Loss of E-cadherin expression resulting from promoter hypermethylation in oral tongue carcinoma and its prognostic significance

Authors


Abstract

BACKGROUND

E-cadherin is expressed on the surface of normal epithelial cells. Loss of E-cadherin expression has been found in cancers and is postulated to facilitate tumor cell dissociation and metastasis. This study evaluated the role of promoter dense methylation in the downregulation of E-cadherin expression in oral tongue carcinoma.

METHODS

E-cadherin expression of 109 oral tongue carcinomas (93 primary tumors, 7 locally recurrent tumors, and 9 metastatic lymph nodes) was evaluated by immunohistochemical staining of tumor tissues. The methylation status of the CpG islands at the promoter region of E-cadherin which flanked five HpaII (methylation sensitive restriction enzyme) digestion sites were evaluated by methylation sensitive polymerase chain reaction in 86 tumors (70 primary tumors, 7 locally recurrent tumors, and 9 metastatic lymph nodes).

RESULTS

Underexpression of E-cadherin was found in 83% of primary tumors, 86% of recurrent tumors, and 89% of nodal metastases. Hypermethylated E-cadherin promoter was found in 64% of primary tumors, 71% of recurrent tumors, and 67% of nodal metastases. Downregulation of E-cadherin expression was found to be related to promoter hypermethylation. Consistently weak expression of E-cadherin by promoter hypermethylation was observed in primary tumors, their corresponding metastatic lymph nodes, and recurrent tumors. Downregulation of E-cadherin expression was a significant poor prognostic factor for survival.

CONCLUSIONS

Methylation of CpG sites at the promoter region played a key role in the inhibition of E-cadherin expression in both primary oral tongue carcinomas and their corresponding recurrences and nodal metastases. The resulting downregulation of E-cadherin expression had adverse effects on the prognosis of patients who were treated by primary surgery. Cancer 2002;94:386–92. © 2002 American Cancer Society.

Carcinoma of the oral tongue is one of the most common head and neck carcinomas. Nodal metastasis is an early event.1, 2 In our previous report, the nodal recurrence rate of untreated N0 neck carcinoma was found to be 47% for early Stage I and II tumors, staged in accordance with the American Joint Committee on Cancer 1997. Local and regional recurrences account for 90% of treatment failures post surgery and radiotherapy.3 The overall survival rate decreases as the carcinoma stage increases, from 75% for Stage I to 22% for Stage IV.3 The extent of locoregional tumor spread cannot be accurately evaluated clinically and radiologically before surgical treatment.4–6 With better understanding of the genetic abnormalities and the finding of more accurate prognostic markers, the management of tongue carcinoma may be improved in the future.

E-cadherin is expressed on the surface of normal epithelial cells. Loss of E-cadherin expression has been found in cancers and is postulated to facilitate tumor cell dissociation and metastasis. Diminished E-cadherin expression has been documented in association with the acquisition of invasiveness in vitro and poor prognosis for many carcinomas.7–9 Studies on the human and murine E-cadherin promoter have led to the characterization of several positive regulatory elements in the 5' proximal promoter, including a CCAAT-box, a GC-rich region, and an enhancer element.10–11 The factors interacting with the proximal GC-rich region and the enhancer element have been identified as AP2 and Sp1 transcription factors.11 Inactivation of E-cadherin is likely associated with aberrant CpG island hypermethylation.12–14 The binding of certain transcription factors has been reported to be inhibited by CpG methylation in their binding sites.15–16 Furthermore, methylation of specific CpG sites in HpaII and FnuDII methylation enzyme recognition sites were found to decrease the activity of retinoblastoma promoter.16

We have previously reported the finding of underexpression of E-cadherin in 85% of oral tongue carcinomas.17 The causal and prognostic relationships between hypermethylation of the promoter and the inactivation of E-cadherin expression in oral tongue carcinoma are still unclear. The current study addressed that question in patients who satisfied all the following criteria: (1) a diagnosis of oral tongue carcinoma, (2) a diagnosis of squamous cell carcinoma, (3) a primary surgical treatment without prior radiotherapy or chemotherapy, and (4) available surgical specimens.

