Loss of E-cadherin expression is found frequently in many types of human malignancies, including mucoepidermoid carcinoma (MEC). CpG methylation in the promoter region has proven important in the regulation of gene expression implicated in malignant transformation. DNA methyltransferases (DNMTs) are the major enzymes involved in establishing genomic methylation patterns. The current study was designed to test the hypothesis that CpG methylation of the promoter region of the E-cadherin gene may inactivate its expression and to examine DNMT 1 (DNMT1) protein expression in MEC.
Genomic DNA was obtained from paraffin embedded sections by laser microdissection in 46 MEC specimens. Methylation status of the E-cadherin promoter was examined by utilizing the methylation-specific polymerase chain reaction assay. To examine E-cadherin and DNMT1 proteins expression levels, the MEC specimens and adjacent epithelial tissues were studied immunohistochemically. Chi-square analysis was used to evaluate the correlation of protein expression and E-cadherin methylation status with clinicopathologic parameters. Comparisons of the survival rate between patients with DNMT1-positive and DNMT1-negative patients were made using Kaplan–Meier analysis.
The data showed that all normal tissues expressed E-cadherin, and no promoter methylation was detected. Of the MEC samples analyzed, methylation allele was found in 33 of 46 samples (72%), and reduced E-cadherin expression was found in 21 of 46 samples (45%). DNMT1-positive expression was observed in 29 of 46 MEC samples (63%). A significant correlation was found between E-cadherin expression and the methylation status of E-cadherin promoter (P = 0.021). In addition, increased DNMT1 expression was correlated with histologic grade, clinical stage, and a poor prognosis in patients with MEC.
E-cadherin has a major role in the maintenance of intercellular junctions between epithelial cells in most organs.1 Substantial evidence shows that reduced or loss of E-cadherin expression is a common event in many human malignancies.2 In our previous study, we demonstrated that E-cadherin loss was found frequently in the most common salivary gland malignancy, mucoepidermoid carcinoma (MEC).3 However, the molecular mechanisms of E-cadherin loss are not clear in MEC.
The expression of some genes can be inactivated frequently by reversible epigenetic events rather than genetic events. Aberrant DNA methylation is one of the most consistent epigenetic mechanisms of gene expression in human malignancies.4 In malignant disease, overall DNA methylation level is lower than in normal tissues.5 However, some loci, such as the gene promoter region, tend to exhibit increased DNA methylation in carcinoma.6 Promoter hypermethylation associated with gene silencing in normally unmethylated 5′-CpG-rich areas is mediated by methyl-binding proteins that bind to methylated cytosines and recruit a complex of proteins that repress transcription, including histone deacetylases.7 The structure of the E-cadherin gene is notable for a dense CpG island that flanks the 5′ transcriptional start site. Decreased expression of the E-cadherin gene has been linked to aberrant methylation of this CpG island in several common human malignancies.8 Although E-cadherin protein expression reportedly was decreased significantly in MEC, no previously published study has investigated the potential role of methylation as the causative etiology in loss of E-cadherin expression in this form of malignant disease.
In normal cells, CpG island methylation patterns in genomes can be maintained precisely through DNA methyltransferases (DNMTs).9 To date, three major enzymes that process DNMT activity—DNMT1, DNMT3a, and DNMT3b10—have been confirmed. DNMT1 is the most abundant DNMT targeted to replication foci and has a preference for hemimethylated DNA substrates.11 It seems to be the main enzyme responsible for copying the methylation pattern after each round of DNA replication.12 Recent investigations show that DNMTs are overexpressed in tumorigenic cells13 and in a few types of human tumors.14, 15 Nevertheless, conflicting results between the DNA methylation status of specific genes and DNMT expression have been reported in the literature.16–18 In addition, data on DNMT expression in human MEC are not available. Therefore, in the current study, we undertook an investigation to determine the extent of E-cadherin methylation and to study the possible relation of aberrant methylation and decreased expression of the gene. Correlations of E-cadherin expression with methylation status, DNMT1 expression, and clinicopathologic parameters also were evaluated.
