Esophageal cancer is one of the most serious cancers, and the outcome is fatal for the great majority of patients, regardless of the disease stage. The incidence of esophageal cancer is particularly high in China, India, Japan, Korea, the United Kingdom and the region around the Caspian Sea. The incidence of esophageal cancer is also rising in the United States, and the American Cancer Society estimates that ∼15,560 new cases of esophageal cancer will be diagnosed in the US during 2007. Despite significant advances in the detection and treatment of esophageal cancer, the prognosis of the disease remains dismal: the 5-year survival rate is about 10%.1 Even in patients who undergo curative surgical resection, the 5-year survival rate is only 25%.
The poor prognosis of patients with esophageal cancer is partially due to the high rate of recurrence, which occurs in approximately half of patients after curative surgical resection, and most patients die from the disease within 2 years of recurrence. Osugi et al.2 reported that more than 80% of tumor recurrence occurs within 24 months after the original resection of esophageal squamous cell carcinoma (ESCC). They also found that the prognosis in patients who experienced recurrence within 24 months after surgery was worse than that in those whose recurrence occurred more than 24 months after esophagectomy. Thus, the development of biomarkers for predicting recurrence and the implementation of efficient treatments for preventing recurrence after surgery in these patients are clearly imperative.
The aberrant methylation of normally unmethylated CpG islands has become widely recognized as a means to induce the transcriptional silencing of tumor suppressor genes in a variety of human cancers. Hypermethylation of CpG islands at the promoter region of more than 20 tumor suppressor genes has been reported in esophageal cancer [reviewed in ref.3]. In addition to clinical and histopathological factors that might aid in the prediction of patient outcome after curative resection of esophageal cancers, the prognostic significance of epigenetic alterations at a promoter region of a gene associated with esophageal cancer has been suggested in several reports.4, 5 We analyzed the aberrant methylation of the CDH1, p16, Wif-1, sFRP1, RARβ2, DAP kinase, integrin α4 genes in 251 ESCCs in order to identify a recurrence-associated epigenetic prognostic indicator after esophagectomy. Among those genes, p16,6–12CDH1,6, 13RARβ2,6, 14, 15DAP kinase,6, 7sFRP17 and Wif-116 genes are known to be frequently inactivated by the aberrant methylation of CpG islands at their promoter regions and to be important in the carcinogenesis of ESCC. Integrin α4 methylation was analyzed in this study due to high prevalence of hypermethylation in colorectal cancer (unpublished).
Material and methods
A total of 251 histologically proven ESCC patients who underwent definitive curative surgical resection at the Samsung Medical Center in Seoul, Korea between May 1994 and September 2001 participated in this study. Patients who had a second primary esophageal cancer or those with R1, R2, or p-T4 suspicious of remnant malignancy were excluded from this study. Patients who underwent preoperative adjuvant treatment or who died in the hospital after the operation were also excluded from this study. Adenocarcinoma was also not analyzed in the present study due to its low prevalence (∼1.5%) in Korea. Written informed consent for the use of paraffin-embedded tissues was obtained from all patients before operation. All procedures used in this study were approved by the Institutional Review Board at the Samsung Medical Center. All available paraffin blocks were reviewed by thoracic pathologists (J. Han & EY. Cho). Demographic and clinical information, including survival and recurrence, was obtained from medical records and follow-up data gathered by a well-trained interviewer. The pathologic stage was determined according to the tumor-node-metastasis (TNM) system of the American Joint Committee on Cancer (AJCC) esophageal cancer staging criteria.17
Esophageal resection was performed with transhiatal esophagectomy in 12 patients (5%), 2-field lymphadenectomy in 211 patients (84%) and 3-field lymphadenectomy in 28 patients (11%). The postoperative follow-up schedule was as follows: 1 month after surgery, every 3 months for the first 2 years, every 6 months thereafter for the next 2 years, and then once annually for life or more frequently if needed. Follow-up evaluation included chest X-ray, chest computed tomography scan (lower neck to upper abdomen), carcinoembryonic antigen and other serum chemistry at every visit. Endoscopy, abdominal ultrasonography, brain magnetic resonance imaging, and bone scan were performed if the patient was symptomatic. In cases whether the patient did not follow the postoperative follow-up schedule, specialized nurses called the patients to check on their health status.