MATERIALS AND METHODS

Patients and Tissues

Formalin fixed and paraffin embedded tumor tissues were taken from archival surgical specimens of 93 patients suffering from squamous cell carcinoma of the oral tongue who were treated at the Queen Mary Hospital, Hong Kong. These patients were all treated by primary surgery without prior radiotherapy or chemotherapy. There were 60 male and 33 female patients. The tumor grades were 32 well differentiated, 54 moderately differentiated, and 7 poorly differentiated tumors. The pathologic tumor stages were 21 patients at Stage I, 24 patients at Stage II, 21 patients at Stage III, and 27 patients at Stage IV. The median follow-up period after surgery for patients who were alive without tumor was 63 months.

Immunohistochemical staining was done for the primary tumors in all 93 patients, corresponding locally recurrent tumors in 7 patients, and metastatic lymph nodes in 9 patients.

A total of 86 tumor tissue samples had good quality DNA extracted from the same paraffin blocks for immunohistochemical study and were suitable for methylation study. These included 70 primary tumors, corresponding locally recurrent tumors in 7 patients, and metastatic lymph nodes in 9 patients.

For negative (weak or hypomethylated) external controls, we used one short-term primary adenoid culture and 11 formalin fixed, paraffin embedded tissue biopsies, six of which were from histologically normal adenoidectomy nasopharyngeal epithelia and five of which were from histologically normal distant mucosal epithelia of oral tongue carcinomas. One oral tongue squamous cell carcinoma cell line (HN106) was used as a positive (hypermethylated) external control. Extraction of DNA was performed as described.12

Immunohistochemistry

Immunohistochemic staining procedure for E-cadherin was performed by standard method using monoclonal antibody.17 The antigen retrieval was performed by microwave in citrate buffer (pH 6). We used the monocloncal antibodies E-cadherin at dilution of 1:100 (Transduction Laboratories, Lexington, KY). The incubation was carried out overnight at 4 °C. Expression of E-cadherin in six normal adenoid epithelia were used as external positive controls and also as negative controls by omitting the monoclonal antibody during the procedure. The normal epithelium adjacent to carcinoma in each specimen should have strong staining and served as internal positive control. The normal tongue epithelia were also used as internal negative controls by omitting the addition of monoclonal antibody during the immunohistochemic staining procedure. Blocks without normal epithelia or normal epithelia with weak staining were excluded from this study.

All tumor cells within the representative slides were examined, and the percentage of tumor cells with positive expression of E-cadherin was evaluated. Semiquantitative four-point scale scoring was used according to the percentage of cells with weak expression of E-cadherin (0, ≤ 10%; 1+, 11-25%; 2+, 26–50%; 3+, > 50%). Weak expression (decreased expression or underexpression) is defined for specimens with a score of 0,1, or 2, and strong expression is defined for a score of 3 or above.

Experimental Design for Methylation Study

The methylation sensitive enzyme digestion and nested PCR method was used in this study. E-cadherin promoter region and the diagrammatic illustration of the methylation sensitive PCR protocol are outlined in Figure 1.

Figure 1.

Schematic outline of the methylation sensitive polymerase chain reaction (PCR) analysis. A) Map of E-cadherin promoter region. B) Restriction enzyme digestion and nested PCR of the promoter region. Of the PCR products shown in gel electrophoresis, the MspI+ lane indicates MspI pre-digested specimen, the HpaII+ lane indicates HpaII predigested specimen, and the HpaII− lane indicates a specimen without enzyme predigestion.

The sequences of the nested PCR primers were: 5' GCTGCTGATTGGCTGTGGCCGG 3' (outer forward primer, W1), 5' CTGCAGCAGCAGCAGCAGCGCC3' (outer backward primer, W2), 5' GTGAACCCTCAGCCAATCAGCGGTAC 3' (inner forward primer, W3) and 5' GCTCCAAGGGCCCATGGCTGGCCG 3' (inner backward primer, W4).