MATERIALS AND METHODS
Patients and Tissue Specimens
Forty-six patients with a diagnosis of MEC were identified from the pathology archives between 1980 and 2004 (Table 1). Clinical stage and follow-up information were obtained on 39 patients. From each tissue block, a series of 5-μm sections were cut. A 5-μm flanking section was stained with hematoxylin and eosin for pathologic evaluation to identify the tumor regions and normal regions. Serial sections were used for immunohistochemistry, and other sections were used for laser microdissection. The histologic grading and clinical stage of MEC were determined as described previously.3
Table 1. Patients' Clinicopathologic Features in Association with E-Cadherin Methylation Status and Immunohistochemical Expression of E-Cadherin and DNA Methyltransferase 1
The 5-μm paraffin sections were deparaffinized by incubation in a xylene solution (twice for 3 min each) followed by rehydration through a series of 95%, 70%, and 50% ethanol solutions. After rinsing in distilled H2O, the sections were stained with hematoxylin and eosin and dehydrated in 50%, 70%, and 95% ethanol (twice for 1 min each) followed by xylene (twice for 1 min each). Then, the slides were used for microdissection on a Leica AS laser microdissection system (Leica Microsystems Wetzlar GmbH, Germany) after they were air-dried. Approximately 5000 cells from representative tumor and normal portions of the slide were microdissected, excluding any surrounding stromal cells. The dissected cells were digested in 50 μL proteinase K buffer (2 mg/mL proteinase K, 10 mM Tris-HCL, 5 mM ethylenediamine tetraacetic acid, and 1% Tween 20) at 45 °C overnight before DNA isolation.
Genomic DNA Isolation and Bisulfite Modification
DNA isolation from laser microdissection cells was performed using the Puregene DNA purification kit (Gentra, Minneapolis, MN). Ten micrograms of glycogen were added as a carrier to help the precipitation of DNA. Bisulfite treatment of purified genomic DNA was used to convert nonmethylated cytosines (C) to uracils (U), ultimately detected as thymidines (T) after polymerase chain reaction (PCR) amplification. The bisulfite reaction was carried out using a CpGenome DNA modification Kit (Intergen, Purchase, NY). In addition, the DNA ultimately was eluted in 20 μL of nuclease-free water and stored at − 20 °C.
Methylation-specific PCR (MSP) of bisulfite-treated genomic DNA was carried out according to the method reported previously.19 Briefly, methylation patterns within the E-cadherin CpG island in exon 1 were determined after sodium bisulfite modification of genomic DNA by a nested PCR approach. In the first round of PCR, primers were 5′-GTTTAGTTTTGGGGAGGGGTT-3′ (sense) and 5′-ACTACTACTCCAAAAACCCATAACTAA-3′ (antisense); and the cycling conditions consisted of an initial denaturation step at 95 °C for 5 minutes followed by the addition of 0.5 unit of Taq polymerase (Roche) and 35 cycles at 95 °C for 40 seconds, at 48 °C for 40 seconds, and at 72 °C for 40 seconds. The size of the product after this initial PCR reaction was 270 base pairs. For the second round of PCR, this product was diluted 1:25 in water, and 2 μL of the dilution were used for MSP. Nested primer sequences for the methylation reaction were 5′-TGTAGTTACGTATTTATTTTTAGTGGCGTC-3′ (sense) and 5′-CGAATACGATCGAATCGAACCG-3′ (antisense), and the primer sequences for the unmethylation reaction were 5′-TGGTTGTAGTTATGTATTTATTTTTAGTGGTGTT-3′ (sense) and 5′-ACACCAAATACAATCAAATCAAACCAAA-3′ (antisense). PCR parameters were the same as those described above, except that the annealing temperatures for the methylation and unmethylation reactions were 62 °C and 60 °C, respectively. The product sizes of the methylation and unmethylation reactions were 112 base pairs and 120 base pairs, respectively. MSP results were divided into 1) unmethylation if no methylation allele was detected, 2) heteromethylation if both methylation and unmethylation alleles were found, and 3) methylation if only methylation allele was detected.
For immunohistochemical studies, the antibodies used included monoclonal antibody anti-E-cadherin (Transduction Laboratory, Lexington, KY) and polyclonal antibody anti-DNMT1 (Santa Cruz Biotechnology, Santa Cruz, CA). Immunodetection for E-cadherin and DNMT1 was performed with a Dako EnVision detection kit (Dako Corporation, Carpinteria, CA). Sections were dewaxed and subjected to antigen heat retrieval. Endogenous peroxidase activity and nonspecific binding were blocked by incubation with 3% hydrogen peroxide and nonimmune serum, respectively. Slides were then incubated sequentially with primary antibodies for 16 hours at 4 °C and peroxidase-labeled polymer (Dako Corporation) secondary antibody for 1 hour. Then, the chromogen diaminobenzidine test was performed to localize positive staining sequentially by light microscopy. For negative control experiments, the primary antibodies were omitted or replaced with an irrelevant antibody, mouse immunoglobulin G1 (Dako Corporation). Sections were counterstained with hematoxylin and coverslipped.