Recurrence or death was evaluated from information obtained from our hospital records and those from other hospitals, as of October 31, 2006. Among the 131 patients with a tumor recurrence, 75 (57%) had a locoregional recurrence and 56 (43%) had a distant recurrence. Locoregional recurrence included local (at the site of the primary tumor, the anastomotic site) and regional (at lymph nodes or mediastinum, upper abdomen and cervical area) recurrence. Simultaneous locoregional and distant recurrence (in distant organs, pleura or peritoneum) was considered to be distant recurrence. Recurrence was suggested by a tumor occurring beyond 1 month after surgical resection and was confirmed by histologic/cytologic or imaging studies. Lymph node metastasis was defined by following criteria: (i) marginal enhancement with central necrosis on computed tomography, or (ii) greater than 1 cm in short-axis diameter, or (iii) persistent enlargement during follow-up. Recurrent tumors were treated by surgery alone (1%), chemotherapy alone (20%), radiotherapy alone (36%), or combined chemo-radiotherapy (5%). Forty-eight patients (38%) did not receive any therapy.
The data from the patients who died from causes other than esophageal cancer, those whose cancers did not recur before the end of the study, or those who were lost during the follow-up period were treated as censored data when calculating recurrence. The 251 patients consisted of 235 men (94%) and 16 women (6%), ranging in age from 39 to 78 years. The mean age at the time of surgical resection for primary esophageal cancer was 61.8 years. The median duration of follow-up after a curative resection was 3.3 years.
DNA extraction from paraffin block
Archival formalin-fixed, paraffin-embedded blocks containing at least 75% neoplastic tissues were cut at 10 μm. Consecutive serial tissue sections from each paraffin block were placed onto positively charged slides before DNA extraction and stained with H&E to evaluate the admixture of tumorous/nontumorous tissues. Areas corresponding to tumor were carefully microdissected from the surrounding normal stromal tissues, collected in autoclaved Eppendorf tubes, incubated overnight at 63°C in xylene for deparaffinization, and then vortexed vigorously. The supernatant was then discarded after centrifugation at full speed for 5 min. Graded ethanol was added to remove residual xylene and then removed by centrifugation. After ethanol evaporation, tissue pellets were carefully re-suspended in small volumes of lysis buffer ATL (DNeasy Tissue kit, Qiagen, Valencia, CA) and the genomic DNA was extracted using a DNeasy Tissue Kit according to the manufacturer's instruction.
Methylation-specific polymerase chain reaction
The methylation status of CpG islands at the promoter region of the CDH1, Wif-1, p16, sFRP1, RARβ2, DAP kinase, integrin α4 genes was determined by methylation-specific PCR (MSP), as described by Herman et al. (Fig. 1).18 Primer sequences and annealing temperatures for the MSP have been previously described by our group and others.4–6 CpGenome™ Universal Methylated DNA (Chemicon, Temecula, CA) was subjected to bisulfite modification and used as a positive-control for the methylated alleles. Bisulfite-modified DNA from peripheral blood lymphocytes of a healthy individual served as a positive control for the unmethylated alleles, and unconverted DNA from normal lymphocytes was used as a negative control for methylated alleles. Negative control samples without DNA were included in every PCR experiment.
The differences in clinicopathological characteristics and the methylation status of the 7 genes studied in patients with and without recurrence were analyzed using the Wilcoxon rank sum test and χ2 test (or the Fisher's exact test) for continuous and categorical variables, respectively. Multivariate logistic regression analysis was carried out in order to investigate the relationship between the development of recurrence and each of independent variables and to calculate odds ratios (ORs), after controlling for potential confounding factors. Covariates with p values <0.25 in the univariate analysis were subjected to multivariate analysis. The effect of promoter methylation on time to the death or recurrence was estimated by the Kaplan-Meier method, and the significance of differences in survival or recurrence between the 2 groups was evaluated by the log-rank test. Cox proportional hazards regression analysis was performed to estimate the hazard ratios of independent variables, after controlling for potential confounding factors. All tests of significance were two-sided, with a 5% type I error rate.
