The incidence of cervical squamous cell carcinoma (CSCC) has decreased constantly during the last several decades, because of the availabilities of organized and sporadic cytological screening. However, the incidence of cervical adenocarcinoma (CA) continues to increase.1, 2 Although still a subject of debate, the behaviors of CA, such as metastasis to lymph node or response to radiation therapy, have been suggested to be quite different from those of CSCC.3, 4 However, the relative infrequency of CA has restricted information on its carcinogenesis.
Epigenetic mechanisms that result in aberrant gene expression are prominent features of many cancer types, and promoter hypermethylation is one of the main epigenetic mechanisms of gene silencing. The silencing of specific genes by the DNA promoter hypermethylation has been suggested to modify the biological characteristics of human cancers. Also, it is believed that unique promoter hypermethylation profiles exist for the various human cancer types, in which some gene changes are shared while others are cancer-type specific.5 Previously, we described the hypermethylation patterns of 15 tumor suppressor genes in uterine cervical cancer.6 During these previous studies, we developed the impression that hypermethylation patterns are dependent on histologic type. This prompted us to assess whether DNA promoter hypermethylation in CA differs from that in CSCC. In addition, we compared the DNA hypermethylation profiles of CA and endometrial adenocarcinoma (EA), because although the 2 are distinct entities, they show considerable overlap in terms of tumor histology. We also investigated whether DNA hypermethylation profiles can differentiate CA and EA.
In the present study, we characterized the DNA methylation statuses of 9 genes. Initially, we analyzed previous reports and compared frequencies of gene methylation in CA and CSCC. This analysis showed that the hypermethylation frequencies of RASSF1A, APC, death-associated protein kinase (DAPK) and CDH1 differ in CAs and CSCCs (Table I). We also examined p16, THBS1, tissue inhibitor of metalloproteinase-3 (TIMP3) and helicase-like transcription factor (HLTF) methylation, because we previously found that their methylation frequencies are higher in CAs (TIMP3, HTLF) and CSCCs (p16, THBS1).6 Finally, we included hMLH1 in the analysis, because hMLH1 methylation was previously found to be common in EAs,15 but not in cervical carcinoma.6
Table I. Analysis of Previously Published Data on Methylation Profiles in CSCCs and CAs
Cases with methylation/cases examined (CSCC vs. CA)
Korean women undergoing treatment for cervical or endometrial carcinoma at the Gynecologic Oncology Division of Seoul National University Hospital were recruited for this study. We studied 113 archival samples of surgically resected carcinomas (62 CSCCs, 30 CAs and 21 EAs). Of the 30 CAs, 4 cases were endometrioid types, and 26 cases were endocervical types. We did not include cervical adenosquamous carcinoma, because it has a heterologous histology. In the 21 EAs, 2 cases were of the serous type, and 1 case was of the mucinous type and the remainders were of the endometrioid type. Patient age and disease stage were not statistically different between these 3 groups. After identifying histologic types on hematoxylin–eosin-stained slides and marking the respective lesions on paraffin blocks, the marked areas were scraped from the paraffin blocks. The collected materials were dewaxed in xylene and rinsed in ethanol, and the dried tissues were digested in a lytic solution containing proteinase K. Genomic DNA was purified using the phenol/chloroform ethanol precipitation method.
Tumor DNAs were modified using the sodium bisulfite method, as described previously.16 Briefly, 2 μg of DNA was denatured with 2 M NaOH, and then treated with 1 mM hydroquinone and 3.5 M sodium bisulfite for 16 hr at 55°C. After purification using a JETSORB gel extraction kit (Genomed, Germany), DNA was treated with 3 M NaOH and subsequently precipitated with 3 volumes of 100% ethanol and a one-third volume of 7.5 M NH4Ac at −20°C. The precipitated DNA was then washed with 70% ethanol and dissolved in distilled water.
A panel of 9 genes were analyzed for methylation state by methylation-specific PCR (MS-PCR). The genes tested were APC, DAPK, CDH1 (E-cadherin), HLTF, hMLH1, p16, RASSF1A, THBS1 and TIMP3. Bisulfite-modified DNA was amplified using primers, specific for either the methylated or unmethylated sequences. The primer sequences of all genes, whether methylated or unmethylated forms, and the PCR conditions used for the MS-PCR were as described previously.6, 17, 18 PCR was performed in 25 μl reaction volumes, containing PCR buffer [16.6 mmol/l (NH4)2SO4, 67 mmol/l Tris (pH 8.8), 6.7 mmol/l MgCl2, 10 mmol/l β-mercaptoethanol], deoxynucleotide triphosphates (0.2 mmol/l each), the primers (10 pmol each) and 1 unit of Taq polymerase. PCR products (7 μl) were separated by electrophoresis on 2.5% agarose gel and visualized under UV illumination by ethidium-bromide staining. Samples showing signals approximately equal to the size marker were scored as methylated. If samples showed faint-positive signals, experiments were repeated 3 times. Only samples that showed consistent positive signals were scored as methylated.
