Causes and Consequences of Microsatellite Instability in Endometrial Carcinoma¹

Progress in elucidating the molecular genetic basis of the HNPCC³ syndrome has established that an inherited mutation in one of several genes encoding components of a family of DNA mismatch repair proteins confers predisposition to colorectal and other cancers. The great majority of HNPCC kindreds for which a specific genetic alteration has been identified are linked to one of two genes, MSH2 or MLH1 (1). Following somatic inactivation of the homologous wild-type allele, the resulting defect in postreplicative DNA mismatch repair leads to MI, believed to represent the molecular phenomenon responsible for tumorigenesis in the affected tissue(s) (2, 3). Sporadic, i.e., nonhereditary, manifestations of several HNPCC component tumor types also exhibit MI in a substantial fraction of cases, especially colorectal (4, 5, 6), gastric (7, 8, 9, 10), and endometrial (11, 12, 13, 14) carcinomas. The underlying molecular genetic defect in sporadic MI+ colorectal cancers appears to be somatic mutational inactivation of either MSH2 or MLH1 in some cases (15, 16, 17), perhaps in as many as 65% (18) of cases, with hypermethylation of the MLH1 promoter and attenuation of MLH1 expression evident in some cases (19). In contrast, the molecular basis for MI in sporadic endometrial cancers remains obscure, with somatic mutations in MSH2 or MLH1 having been described in only four MI+ endometrial cancers to date (20, 21, 22).

In tumors affected by MI, the mechanism of tumorigenesis is believed to involve frameshift mutations of microsatellite repeats within coding regions of genes, the inactivation of which contributes to tumor development. Genes commonly affected through this mechanism in MI+ gastrointestinal cancers include TGFβRII (23, 24), IGFIIR (25), BAX (26), MSH3 and MSH6 (27), E2F4 (28), and BRCA1 and BRCA2 (29). Limited analyses of MI+ endometrial cancers suggest that mutations of TGFβRII (30) and IGFIIR (31) may be much less common than in MI+ gastrointestinal cancers, but the other loci have yet to be examined in endometrial cancers. Of interest in this context is the recent demonstration that PTEN tumor suppressor gene mutations are found in 30–50% of all endometrial carcinomas (32, 33, 34) and are more common in MI+ than in MI− tumors (32, 33). PTEN mutations are rare in MI+ gastrointestinal cancers (33), suggesting that PTEN may represent a mutational target of MI specific to endometrial tumorigenesis.

The purpose of this study was to perform a comprehensive analysis of the molecular genetic causes and consequences of MI in a series of sporadic endometrial carcinomas. The MI+ tumors were so classified using a panel of microsatellite markers developed together with a consensus recommendation for the definition of MI (35, 36). To determine the extent to which somatic mutations in known mismatch repair genes contribute to MI in sporadic endometrial cancers, complete mutational analyses were performed on the mismatch repair genes MSH2 and MLH1. The same tumors were then examined for evidence of somatic frameshift mutations presumed to result from defective mismatch repair and included analyses of coding region microsatellites within the genes TGFβIIR, IGF2R, BAX, E2F4, MSH3, MSH6, BRCA1, and BRCA2. A complete mutational analysis of the PTEN gene was also performed in all MI+ tumors.

This study was approved by the Institutional Review Board of the Memorial Sloan-Kettering Cancer Center. All tumors and corresponding normal tissues were from patients that underwent primary endometrial cancer surgery at this institution, prior to chemotherapy or radiation therapy, and were obtained as frozen specimens from the Department of Pathology. Pathological review was performed on every case to confirm a diagnosis of endometrial carcinoma. Genomic DNA was isolated using standard procedures (37). All molecular genetic analyses were performed on the same DNA sample from each tumor.