Clinicopathological characteristics and recurrence
The associations between disease recurrence and the clinicopathological features are listed in Table I. At a median follow-up of 39 months, 131 patients (52%) had developed disease recurrence: distant recurrence in 43% and locoregional recurrence in 57%. The mean ages of the patients with and without recurrence were similar (p = 0.19). Recurrence was more frequent in the male patients (53%) than in female patients (38%), but the difference was not statistically significant (p = 0.22). Poorly or moderately differentiated carcinomas were found to recur more frequently than well-differentiated tumors (p = 0.04). Tumors located in the upper or middle esophagus had a tendency to recur more frequently than those in the cervical or lower esophagus, but this difference was not statistically significant (p = 0.17).
Table I. Clinicopathological Characteristics (N = 251)
Patients with lymphatic invasion had a higher risk of recurrence than those without (76% vs. 49%, p = 0.007). Recurrence was associated with perineural invasion (p = 0.01) but not with vascular invasion (p = 0.06). A significant relationship was also found between recurrence and pathologic stage (p = 0.03). Recurrence occurred in 37% of the stage I cancers, in 48% of the stage II cancers, in 61% of the stage III cancers and in 67% of the stage IV cancers. The TNM grades were also significantly associated with recurrence (data not shown). Postoperative adjuvant therapy was performed in 69 (27%) of 251 patients: chemotherapy for 52 patients, radiotherapy for 15 patients, and chemo-radiotherapy for 2 patients. No significant relationship was found between recurrence and postoperative adjuvant therapy (p = 0.63).
Relationship between recurrence and CpG island hypermethylation
The relationship between recurrence and the CpG island hypermethylation of the 7 genes studied was investigated to identify biomarkers that would be useful for predicting recurrence-related prognosis after surgery in ESCC patients. Of the 251 ESCCs, CpG island hypermethylation was detected in 52% for p16, 25% for RARβ2, 43% for CDH1, 21% for integrin α4, 57% for sFRP1, 38 % for DAP kinase and 35% for Wif-1. Because tumor recurrence was found to be significantly associated with pathologic stage at the time of curative resection (p = 0.03; Table I), we stratified the data according to pathologic stage and analyzed the relationship between the hypermethylation of genes and the development of recurrence (Table II). For stage I cancers, recurrence was significantly associated with the CDH1 methylation, but not with the hypermethylation of the other genes studied. Recurrence was found in 61% (11 of 18) of patients with CDH1 methylation and in 24% (8 of 33) of those without, and this difference was statistically significant (p = 0.009).
Table II. Association Between Recurrence Afteresophagectomy and CpG Island Hypermethylation of Genes in Stage I and Stage II ESCCs
For stage II cancers, integrin α4 methylation was significantly associated with recurrence (p = 0.03). Recurrence occurred in 68% (15 of 22) of patients with integrin α4 methylation and in 41% (29 of 70) of those without. The prevalence of recurrence for stage III or stage IV cancers was not significantly different between patients with and without the hypermethylation of any gene (data not shown). Using recursive partitioning analysis, we also analyzed the relationship between the co-hypermethylation of any of the genes and recurrence, but found no relationship between them at all stages (data not shown).
Multivariate logistic regression analysis of recurrence
The data was stratified according to pathologic stage, and stratified multivariate logistic regression analysis was performed to control for the potential confounding effects of variables, such as age, sex, differentiation, tumor location, lymphatic invasion, vascular invasion and perineural invasion, postoperative adjuvant therapy, and to calculate the ORs (Table III). The risk of recurrence in stage I cancers with CDH1 methylation was determined to be 5.26 (95% CI = 1.48–18.67, p = 0.01) times greater than that of those without CDH1 methylation. For the stage II cancers, patients with integrin α4 methylation were found to be at a 3.03 (95% CI = 1.09–8.37, p = 0.03) times greater risk of recurrence than those without. However, there was no relationship between the hypermethylation of any gene and the risk of recurrence in the stage III and stage IV cancers (data not shown).