Each specimen had three sets of PCR to amplify a 216 bp fragment:

  • Each specimen had PCR without any restriction enzyme digestion to serve as a control of PCR amplification. A 216 bp PCR product should have been amplified in each specimen;

  • Each specimen had PCR following MspI methylation insensitive enzyme digestion. The MspI methylation insensitive enzyme cut the CCVGG recognition site and served as a control for successful enzyme digestion. No PCR product should be amplified after Msp I digestion in each specimen;

  • Each specimen had PCR following HpaII methylation sensitive restriction enzyme digestion. There were five CCVGG methylation sensitive digestion sites within the amplified fragment. In the absence of promoter methylation or weak methylation, the cutting of one or more of these CCVGG sites by HpaII would prevent PCR amplification. The presence of promoter hypermethylation of all 5' CCGG sites would prevent HpaII restriction enzyme digestion, and we could thus amplify a 216 bp PCR product.

PCR Analysis of the E-cadherin Promoter

PCR was used to amplify 200 ng of either MspI/HpaII digested or undigested DNA. The PCR reaction was performed in a total volume of 25 μL containing 0.5 μM oligomers 1 and 2, 150μ M each dNTP, 2.5mM MgCl2, 50mM KCl, 10mM Tris-HCl, 5% DMSO, and 0.8 U of AmpliTaq Gold polymerase. Samples were amplified for 40 cycles made up of 45 seconds at 94 °C, 30 seconds at 57 °C, and 30 seconds at 68 °C, with initial denaturation at 94 °C for 12 minutes. One-fifteenth volume of this PCR mixture was added into another reaction mixture. The amplification was repeated as described above, except internal primers 3 and 4 were used. The PCR products were then analyzed by an 8% acrylamide/bis gel.

RESULTS

Evaluation of E-cadherin Expression by Immunohistochemistry

Of the 93 primary tumors, the E-cadherin expression scores were 37 tumors with a score of zero, 25 tumors with a score of 1, 15 tumors with a score of 2, and 16 tumors with a score of 3. Of the seven locally recurrent tumors, there were two tumors with a score of 0, one tumor with a score of 1, three tumors with a score of 2, and one tumor with a score of 3. Of the nine metastatic lymph nodes, there were four tumors with a score of 0, one tumor with a score of 1, three tumors with a score of 3, and one tumor with a score of 3. Illustration of the E-cadherin expression is shown in Figure 2. Weak expression of E-cadherin was found in 83% of primary tumors, 86% of locally recurrent tumors, and 89% of metastatic nodes.

Figure 2.

Immunohistochemical staining of E-cadherin of oral tongue carcinoma. The carcinoma cell membranes are stained brown. Carcinoma cells with downregulation of E-cadherin have no membranous stain.

Absence of Hypermethylation of E-cadherin Promoter Region in Normal Nasopharyngeal and Oral Tongue Epithelia Controls

The application of methylation sensitive PCR analysis was demonstrated on 11 normal controls, which included six normal adenoid tissue biopsies and five histologically normal oral tongue mucosal tissues at margin of surgical specimens, all of which were also preserved in formalin and embedded in paraffin (Fig. 3). Fragments of 216 bp were detected in all 11 normal controls without enzyme pretreatment. These fragments indicated the feasibility of the PCR assay on degraded DNA from tissue biopsies already preserved in formalin. Conversely, no band was detected from HpaII pre-digested samples. This indicated that the E-cadherin promoter in all six adenoid epithelia and five surgical margin histologically normal oral mucosa controls were not hypermethylated, and could be cut by HpaII, which means that the fragment was not amplifiable by PCR. As expected, no detectable band was observed from MspI pre-digested samples, indicating complete restriction enzyme digestion.

Figure 3.

DNA from six adenoid tissue biopsies (specimens N1 to N6) were amplified by polymerase chain reaction (PCR) after enzyme digestion. H− lanes indicate PCR control without any restriction enzyme predigestion of DNA. H+ lanes indicate HpaII predigestion of specimen DNA. M+ lanes indicate MspI predigestion of specimen DNA. MW indicates 50 bp molecular weight ladder.