The assessment of MI was based on the use of a panel of microsatellite markers developed by a multicenter consortium for diagnostic reliability and the standardization of criteria for definition of MI (35, 36). Tumors were classified as MI+ if at least three of five (60%) markers in the primary panel (BAT26, BAT40, APC, Mfd15, and D2S123) displayed evidence of mutant alleles in tumor DNA compared to corresponding normal tissue DNA. Reactions for PCR amplification were carried out in a volume of 20 μl containing 50 ng of genomic DNA, 1.5 mm MgCl₂, 50 mm KCl, 10 mm Tris-HCl (pH 8.3), 200 μm each dNTP, 0.8 μm each primer, and 0.75 units of Taq polymerase (Perkin-Elmer Corp., Foster City, CA). One primer was end-labeled with [γ-³³P]ATP by polynucleotide kinase using the KinAce-It kit (Stratagene, La Jolla, CA) and column purified prior to PCR. Thirty PCR cycles were performed, consisting of 30 s at 94°C, 30 s at 55°C, and 30 s at 72°C, followed by a 7-min extension at 72°C. The PCR products were processed by diluting 1:1 in denaturing loading buffer (95% formamide, 10 mm NaOH, 0.05% xylene cyanol FF, and 0.05% bromphenol blue), heated at 90°C for 2 min, and 5 μl were electrophoresed in 6% polyacrylamide gels containing 8.3 m urea for 2–3 h at 70 W. The gels were fixed in 10% methanol-10% acetic acid, dried, and exposed to Hyperfilm MP autoradiography film (Amersham) for 3–24 h at room temperature.

The complete coding regions and exon-intron junctions of MSH2 and MLH1 were screened for mutations in all MI+ tumors by SSCP analysis followed by sequence analysis of all potential variants. Intron-based primers for PCR amplification of the 16 exons of MSH2 and 19 exons of MLH1 were as described previously (38, 39). Each PCR was carried out in a volume of 20 μl containing: 50 ng of cDNA; 1.5.mm MgCl₂; 50 mm KCl; 10 mm Tris-HCl (pH 8.3); 200 μm each dATP, dGTP, and dTTP; 20 μm dCTP; 1 μCi of [α-³³P]dCTP (6,000 Ci/mmol, NEN); 0.8 μm each primer; and 1 unit of Taq polymerase (Perkin-Elmer Corp.). Amplification was performed in a Perkin-Elmer 9600 thermal cycler for 35 cycles consisting of 20 s at 94°C, 20 s at 55°C, and 30 s at 72°C. PCR products were diluted 1:10 in denaturing loading buffer[95% formamide, 10 mm EDTA (pH 8.0), 0.02% xylene cyanol FF, and 0.02% bromphenol blue], heated at 94°C for 5 min, and placed on ice, and 4 μl were loaded for electrophoresis. Gels for SSCP analysis consisted of 0.5× MDE solution (FMC BioProducts) and 0.6× TBE buffer, and were run in 0.6× TBE buffer at 6–8 W for 16 h at room temperature. Gels were dried and exposed to Hyperfilm MP autoradiography film (Amersham) for 4–16 h.

For sequence analysis, variant and wild-type bands were excised from SSCP gels following autoradiography and suspended in 50 μl of TE buffer for 1 h at room temperature. One μl of the eluted DNA sample was used as a template for PCR amplification under conditions identical to those described above for SSCP analysis, except that all dNTPs were at 200 μm and radiolabeled dCTP was omitted. Each PCR product was electrophoresed in its entirety in NuSieve 3:1 agarose (FMC BioProducts), visualized with ethidium bromide, excised from the gel, and purified using the Qiaex II gel extraction kit (Qiagen). Two ng of each DNA template were sequenced using the Thermo Sequenase radiolabeled terminator cycle sequencing kit (Amersham). All products were sequenced in both directions, using one or the other PCR primer. Following the sequencing reactions, 6 μl of each termination reaction were added to 4 μl of stop solution and heated to 70°C for 10 min, and 4 μl were loaded into each gel lane. Sequencing gels consisted of 6% polyacrylamide and 7 m urea in TBE buffer. After electrophoresis at 70 W for 2–3 h, gels were fixed in 10% methanol-10% acetic acid, dried, and subjected to autoradiography as above for 16 h.

Methylation of the MLH1 promoter region was examined using the procedure described previously (19), with the following modifications. Following restriction endonuclease digestion, 12.5 ng of DNA were used as template for PCR amplification in a volume of 20 μl. One PCR primer (25266) was radiolabeled with [γ-³³P]ATP as described above, and half of each PCR product (10 μl) was electrophoresed in a 10% polyacrylamide/TBE gel (Bio-Rad). Gels were dried and subjected to autoradiography as above.

Mutations were detected in genes containing coding region microsatellites by PCR amplification of the appropriate segment of each gene followed by gel electrophoresis as described for the MI analysis above. Primers for PCRs were as described previously for TGFβRII (30), IGFIIR (25), BAX (26), E2F4 (28), MSH3 and MSH6 (27), and BRCA1 and BRCA2 (29). Mutation analysis of the PTEN gene was performed by screening the entire coding region and exon-intron junctions by SSCP and sequencing analyses as described above for the mismatch repair gene mutation analysis. Eleven sets of intron-based PCR primers were used to amplify the nine exons of PTEN as described (34).