Table III. Multivariate Logistic Regression Analysis1of Recurrence in Patients with Stage I (N = 51) or Stage II (N = 92) ESCCs
Recurrence-free survival, survival after recurrence, and overall survival were analyzed in patients with and without the hypermethylation of any of the 7 genes studied. Kaplan-Meier survival estimates are shown in Figure 2. The recurrence-free 5-year survival rates were 63% for stage I cancers, 54% for stage II cancers, 41% for stage III cancers and 29% for stage IV cancers, and this difference was statistically significant (p < 0.001). Hypermethylation of CDH1 and integrin α4 genes was significantly associated with poor recurrence-free survival in stage I (Fig. 2a) and in stage II cancers (Fig. 2b), respectively. The recurrence-free 5-year survival rates for the 51 stage I cancers were 42 and 71% in patients with and without CDH1 methylation, respectively, this difference was statistically significant (p = 0.01; Fig. 2a). The recurrence-free 5-year survival rates for the 92 stage II cancers were 39% and 61% in patients with and without integrin α4 methylation, respectively (p = 0.01; Fig. 2b).
Survival after recurrence was examined with respect to the (co-) hypermethylation of any gene. Of the 19 patients that did experience recurrence in stage I, survival after recurrence was found to be extremely poor in patients with Wif-1 methylation (Fig. 2c). The median survival time after recurrence was 6.5 and 24 months in the recurrent stage I cancers with and without Wif-1 methylation, respectively. This difference was statistically significant (p = 0.0002). However, for the stage II cancers, the hypermethylation of any gene was not associated with survival after recurrence (data not shown).
The overall 5-year survival rates for stage I, II, III and IV cancers were 68, 53, 31 and 19% (p < 0.001), respectively. For nonrecurrent stage I cancers, hypermethylation of any gene was not associated with overall survival (data not shown). However, co-hypermethylation of the CDH1 and Wif-1 genes was significantly associated with overall survival in recurrent stage I cancers. The median survival time in patients with and without co-hypermethylation of both genes was 13 months and 45 months, respectively (p < 0.0001) (Fig. 2d). For the stage II cancers, hypermethylation of any gene studied was not associated with overall survival. In addition, hypermethylation of any gene in stage III and stage IV cancers was not associated with recurrence-associated survival (data not shown).
Cox proportional hazards analysis
Stratified Cox proportional hazards regression analysis was performed to determine whether hypermethylation of the CDH1,Wif-1, or integrin α4 genes was an independent prognostic factor, after controlling for potential confounding factors (Table IV). Recurrence types (locoregional or distant metastasis) and the presence of postoperative adjuvant therapy were not adjusted because the variables were not associated with survival in univariate analysis (data not shown). Recurrence-free survival in the stage I cancers was poorer in patients with CDH1 methylation than in those without CDH1 methylation (HR = 3.13, 95% CI = 1.21–8.09; p = 0.02). For the stage II cancers, integrin α4 methylation was significantly associated with short recurrence-free survival (HR = 2.12, 95% CI = 1.13–3.98; p = 0.03). The hazard of failure after recurrence in the stage I cancers was about 13.17 (95% CI = 2.46–70.41; p = 0.003) times higher in patients with Wif-1 methylation than in those without Wif-1 methylation, but not in the stage II cancers (HR = 0.66, 95% CI = 0.32–1.34; p = 0.25). The recurrent stage I cancers with co-hypermethylation of the CDH1 and Wif-1 genes were found to carry a 17.23 (95% CI = 3.14–94.45; p = 0.001) times greater risk of failure than those without co-hypermethylation of both genes. However, there was no relationship between hazard of failure and the hypermethylation of any gene in the stage III and stage IV cancers (data not shown).
Table IV. Stratified Cox Proportional Hazards Analysis1 in Stage I (N = 51) or Stage II (N = 92) ESCCs
To identify a recurrence-associated epigenetic prognostic indicator after esophagectomy in patients with ESCC, we retrospectively analyzed the relationship between recurrence and survival in patients with ESCC, and the promoter methylation of the 7 genes. It is critical for transformed residual cancer cells to acquire the ability to replicate limitlessly and invade adjacent tissues for recurrence. Of 7 genes examined, 1 group (p16 and RARB2) is involved in limitless replicative potential and the other (CDH1, integrin, Wif-1, sFRP1, DAPK) is involved in tissue invasion and metastasis. In addition to the contribution of CpG island hypermethylation to ESCC development, DNA methylation is also known to occur at precursor lesions of ESCC and to contribute to the progression of the dysplasia-carcinoma sequence in the development of ESCC. Hypermethylation of p16, CDH1 and RARβ2 genes occurs as early as basal cell hyperplasia or dysplasia in the development of ESCC.6, 11, 14 Ishii et al.7 also reported that hypermethylation of sFRP1 and DAPK is present at baseline low levels in normal and background non-neoplastic esophageal epithelium.