Assessment of Immunoreactivity
Using a semiquantitative scale that was described previously,20 an exclusively membranous pattern of staining was considered positive for E-cadherin expression. Both the intensity of staining and the distribution as an approximate percentage of positive tumor cells were considered in the assessment. Briefly, the staining intensity was graded subjectively as weak, moderate, or intense; and distribution was categorized as focal (< 10%), regional (10–50%), and diffuse (> 50%). Tumors that stained in moderate diffuse, intense regional, or intense diffuse patterns were considered “preserved.” Tumors that showed other patterns of staining (weak focal, weak regional, weak diffuse, moderate focal, moderate regional, and intense focal) and tumors that showed no staining were categorized as reduced expression.
Immunohistochemistry results were analyzed using the chi-square test for a table. The correlations between E-cadherin, DNMT1 expression, and/or methylation status of E-cadherin and clinicopathologic parameters were determined using the Fisher exact test. Cumulative survival curves were calculated by the Kaplan–Meier method according to the DNMT1 protein staining results, and the Log-rank test was used to compare statistical differences between groups. The endpoint of the current analysis was disease-related death. Differences and correlations at P values < 0.05 were considered significant.
E-Cadherin Methylation Status
To determine whether E-cadherin promoter hypermethylation occurred in tumor tissue compared with normal epithelial tissue, MSP was performed to determinate the methylation status of E-cadherin. Methylation of the E-cadherin gene promoter was tested first in eight normal epithelial samples. No E-cadherin methylation of the normal epithelial tissue was observed (Fig. 1A). Next, we investigated E-cadherin promoter methylation in 46 MEC samples from the same series. Methylation alleles were present in 33 of 46 samples (72%) (Table 1). The methylation status of E-cadherin promoter in low-grade, intermediate-grade, and high-grade, local, and advanced tumors is presented in Table 1. No significant correlations were noted between the methylation status of E-cadherin and tumor stage or histologic grade.
E-Cadherin and DNMT1 Staining
Immunostaining of E-cadherin was performed on all 46 MEC tumor samples and their corresponding normal tissues. E-cadherin was detected in the cell membranes of epithelia from all noncancerous mucosa. In our series of MEC, the staining was heterogeneous, with tumor cells demonstrating a diffuse decrease in membranous E-cadherin staining compared with the cells in normal, paired control tissues. Reduced E-cadherin immunohistochemical expression was observed in 21 of 46 tumor samples (45%), compared with 25of 46 tumor samples (55%) that exhibited preserved E-cadherin protein expression. The staining results of E-cadherin expression according to clinicopathologic parameters are shown in Table 1. No significant correlations were observed between E-cadherin expression and gender, location of tumor, histologic grade, or clinical stage (Table 1).
To determine whether the 5′-CpG island methylation in the E-cadherin promoter was associated with E-cadherin gene expression, we analyzed the correlation between E-cadherin expression and CpG methylation in 46 tissue specimens from patients with MEC. In the normal epithelium, in which E-cadherin expression was normal, all samples exhibited unmethylation. In tumor tissues, in which E-cadherin was heterogeneous, the promoters were mixed with methylated and unmethylated alleles (Fig. 1). Among the 21 MEC samples with reduced E-cadherin expression, unmethylated, heteromethylated, and methylated promoters were detected in 4 samples, 8 samples, and 9 samples, respectively. Among the 25 MEC samples with preserved E-cadherin expression, unmethylated, heteromethylated, and methylated promoters were detected in 9 samples, 14 samples, and 2 samples, respectively. There was a significant correlation between DNA methylation of CpG islands of the E-cadherin gene and reduced E-cadherin expression in the MEC samples (P = 0.021) (Table 2).
Table 2. Methylation Status of E-cadherin Promoter in Patients with Reduced and Preserved E-Cadherin Expression
Immunoreactivity for DNMT1 was detected in the nuclei, but not in the cytoplasm or cell membranes, of cells in the parabasal layer of epithelia, lymphocytes (which acted as the positive internal control for the analysis), and tumor cells (Fig. 2). To discriminate definitively positive samples from samples with background staining, if > 30% of the cells in a tissue sample exhibited nuclear staining, then the sample was categorized with positive immunoreactivity.16 Among the 46 MEC samples, 29 samples (63%) were positive for DNMT1. Correlations between the incidence of nuclear immunoreactivity for DNMT1 and the clinicopathologic features of patients with MEC are shown in Table 1. DNMT1 protein expression was associated significantly with histologic grade (P = 0.003) and clinical stage (P = 0.005). We also analyzed DNMT1 expression as it related to E-cadherin expression and methylation status. However, no significant correlation between DNMT1 protein expression and the preserved versus reduced E-cadherin immunostaining or methylation status of E-cadherin was observed (P = 0.562 and P = 0.307, respectively) (Table 3).