A consecutive series of 57 endometrial carcinomas were examined for the presence of MI using the diagnostic criteria described in “Materials and Methods.” Fourteen (25%) tumors displayed unambiguous evidence of MI, as shown in Fig. 1. These tumors were examined further for the molecular etiology and downstream molecular effects of this phenomenon. A complete mutation screen of MSH2 and MLH1 failed to reveal a clearly deleterious mutation in any tumor. Two sequence alterations of unknown significance were identified in MLH1: a somatic missense mutation at codon 718, CAC(His) → TAC(Tyr) (Fig. 2) and a 2-bp insertion at position −13 (relative to exon 11) of intron 10 that did not obviously affect the splice junction.

To determine whether methylation of the MLH1 promoter region was associated with MI in these tumors, paired DNA samples from each of the 14 MI+ endometrial cancers and their corresponding normal reproductive tract tissues were examined. Methylation of the MLH1 promoter was not observed in any of the normal tissues but was seen in 10 of 14 (71%) of the MI+ cancers (Fig. 3). Adequate clinical material was not available to perform an expression analysis of MLH1 in these tissues; thus, 28 MI− cancers were examined to determine the extent of correlation of MLH1 promoter methylation with the MI phenotype. Only two (7%) of these tumors displayed bands of weak autoradiographic intensity corresponding to a partially methylated MLH1 promoter. Thus, there was a highly significant correlation of MLH1 promoter methylation with the presence of MI in the corresponding tumor (P = 0.01, Fisher’s exact test, two-sided).

The MI+ tumors were next examined for the presence of frameshift inactivating mutations in microsatellite-containing genes believed to be the result of MI (Fig. 4). Generally, mutations in these genes were uncommon in MI+ endometrial cancers, with a range of 0–29% of tumors containing such alterations (Table 1). Microsatellite repeats within the TGFβRII, MSH3, MSH6, BRCA1, and BRCA2 genes appeared rarely, if ever, mutated, whereas mutations in the BAX gene were most common. A mutation analysis of the complete PTEN coding region was also performed (Fig. 5). Four of 14 (29%) MI+ endometrial cancers were affected by inactivating mutations of PTEN: two of these four mutations were frameshift alterations, one of which occurred in a coding region microsatellite; one was a nonsense mutation; and one was a nucleotide substitution at the splice acceptor site of intron 6 (Table 2).

The finding that 25% of sporadic endometrial carcinomas display MI is consistent with previous estimates of this phenomenon, which have ranged from 17 to 23% (11, 12, 13, 14). A comprehensive analysis of these MI+ tumors for the underlying molecular defect responsible for MI revealed that somatic mutations in the mismatch repair genes known to sustain somatic alterations in MI+ gastrointestinal tumors, MSH2 and MLH1, are likely to be very rare. These findings are also consistent with previous studies in which MSH2 and MLH1 were screened for mutations in MI+ endometrial cancers; only four somatic mutations, two in MSH2 and two in MLH1, have been identified in a total of 51 MI+ endometrial cancers (20, 21, 22). Collectively, these data imply strongly that the molecular etiology of MI in sporadic endometrial cancers involves an alternative mechanism(s) to that operative in HNPCC and sporadic gastrointestinal cancers.

Inactivation of the MLH1 gene through promoter hypermethylation has been proposed as a mechanism for the induction of MI in sporadic colorectal cancers exhibiting attenuated expression of MLH1 (19). A recent detailed analysis of this phenomenon demonstrates that MLH1 promoter methylation is common in colorectal cancer cell lines and primary tumors with mismatch repair deficiencies, that the promoter methylation is correlated with decreased expression of MLH1 at both the mRNA and protein levels, and that demethylation of the MLH1 promoter results in the reexpression of MLH1 in cell lines (40). This phenomenon has also been observed in the endometrial carcinoma cell line AN3CA (19, 41). Our data indicate that the MLH1 promoter is extensively methylated in the majority (71%) of sporadic MI+ endometrial cancers. Although we were unable to examine MLH1 expression in these tumors because of inadequate clinical material, the MLH1 promoter methylation was highly correlated with the presence of MI because only 7% of MI− tumors displayed MLH1 methylation. Furthermore, only a fraction of cells in these two MI− cases appeared to be affected by this phenomenon. Results from the colorectal carcinoma analysis are very similar, in that 2 of 21 (10%) sporadic MI- tumors exhibit methylation of MLH1 (40).