No association was found between the hypermethylation of the 7 genes studied and the recurrence-related prognosis in stages III and IV cancers. However, hypermethylation of CDH1 and integrin α4 genes was found to be associated with the recurrence in stage I and stage II cancers, respectively. It is unclear why epigenetic molecular indicators associated with the risk of recurrence in ESCC differed according to the pathologic stage of the disease. E-cadherin encoded by CDH1 is crucial in connecting cells and intergrin in mediating cell attachment to extracellular matrix (ECM). Although cell-to-cell and cell-to-ECM adhesion status was not evaluated in stage I and stage II cancers, one possibility is that hypermethylation of genes for the development of recurrence depend on the statuses of cell-to-cell and cell-to-ECM adhesion in each stage.
Although stage I esophageal cancer spreads slightly deep, it is usually found only in the top layers of cells lining esophagus and does not extend to nearby tissues, lymph nodes, or other organs. Therefore, CDH1 methylation in stage I cancers might contribute to the conversion of an epithelial cell to a fibroblastoid phenotype (epithelial-mesenchymal transition) by causing a loss of cell-to-cell contact and may eventually increase the risk of invasion. For stage II cancers, a loss of cell-to-cell contact might already occur in the most of stage II cancers, followed by the invasion of muscularis propria (T2) or adventitia (T3) irrespective of lymph node invasion. Integrin plays a crucial role in the attachment of cells to the ECM. Therefore, it is likely that integrin α4 methylation may contribute to the development of recurrence through the loss of cell-to-ECM adhesion in stage II cancers.
In the present study, CDH1 methylation in stage I cancers was associated with short recurrence-free survival but not with survival after recurrence, suggesting that CDH1 methylation may play a role at the early stage of recurrence development. E-cadherin mediates cell adhesion by forming a complex with β-catenin at plasma membrane and also sequesters β-catenin from the cytoplasm. The introduction of CDH1 into a cell line lacking E-cadherin is known to help sequester β-catenin and thus reduce the transcription of WNT target genes. The loss of E-cadherin increases β-catenin-dependent transcription19–21 by increasing the level of cytoplasmic β-catenin through the direct release of β-catenin into the cytoplasm.22, 23 However, several groups have also reported that the loss of E-cadherin expression did not result in constitutive β-catenin/TCF transcriptional activation. The inactivation of E-cadherin does not significantly increase the level of free cytosolic β-catenin, probably because the excess cytoplasmic β-catenin is rapidly removed by an intact degradation system. Accordingly, the lack of association between CDH1 methylation and the survival after recurrence in this study may result from the lack of constitutive β-catenin activation.
The epigenetic regulation of Wif-1 has been reported in a variety of cancers, such as breast cancer,24 bladder,25 colon,26, 27 malignant pleural mesothelioma,28 lung29 and esophageal cancer.17 However, their prognostic significances in ESCC remain to be elucidated.
The co-hypermethylation of CDH1 and Wif-1 observed was associated with poor overall survival in recurrent stage I cancers, suggesting an additive effect of 2 genes in the WNT/β-catenin pathway. The signaling function of β-catenin is principally regulated through the alteration of its stability. The components of the Wnt and cadherin pathways are linked through the activities of β-catenin and are known to combine β-catenin signaling. In the absence of Wnt signaling, β-catenin is rapidly degraded by a destruction complex consisting of APC, axin, glycogen synthase kinase 3-beta (GSK3β) and casein kinase.30 Wnt signaling induces the stabilization of the ‘free’ cytoplasmic pool of β-catenin by inhibiting β-catenin degradation which leads to the nuclear accumulation of β-catenin. The expression of the Slug/Snail family proteins, which are repressors of CDH1 gene transcription,31 lowers the threshold for activating Wnt signaling and thereby amplifies and/or sustains Wnt signaling by accumulating β-catenin.32 Although the order in which hypermethylation of Wif-1 and CDH1 genes occurred during the process of recurrence was not studied, it is reasonable to speculate that Wif-1 methylation occurs after CDH1 methylation in recurrent stage I cancers and contributes to disease progression by preventing β-catenin degradation and thereby amplifying the expression of β-catenin target genes.
