).

We have shown that MGMT methylation is strongly associated with sporadic MSI-L cancers, though is not restricted to this subset. Additionally, K-ras mutation and MGMT are highly correlated within MSI-L cancers. These observations are consistent with the suggestion that methylation and subsequent inactivation of MGMT overloads the mismatch repair system resulting in a mild mutator phenotype and, in turn, predisposing to mutation in K-ras. We could not confirm a selective association between G to A transition in K-ras and MGMT methylation(18).

The high frequency of MYCL mutation can only be partially explained by the loss of MGMT function in view of the high rate of MYCL mutation in cancers with unmethylated MGMT (Table 1). An alternate mechanism may explain MYCL mutation in cancers with functioning MGMT. MSS cancers showing MGMT methylation could be underdiagnosed MSI-L cancers, because the microsatellite markers used in this study are not 100% sensitive for MSI-L status. K-ras is mutated more frequently in MSS cancers with MGMT methylation,although the trend falls short of significance. The combination of MGMT silencing and loss of DNA mismatch repair proficiency in MSI-H cancers might lead to a mutational burden that would be unsustainable and would therefore trigger apoptosis. This would explain the infrequent finding of both MGMT methylation and K-ras mutation in MSI-H colorectal cancer (11). Nevertheless, MGMT methylation was demonstrated in 3 of 23 MSI-H cancers and immunohistochemical loss occurred in 2 of 5 MSI-H serrated polyps, indicating that mutual exclusivity of MGMTsilencing and MSI-H is not absolute.

K-ras mutation appears to initiate the development of crypt hyperplasia and serration that characterizes common types of ACF that are believed to be the forerunners of hyperplastic polyps and some adenomas (12). Paradoxically, the frequency of K-ras mutation is higher within ACF than in hyperplastic polyps (9). It is possible that ACF with potential for progression may show MGMT methylation (and may subsequently develop mutations including K-ras), whereas ACF that do not progress are initiated by K-ras mutation in isolation. Immunohistochemical staining patterns for MGMT in hyperplastic polyps,mixed polyps, and serrated adenomas indicate that only a subset of these lesions is likely to be initiated by MGMT methylation. Despite the fact that most MSI-L polyps included dysplastic areas, only 5 of 20 MSI-L polyps (25%) showed a loss of expression of MGMT. Moreover, in seven of the nine polyps showing a loss of MGMT staining,the loss was limited to small subclones. It is possible, however, that MGMT methylation is a fluctuating phenomenon within early lesions, with demethylation occurring after the establishment of mutated clones. Such a “hit and run” mechanism would account for mutations (including MSI) occurring in the absence of MGMTsilencing.

MSI-L status in colorectal cancer appears to have both biological and clinical significance. We have shown previously that distant metastases at the time of diagnosis of MSI-H, MSI-L, and MSS colorectal cancers were present in 4, 21, and 14% of cases, respectively, although the trend fell short of significance (23). It is nevertheless interesting that mutation in K-ras is associated with methylation of MGMT (18) and with a poor prognosis (26). These observations are now beginning to demarcate a distinct molecular pathway of colorectal tumorigenesis that differs from the classic adenoma-carcinoma sequence.