Sensitivity and Specificity of Methylation Sensitive PCR Analysis

In view of the heterogeneity of tissue specimens composed of a mixture of methylated and unmethylated cells, we tested the sensitivity and specificity of PCR by mixing a different concentration of methylated DNA from one head and neck squamous cell carcinoma (HNSCC) cell line with unmethylated DNA from a short-term primary normal epithelial cell culture (Fig 4). The specificity of the assay was demonstrated on the samples mixed with different concentrations of DNA from normal epithelial cells and from HNSCC cells. The intensity of the 216 bp fragment was more visible as the methylated DNA concentration was gradually increased. The sensitivity of this assay was determined by the detection of a 216 bp fragment from 50 ng of methylated DNA (Fig 4, lane 3). The results showed that the presence of unmethylated DNA from normal cells or tumor cells had little influence on the amplification of DNA from hypermethylated tumor cells (Fig 4, lanes 1 and 2). We applied 200 ng DNA from tissue specimens for the subsequent PCR analysis.

Figure 4.

Sensitivity and specificity of methylation sensitive polymerase chain reaction (PCR) analysis. After HpaII pre-digestion, the mixtures of different concentration of DNA from normal adenoid epithelia and head and neck squamous cell carcinoma cell culture were amplified by PCR.

Correlation of E-cadherin Expression with Methylation Status of E-cadherin Promoter in Oral Tongue Carcinomas

Methylation status of E-cadherin gene was evaluated on 70 primary tumors, 7 recurrent tumors, and 9 metastatic nodes, and the representative results are shown in Figure 5. Of the 70 primary tumors, 45 (64%) had hypermethylation. Hypermethylation was found in 76% (44 out of 58) tumors with weak E-cadherin expression, compared with 8% (1 out of 12) tumors with strong E-cadherin expression (chi-square, P <0.001; odds ratio = 0.029, 95% confidence interval 0.003-0.244).

Figure 5.

Representative results of methylation sensitive polymerase chain reaction analysis of primary tumors, metastatic lymph nodes, and recurrent tumors. The corresponding immunohistochemical staining scores are also shown at the bottom.

Of the seven recurrent oral tongue carcinomas, five (71%) had cells with methylated E-cadherin promoter regions. Of the six tumors with weak expression of E-cadherin, four (67%) had hypermethylation. All seven tumors showed concordant methylation results with their corresponding primary tumors.

Of the nine metastatic lymph nodes, six (67%) had tumor cells with hypermethylation of E-cadherin promoter regions. Among the eight metastatic nodes with weak E-cadherin expression, six (75%) had promoter hypermethylation. Of the one metastatic lymph node with strong expression of E-cadherin, no hypermethylation was detected. Eight metastatic nodes showed concordant hypermethylation results consistent with their corresponding primary tumors. One tumor with weak E-cadherin expression in both the lymph node and the primary tumor showed promoter hypermethylation in the primary tumor but not in the corresponding metastatic lymph node.

Clinicopathologic Significance of E-cadherin Underexpression from Hypermethylation

Primary tumor expression of E-cadherin was evaluated for the correlation with tumor grade, pathologic stage, pathologic T stage, and nodal metastasis (defined by either the presence of pathologic evidence of nodal metastasis before treatment or nodal recurrence not related to local recurrence) by appropriate univariate analysis, all parameters were not statistically significant (all with respective P values > 0.05).

The 5-year tumor–free actuarial survival rate was 52% for patients with weak E-cadherin expression, compared with 79% for patients with strong E-cadherin expression (Life-table method, Wilcoxon statistics, P = 0.037). The survival curves are illustrated in Figure 6. The pTNM stage, pT stage, and pN stage were also significant prognostic factors for survival in univariate analysis (Life-table method, Wilcoxon statistics, P < 0.0001, P = 0.0001 and P = 0.0005 respectively). Other factors, including gender and tumor grade, were not statistically significant (Life-table method, Wilcoxon statistics, all with P > 0.1). Both Cox regression and logistic regression multivariate analysis recruiting pTNM stage, pT stage, pN stage, and E-cadherin expression showed that pT stage, pN stage, and E-cadherin were significant independent prognostic factors for tumor-free survival.

Figure 6.

Actuarial survival curves of patients with strong (black square) and weak (empty circle) expression of E-cadherin.