Table 3. Correlations between DNA Methyltransferase 1 Expression, E-Cadherin Expression, and Methylation Status of the E-Cadherin Promoter
E-cadherin (no. of patients)
Methylation status (no. of patients)
DNMT1: DNA methyltransferase 1; R: reduced expression; P: preserved expression; U: unmethylated; H: heteromethylation; M, methylation.
Negative (n = 17)
Positive (n = 29)
Total (n = 46)
Follow-up information was available on 39 patients, and the period ranged from 3 months to 125 months (mean, 42.5 mos). Patient who had negative DNMT1 expression had a longer survival (mean, 71.9 mos) compared with patients who had positive DNMT1 expression (mean, 31.9 mos). Kaplan–Meier survival curves were plotted for patients according to their DNMT1 protein expression levels. The cumulative survival rate of patients with DNMT1-positive MEC was significantly lower compared with the rate of patients with DNMT1-negative MEC (Fig. 3) (P = 0.0349; log-rank test).
Homotypic cell-cell adhesion of differentiated epithelial tissues largely is governed by E-cadherin.21 Diminished E-cadherin expression has been related to the acquisition of invasive properties within experimental tumors and in many advanced-stage human carcinomas.2 In our previous study, the perturbation of E-cadherin occurred frequently in MEC.3 To date, however, the regulatory mechanisms responsible for altered levels of this protein in MEC have not been elucidated. Mutation inactivation, one possible mechanism, that results in the uniform loss of E-cadherin expression has been detected only in other types of tumors.22, 23 However, according to the studies reported to date, the loss of E-cadherin expression does not occur as a result of mutations or deletion at the E-cadherin gene locus in MEC.24 Actually, in most common carcinomas, including MEC, the loss of E-cadherin expression is heterogeneous and transient,3, 25, 26 suggesting that the mechanism for E-cadherin loss in these tumors does not involve irreversible genetic alterations. Recent reports have indicated that aberrant hypermethylation of the E-cadherin promoter region, coupled with alterations in chromatin structure and transcription factor activity, may conspire to extinguish E-cadherin expression.8, 27 In the current study, we used laser microdissection to obtain pure populations of tumor cells from tumor samples. This method can eliminate contamination of the tumor or normal epithelial cells from the surrounding fibroblastic stroma cells. Our results demonstrated that promoter methylation was observed frequently and was correlated significantly with the loss of E-cadherin in MEC. Thus, the epigenetic silencing of the E-cadherin promoter by hypermethylation is an important mechanism for the silencing of this gene in MEC.
The current data demonstrate a significant association between decreased E-cadherin expression and E-cadherin gene promoter hypermethylation in the samples examined (Table 2). However, the aberrant gene promoter methylation cannot explain all loss of E-cadherin expression in MEC: As shown in Table 2, 4 samples without the methylated allele also were reduced for E-cadherin expression. Indeed, this phenomenon is also noted in other types of human malignancies28, 29; for example, the loss of E-cadherin expression in breast carcinoma occurs frequently, even in neoplasms without E-cadherin methylation.29 This may explain the finding that E-cadherin expression may be repressed by a mechanism other than promoter hypermethylation. For example, it has been shown that a novel glycoprotein, dysadherin, is a potential suppressor of E-cadherin.30 In addition, E-cadherin may be repressed by Snail transcription factor.31 Conversely, even when methylation occurred, some MEC samples also expressed E-cadherin. Our data showed methylated E-cadherin in 33 of 46 MEC samples (72%), whereas reduced E-cadherin expression was observed in 21 of 46 samples (46%). Results from recent studies may explain this discrimination between methylation and protein expression.32–35 First, methylation density influences transcription repression. Another report, for example, demonstrated an antagonistic relation between DNA methylation density and transcriptional activity.32–34 In tumor tissue, Graff et al. showed that, during disease progression, the expression of E-cadherin is diminished, and the density of promoter methylation is increased markedly.34 These findings open the possibility that the extent of the hypermethylation, rather than its presence or absence, is attributable to the loss of E-cadherin expression in MEC. Second, both other methyl-binding proteins and chromatin alterations are required to silence the gene.7, 35 For example, Darwanto et al. demonstrated that the loss of MeCP2 expression, a methyl-binding protein, was correlated with E-cadherin reexpression even in the presence of E-cadherin promoter methylation.35 Thus, it remains to be investigated whether such repression mechanisms and molecules may lead to the loss of E-cadherin expression in MEC.