Finally, we examined the molecular consequences of MI in this series of endometrial cancers by quantitating the extent of frameshift mutations in microsatellite-containing genes known to be altered in gastrointestinal cancers. None of the genes examined were found to be commonly mutated. The TGFβRII gene is altered in the majority of MI+ colorectal and gastric carcinomas, 90% and 71%, respectively (24, 30), but we failed to identify any MI+ endometrial cancer with this alteration. A previous study reported a low incidence of TGFβRII mutations in MI+ endometrial compared to MI+ colorectal cancers (30). Together, these data indicate that mutational inactivation of TGFβRII does not provide a significant contribution to the tumorigenic process in the endometrium.

The mismatch repair genes MSH3 and MSH6, which both containing coding region microsatellites, are inactivated in 30–40% of MI+ colorectal (27) and 50–60% of MI+ gastric (29) carcinomas. These observations suggest a progressive model in which “primary” mutations in mismatch repair genes such as MSH2 or MLH1 lead to MI, a result of which is then “secondary” mutation of additional mismatch repair genes, such as MSH3 and MSH6, with the acceleration of genomic instability and the accumulation of additional mutations in cancer-related genes (27). Interestingly, however, we observed only one MSH6 mutation and no MSH3 mutations in MI+ endometrial cancers, again suggesting that discrete pathways exist for tumorigenesis in association with MI in the gastrointestinal tract and endometrium.

Somewhat more common were frameshift mutations of the regulator of apoptosis BAX, detected in 29% of MI+ endometrial cancers compared to 50% of MI+ colorectal (26) and 67% of MI+ gastric (29) carcinomas. The growth factor receptor IGFIIR appears to be mutated at a similarly low frequency in MI+ endometrial (15%) and gastrointestinal (6–25%) cancers (25, 31). Mutations of the transcription factor E2F4 are less common in MI+ endometrial (20%) than gastrointestinal (31–65%) cancers (28, 42), and frameshift mutations of microsatellites within the tumor suppressors BRCA1 and BRCA2 are equally rare in endometrial (0–7%) and gastrointestinal (0–12%) cancers (29). These data imply that a significant proportion of the molecular targets of MI in endometrial cancer remains to be identified.

The PTEN tumor suppressor gene is a plausible candidate for such a target of MI in endometrial carcinoma. The gene is frequently inactivated in endometrial cancers (32, 33, 34), and most tumors with mutations display the MI phenotype (32, 33). Furthermore, the coding region of PTEN contains two [A]₆ tracts (43, 44), which, although of shorter length than the typical microsatellite, represent potential mutational targets in tumors with defective mismatch repair. Only one of the MI+ endometrial cancers in our series displayed a frameshift mutation in one of these poly(A) tracts, however. An additional two MI+ tumors contained nucleotide substitution mutations, and one MI+ tumor contained an insertion mutation that are not obviously related to the MI phenotype. Overall, our data suggest that the majority of MI+ endometrial carcinomas do not sustain PTEN mutations during tumor development. Furthermore, the mutations that occur are not obviously attributable to MI; this view is supported by a previous observation that frameshift mutations in a poly(A) tract of PTEN are found in MI− endometrial cancers as well (33).

In summary, these and previous published data support a model in which ∼25% of all sporadic endometrial carcinomas are characterized by MI, and in which the great majority of these tumors are not causally associated with somatic mutations in the mismatch repair genes responsible for HNPCC and some sporadic gastrointestinal cancers. Instead, our data suggest that MI in most of these cancers develops as a result of MLH1 inactivation through promoter hypermethylation. Further studies will be required to establish a cause-and-effect relationship between MLH1 hypermethylation and MI in endometrial carcinoma, but the strong statistical correlation between these two phenomena and the demonstration of a direct mechanistic relationship in colorectal cancer cells (40) suggest that such a direct relationship in endometrial cancer is likely. Finally, our data imply that the majority of molecular genetic targets for MI in endometrial tumorigenesis remain to be identified.

We are grateful to Dr. William L. Gerald for assistance in obtaining the clinical material used in this study and to Dr. William J. Hoskins for his support of this research program.