DISCUSSION

The current study shows that E-cadherin promoter hypermethylation occurred frequently (64%) in oral tongue carcinoma and was the main cause of diminished expression of E-cadherin. The finding of E-cadherin downregulation as a result of the promoter hypermethylation is also consistent with findings in carcinomas of the breast, prostate, thyroid, and liver.14, 24–26

The selected promoter region (positions −225 to 15) in the current study is located closely to the transcription start site of E-cadherin. The current results are consistent with other studies which indicated the importance of this region for E-cadherin transcription.27, 28 In a study by Behrens et al, an upstream fragment (positions −178 to +92) was found to mediate the strong expression of a chloramphenical acetyl-transferase reporter gene in epithelial cells, whereas this promoter was either inactive or much less active in non-epithelial cells.27

The selected promoter region contains 23 CpG sites that flanks five HpaII and three HhaI digestion sites. It also contains part of the FnuDII recognition site, as well as activating transcription factor (ATF) and Sp1 recognition sites. It has been documented that methylation of specific CpG sites in some methylation enzyme recognition sites can abolish promoter activity. The methylation of HpaII and FnuDII CpG sites were found to decrease the activity of the retinoblastoma promoter.10 It has also been documented that methylation of CpG inhibits the binding of transcription factor, thus resulting in transcriptional inactivation.29 One possible mechanism through which DNA methylation can regulate gene expression is the prevention of binding of transcription factors by the addition of 5-methyl cytosine, which protrudes into the major groove of the DNA helix.30 It has been reported that, in an in vitro system, specific transcription factors bind with less affinity to methylated target sequences.31 ATF is found to be essential in the expression of the retinoblastoma gene. However, the binding of ATF is inhibited when those recognition sequences are CpG methylated.16 Despite previous reports that binding of the Sp1 transcription factor is not significantly affected by methylation of the CpG dinucleotide,32, 33 methylation of both cytosines (m)Cp(m)CpG within its binding site, 5'-GGGCGG (lower strand, 5'-CCGCCC), was found to inhibit binding by 95%.34 Since the selected promoter region for analysis includes two important transcription factors, ATF and SP1, the binding of these transcription factors are inhibited by the promoter dense methylation and lead to the inactivation of E-cadherin expression. The current data highly suggest that the binding of transcription factors ATF and SP1 is essential for the expression of E-cadherin and is inhibited by dense methylation at the E-cadherin promoter region in oral tongue carcinoma.

The high frequency of aberrant methylation of the promoter region in association with E-cadherin transcription silence was consistently found in primary tumors and their corresponding nodal metastases and local recurrences. With only a few exceptions, most locally recurrent tumors and metastatic lymph nodes showed concordant results consistent with their corresponding primary tumors. This suggests that methylation in association with loss of E-cadherin expression can persist in invasive and metastatic lesions during malignant progression. Similar conclusions have been documented in several studies.24, 35 Since each tumor has a heterogeneous population of tumor cells, and there are many other genetic abnormalities in relation to invasion and metastasis, discordant patterns of methylation and expression can be found in a small percentage of tumors. This is in agreement with several observations.24, 36

It has been reported that the loss of E-cadherin expression is correlated with tumor grade in some poorly differentiated tumors compared with well-differentiated tumors.24, 37 In the current study, there was no significant correlation of reduced E-cadherin expression with the loss of differentiation in oral tongue carcinoma. However, the loss of E-cadherin was an important prognostic factor for survival. The downregulation of the adhesion molecule facilitates the dissemination of cancer cells. The higher invasive and metastatic potentials ultimately lead to the higher risk of recurrence and tumor related death after treatment for oral tongue carcinoma. This invasive and metastatic potential may not equally affect cancers treated with chemotherapy or radiotherapy compared with surgery. This is particularly important for oral tongue carcinoma, which is primarily treated with surgery. The poor prognosis, in association with weak expression of E-cadherin, may be a clinically useful tumor marker in conjunction with other clinicopathologic parameters in the decision for more aggressive management of these patients. A wider surgical resection margin or the addition of adjuvant radiotherapy or chemotherapy may be other options for consideration for patients with poor prognosis. However, the efficacy requires further evaluation.

In conclusion, methylation of CpG sites at the promoter region of E-cadherin is an important cause of downregulation of E-cadherin expression in oral tongue carcinomas. The methylation patterns are consistently found in primary tumors and their corresponding metastatic lymph nodes and recurrent tumors. The downregulation of E-cadherin as a result of hypermethylation has significant adverse effect on the survival of patients with surgical treatment.

Ancillary