In vertebrates, DNA methylation originates from an enzymatic transfer of a methyl group provided by the methyl donor, S-adenosylmethionine. A family of enzymes termed DNMTs catalyzes the conversion of cytosine to 5′-methylcytosine.36 DNMT1 was the first enzyme to be isolated as a mammalian DNMT.37 In tumors, controversial results between DNA methylation status and DNMT expression have been reported.15–18, 38 Some studies have shown a significant correlation between hypermethylated CpG islands and strong DNMT expression.15, 16 However, in other studies, investigations of CpG island methylation status failed to confirm this correlation.17, 18, 38 In the current study, we compared the expression of DNMT1 between MEC samples with and without hypermethylation of the E-cadherin promoter and between E-cadherin-positive versus E-cadherin-negative stained samples. The amount of DNMT1 expression did not have any obvious correlation with the expression and/or methylation status of the E-cadherin gene. There are several possible explanations for these findings: Although DNMT1 is a major DNMT in humans, 2 other enzymes, DNMT3a and DNMT3b, also possess DNMT activity.4 Genomic methylation patterns may be established through cooperation among these three DNMTs.39, 40 Moreover, studies have indicated that DNMT expression is coordinated in many malignancies,12 and DNMTs have different expression levels in variable tumors or in a tissue-specific manner.12, 13 Therefore, it is hypothesized that other DNMTs are required in silencing E-cadherin in MEC. In addition, methylation and E-cadherin expression are dynamic during tumor progression,34 and it is believed that abnormal DNMT expression is an early event in the development of tumors.41, 42 However, we examined their correlation only at a particular point in time. It is possible that the methylation change may occur with transiently up-regulated DNMT1 when the tumors developed; then, the aberrant methylation patterns were maintained thereafter with the relatively low, but still sufficient, enzyme levels. Actually, it is enzymes that carry out the majority of chemical reactions in cells, and the enzymatic activity of DNMT1 is regulated by both posttranscriptional and posttranslational mechanisms.43 In the current study, we examined whether the methylation status of E-cadherin was associated with the expression of DNMT1, but not with the enzymatic activities. Thus, it also is possible that the activity (and not the amount) of the enzyme is more important in this tumor. Further research will be needed to investigate the association of enzymatic activity of DNMT1 with the methylation status of E-cadherin. Previous studies also suggested that the aberrant promoter hypermethylation of the tumor suppressor genes was caused by locally restricted and gene-specific regulations involving DNMTs and their interaction proteins.4, 44 DNMT1 can interact with a number of different proteins that clearly may affect its nuclear localization as well as access to its DNA target sites in chromatin.44–46 For example, DNMT1 targets replication foci, in which DNA methylation patterns are copied from another strand by binding to proliferating cell nuclear antigen (PCNA).45 Taken together, these results indicate that DNMT1 expression alone may not explain regional DNA hypermethylation in malignancies. Certain unknown components of the DNA methylation machinery potently may target DNMT1 to substrate DNA, or DNMTs other than DNMT1 also may participate in regional DNA hypermethylation in MEC.
Although the DNMT1 immunostaining results did not correlate with E-cadherin methylation or expression, our data agreed with studies in colon, prostate, and hepatic carcinomas15, 47 that increased DNMT1 protein expression is a relatively common event in MEC and that the incidence of nuclear DNMT1 immunoreactivity is correlated significantly with the histologic grade and clinical stage of MEC. Furthermore, our current results demonstrated that patients with MEC who exhibit increased DNMT1 protein expression had a significantly lower survival rate compared with patients who did not exhibit such increased expression. In an earlier study, PCNA, a DNMT1 interaction protein, was overexpressed and correlated significantly with survival in patients with MEC.48 Therefore, increased DNMT1 and interactive protein expression may play a role in the malignant progression of MEC. Taken together, our results provide a clue that methylation may be involved in tumor progression, and protein expression analysis of DNMT1 may be a biologic predictor of poor prognosis for patients with MEC.
In conclusion, the current results demonstrated that promoter methylation may involve in the loss of E-cadherin expression in MEC. Up-regulation of DNMT1 may play a role in the development and prognosis of patients with MEC. Although up-regulation of DNMT1 expression is important, additional mechanisms and/or factors also may involve gene silencing of E-cadherin. Further studies on the correlation between DNMT1 and other components of the DNA methylation machinery in tissue specimens may extend our understanding of the basis of regional DNA hypermethylation during MEC carcinogenesis.