The χ2 test and Fisher's exact test were used to compare the DNA promoter hypermethylation frequencies in the 9 aforementioned genes in CSCCs, CAs and in EAs. Bonferroni corrections were used to adjust for multiple comparisons among the 9 genes, and a p threshold of 0.0055 was used to determine statistical significance. All statistical analyses were performed using SPSS for Windows Release 12.0 (SPSS, Chicago, IL).
To confirm the presence of a distinct difference between gene methylation profiles in CSCC and CA, we investigated the DNA methylation profiles of 9 tumor suppressor genes, using MS-PCR. A total of 87 (94.6%) of the 92 cervical carcinomas (62 CSCCs and 30 CAs) showed methylation in at least 1 of these loci, and the proportion of cases that showed methylation in at least 1 of these loci was similar for CSCC and CA (95.2% vs. 93.3%, respectively). In CSCC, the methylation frequency at each gene locus varied from 0 to 46.2%, whereas it varied from 0 to 53.8% in CA.
In CSCC, a high frequency of methylation (>40%) was detected for DAPK (77.4%), CDH1 (80.6%) and THBS1 (40.3%), whereas p16 (19.4%) showed an intermediate frequency of methylation (15–40%). The remaining loci showed methylation frequencies of <15%. Meanwhile, in CA, high frequencies of methylation were found for DAPK (46.7%), CDH1 (43.3%), APC (40.0%), TIMP3 (53.3%) and HLTF (43.3%), and intermediate frequencies were found for RASSF1A (33.3%) and THBS1 (23.3%). The remaining loci showed methylation frequencies of <15%. A comparison of the DNA promoter hypermethylation profiles of CSCC and of CA is shown in Table II. When Bonferroni corrections were used to adjust for multiple comparisons among the 9 genes, only the hypermethylation frequencies of DAPK, CDH1, HLTF and TIMP3 were found to be significantly different between these 2 histologic types. DAPK and CDH1 were more frequently methylated in CSCC than in CA (p = 0.005 and 0.001, respectively). On the other hand, HLTF and TIMP3 were more frequently methylated in CA (p < 0.001 and 0.001, respectively). Moreover, the hypermethylations RASSF1A and APC were more frequent in CAs than in CSCCs, but this was not significant (p = 0.008 and 0.009, respectively).
Table II. Comparison of The DNA Promoter Hypermethylation Frequencies in CSCC and CA
To investigate whether there is a distinct difference between the methylation profiles of CA and EA, we assessed the DNA methylation profiles of 21 cases of EA. A total of 20 (95.2%) of the 21 EAs showed methylation in at least 1 of these loci, and methylation frequencies of each gene locus varied from 0 to 69.2%. In particular, high frequencies of methylation (>40%) were detected in RASSF1A (81.0%), E-cadherin (42.9%) and HLTF (47.6%). And, intermediate frequencies of methylation (15–40%) were detected in APC (19.0%), hMLH1 (23.8%) and TIMP3 (19.0%), whereas p16 showed no sign of methylation. A comparison of the DNA promoter hypermethylation profiles of CAs and EAs is shown in Table III. After applying Bonferroni correction, only RASSF1A was found to be significantly more frequently hypermethylated in EAs than in CAs (p = 0.001), and although not statistically significant, DAPK and TIMP3 were less frequently hypermethylated in EAs (p = 0.019 and 0.02, respectively).
Table III. Comparison of The DNA Promoter Hypermethylation Frequencies in CAs and EAs
In addition, we observed a decreasing trend of DAPK hypermethylation and an increasing trend of RASSF1A hypermethylation on moving from CSCCs to EAs (ptrend < 0.001 and 0.001, respectively). These trends are illustrated in Figure 1.