Both sFRP1 and WIF-1 (Wnt inhibitory factor 1) are secreted extracellular molecules that block the Wnt signaling pathway primarily by directly binding to Wnt proteins and thereby preventing their access to cell surface receptors.33 The survival after recurrence in stage I cancers was significantly associated with Wif-1 methylation, but not with sFRP1 methylation. Several observations suggest that the levels of β-catenin and a variety of growth factor/receptor pathways, such as TGFβ signaling, coordinate the effect of Wnt antagonists on transcriptional activity of Wnt target genes.34, 35 Transfection of sFRP1 in HCT116 and SW480 cells decreases the levels of β-catenin, and resulted in growth inhibition and apoptosis. However, sFRP1 is also known to have inhibitory effects on engineered HCT 116 cells that contain a single wild-type β-catenin allele, whereas parental HCT116 cells with both wild-type and mutant alleles or HCT116 cells that contain only a mutant allele are insensitive to sFRP1.35 This observation suggests that the silencing of sFRPs would only be advantageous for growth before mutations in β-catenin have occurred. At later stages of tumorigenesis when cancer cells constitutively express high levels of β-catenin, the disruption of Wnt upstream of β-catenin would have only a minor effect on cell growth. In addition, the transcriptional levels of Wnt target genes activated by Wnt signaling alone, or by both Wnt and TGFβ signaling are also different.34, 35 Therefore, further study is needed to investigate the expression levels of TGFβ and its downstream effectors, TCF/LEF and β-catenin, at each pathologic stage in order to understand the exact role of Wif-1 methylation in recurrence-associated survival according to pathologic stages.
A portion of a preneoplastic lesion (i.e., a field) with genetically altered cells may remain in patients after surgical resection of the initial tumor, which is a high risk factor for the development of another carcinoma. However, data concerning the genetic alteration of preneoplastic cells in a field were not available in this study. Thus, we could not discriminate between a recurrent carcinoma that developed from residual cancer cells that were left in or near the “macroscopically normal” surgical margins and a second field tumor that developed from preneoplastic precursor cells clonally related to the cells of the index tumor,36 which is in line with the “field cancerization” concept37 or a true second primary tumor defined as an independently evolved carcinoma. Some of the second field tumors or second primary tumors might have been misclassified as locally recurrent tumors. The present study was severely limited by the small number of sample and therefore there may be a possible risk of a false positive conclusion. Additional studies on a large scale are needed to differentiate between a second field tumor or a second primary tumor and a recurrent tumor in order to validate the clinical usefulness of the genes.
In this study, a critical issue was to determine whether the strength of association between recurrence and methylation was uniform across stages. We tested that the population ORs were constant across stages. Breslow-Day test for homogeneity of the OR was applied, and we found evidences of heterogeneity in the cases of CDH1 (p = 0.02) and integrin α4 (p = 0.04). Therefore, the variable “stage” might be an effect modifier in this study, and we treated the data in the various contingency tables as if they have been drawn from distinct populations. We computed a different OR for each stage rather than a single summary values for the overall relative odds. In addition, we did not consider multiple comparisons problems in this study since we selected 7 genes on the basis of hypothesis. However, tests such as Bonferroni or FDR for multiple comparisons should be performed in a study of multiple genes selected randomly.
In conclusion, the present study suggests that hypermethylation of CDH1 and integrin α4 genes may be significantly associated with a recurrence-associated prognosis in stage I and stage II ESCCs, respectively. Thus, epigenetic prognostic indicators associated with the recurrence of ESCC may differ according to the pathological stage of the disease.
The authors thank Ms. Eun-Kyung Kim for her assistance in the collection and management of the data and Mr. Hoon Suh for his assistance in sample collection.