The present study shows that the DNA promoter hypermethylation profiles differ in different histologic types of uterine carcinoma. Our findings encourage us to speculate that some DNA hypermethylation profiles favor the carcinogenesis of specific uterine carcinomas. In the present study, DAPK and CDH1 promoter hypermethylations were found to be a discriminating characteristic of CSCC. Given that CDH1 methylation is more commonly found in normal cervical tissues than in DAPK,19 it is more likely that DAPK methylation is more closely related to the carcinogenesis of CSCC. Yang et al. reported that CDH1 methylation is more frequent in CSCCs than in CAs (68.9% vs. 37.5%), which is consistent with our data,10 and Dong et al. reported similar data (24% vs. 10%). DAPK methylation is frequently found in human solid tumors, such as lymphoma, lung cancer and head and neck cancer.5, 20, 21
DAPK is a proapoptotic serine/threonine kinase and is involved in apoptosis. Recently, it was demonstrated that DAPK counteracts oncogene-induced transformation by activating a p19ARF/p53-dependent apoptotic checkpoint.22 These results suggest a role for DAPK in an early apoptotic checkpoint designed to eliminate premalignant cells during cancer development, further suggesting that early checkpoint failure by DAPK inactivation has a more important role in the oncogenesis of CSCCs than that of CAs.
The novel finding of the present study is that DNA hypermethylations of HLTF and TIMP3 are more frequent in CAs than in CSCCs. The TIMP3 gene is located on chromosome 22q12.3, and abrogates matrix metalloproteinase (MMP) activity by binding covalently to its active site.23 Moreover, it is believed that TIMP3 downregulation contributes to tumor growth and tumor invasion by allowing MMP activity upregulation in the extracellular matrix.24, 25 Recent studies on the methylation-associated silencing of TIMP3 have suggested that it has a tumor suppressive role in different cancers.26 The frequency of TIMP3 methylation in CAs has been reported to be 0% (0 of 6 cases),27 which is inconsistent with the findings of the present study. However, it is not possible to subject these 6 cases to frequency analysis by histologic type, as our finding needs further confirmation. HLTF (a SWI/SNF family protein) participates in chromatin remodeling and facilitates transfactor interactions with nucleosomes.28 Loss of HLTF expression in association with its promoter methylation was reported in colon cancers.29 Therefore, further investigation about the association between the expression of these 2 genes and methylation status in CAs are warranted.
In the present study, DNA hypermethylations of RASSF1A and APC were more frequent in CAs than in CSCCs, although this did not reach significance. This result is consistent with previous reports. Cohen et al. reported that the methylation of the RASSF1A promoter is significantly more frequent in CAs than in CSCCs (45% vs. 0%), and although not statistically significant, Kuzmin et al. reported similar results (24% vs. 10%). Another study found no significant difference between RASSF1A methylation frequencies in CA and CSCC.13 However, if the 4 studies mentioned earlier are metaanalyzed as 1 data set, RASSF1A methylation becomes significantly more frequent in CAs than in CSCCs (29 of 101, 28.7% vs. 20 of 168, 11.9%, p = 0.0009). Regarding APC hypermethylation, Dong et al. observed that APC is more frequently methylated in CA than in CSCC (59.1% vs. 12.9%).
As mentioned earlier, morphological overlap can make the histological differentiation of EA and cervical CA difficult. Therefore, the distinct DNA hypermethylation patterns of these 2 tumors are of clinical importance. According to our data, the frequency of RASSF1A methylation gradually increased in the order CSCC < CA < EA. Previously, it was reported that RASSF1A was highly methylated (14 of 15, 93.3%) in 15 cases of EA, which is similar to that found in the present study.30RASSF1A is frequently inactivated in a variety of primary human solid tumors, including lung, breast, ovarian, prostate and bladder.9, 31, 32, 33 Since RASSF1A is located at chromosome 3p, epigenetic inactivation of 1 RASSF1A allele, followed by LOH at 3p, may result in the inactivation of RASSF1A. It is well known that LOH at 3p is frequently observed in EAs and uterine cervix.34, 35, 36 Moreover, a recent study reported that 67% (8 of 12) of cervical cancers with hypermethylated RASSF1A showed concomitant LOH at 3p21.13 Our data suggests that RASSF1A inactivation may have an important role in EAs and that the epigenetic inactivation of RASSF1A, in combination with LOH, plays a more important role in the development of EAs.
Although the present study shows correlations between the DNA methylations of some tumor suppressor genes and histologic uterine cancer types, it should be noted that, because of insufficient statistic power caused by sample size constraints, we cannot exclude a possible association between histologic type and other less well-correlated genes. To assess this possibility, further studies are needed. In addition, the present study describes the presence of methylated alleles, but does not comment on the degree of methylation. To determine methylated alleles quantitatively, microdissection of epithelial cells and quantitative assays are required.
This study demonstrates that the various types of uterine cancer, i.e., CSCCs, CAs and EAs, show unique promoter hypermethylation profiles. Our findings suggest that certain combinations of promoter hypermethylations predispose tissues to the development of specific, human uterine cancer types. In addition, our data suggest that the carcinogenesis of CSCC and CA might proceed via different epigenetic mechanisms.