Fax: (011) 02 34103649
Aberrant methylation of H-cadherin (CDH13) promoter is associated with tumor progression in primary nonsmall cell lung carcinoma
Article first published online: 21 SEP 2005
Copyright © 2005 American Cancer Society
Volume 104, Issue 9, pages 1825–1833, 1 November 2005
How to Cite
Seuk Kim, J., Han, J., Shim, Y. M., Park, J. and Kim, D.-H. (2005), Aberrant methylation of H-cadherin (CDH13) promoter is associated with tumor progression in primary nonsmall cell lung carcinoma. Cancer, 104: 1825–1833. doi: 10.1002/cncr.21409
- Issue published online: 17 OCT 2005
- Article first published online: 21 SEP 2005
- Manuscript Accepted: 31 MAY 2005
- Manuscript Revised: 29 APR 2005
- Manuscript Received: 9 FEB 2005
- Samsung Biomedical Research Institute. Grant Numbers: B-A5-102, B-A5-013
- nonsmall cell lung carcinoma;
Abnormalities in the H-cadherin gene have been reported in several human malignancies, including nonsmall cell lung carcinoma (NSCLC). Aberrant methylation of the H-cadherin promoter also has been reported in NSCLC, but its clinical significance remains to be elucidated.
The authors studied H-cadherin methylation in 305 patients with NSCLC to gain a further understanding of the clinicopathologic and prognostic significance of H-cadherin methylation in patients with NSCLC. The methylation status of the H-cadherin gene was investigated by using methylation-specific polymerase chain reaction analysis in paraffin blocks from 305 patients with NSCLC. Ki-67 expression was assessed by immunohistochemical staining. All statistical analyses were 2-sided with a 5% Type I error rate.
H-cadherin methylation was observed in 130 of 305 tumor samples (43%). The prevalence of H-cadherin methylation was associated significantly with pathologic stage and was observed in 44% of patients with Stage I disease, in 23% of patients with Stage II disease, in 59% of patients with Stage III, and in 88% of patients with Stage IV disease (P = 0.001). H-cadherin methylation occurred with a 2.71 times greater prevalence (95% confidence interval [95% CI], 1.21–6.09; P = 0.01) T2 tumors than in T1 tumors and with a 3.78-fold greater prevalence (95% CI, 1.05–13.59; P = 0.04) in T3 tumors than in T1 tumors. However, lymph node metastasis was related inversely with H-cadherin methylation (odds ratio = 0.51; 95% CI, 0.28–0.95; P = 0.03), and H-cadherin methylation was not associated with the Ki-67 labeling index (P = 0.53) or with tumor size (P = 0.89). No relation was found between H-cadherin methylation and survival in patients with Stage I NSCLC (P = 0.51) or in patients with Stage II NSCLC (P = 0.46).
The current findings suggested an association between H-cadherin methylation and tumor progression in NSCLC but had no prognostic significance in patients with early-stage NSCLC. In addition, H-cadherin methylation may be a valuable candidate molecular marker for the early detection of NSCLC. Cancer 2005. © 2005 American Cancer Society.
Lung carcinoma is one of the most common causes of cancer deaths in the world. Despite much progress in the treatment and detection methods of lung carcinoma, the prognosis remains poor. This situation results largely from micrometastasis, which is present in greater than two-thirds of patients at the time of diagnosis.1 Therefore, it clearly is imperative that efficient diagnostic methods be developed to detect lung carcinoma at the earliest stages, during which curative surgical resection remains feasible. One promising approach to early detection involves the identification of lung carcinoma-specific molecular biomarkers, especially those associated with the initiation and progression of lung carcinoma in its earliest stage. Screening tests on long-term smokers used to date (radiography and sputum cytology) have failed to reduce lung carcinoma mortality.2
The cadherins are a class of tumor suppressor genes involved in the pathogenesis of lung carcinoma. They are a superfamily of Ca2+-dependent intercellular adhesion molecules that play a role in maintaining intercellular connection in a homophilic and subclass-specific manner.3 It was found that a loss of cadherin expression led to transition from benign tumor, to invasive tumor, and the subsequent metastatic dissemination of tumor cells by causing changes in cell-cell and cell-matrix adhesion.4, 5 The H-cadherin gene, a new member of the cadherin superfamily, was isolated recently and has been mapped to 16q24.6 The introduction of H-cadherin in human breast carcinoma cells reduced their invasive potential and markedly decreased their growth rate; in addition, it induced the reversion of morphology from an invasive type to a normal cell-like type.7, 8 Abnormalities in the H-cadherin gene have been identified in human malignancies, including lung carcinomas,9–12 and H-cadherin expression is diminished in lung and breast carcinomas.7, 9 Moreover, H-cadherin expression is associated with tumorigenicity in nonsmall cell lung carcinoma (NSCLC) and frequently is silenced by promoter methylation of the H-cadherin gene in NSCLC.9, 10, 12, 13
Aberrant methylation of CpG islands within the promoters of tumor suppressor genes is one of the most frequently acquired epigenetic changes during lung carcinoma carcinogenesis.13–16 In the current study, we investigated the aberrant methylation of H-cadherin promoter in 305 primary NSCLCs to assess the clinicopathologic and prognostic significance of H-cadherin methylation.
MATERIALS AND METHODS
Study Population and Specimens
In total, 305 patients who underwent curative surgical resection for NSCLC at the Department of Thoracic Surgery, Samsung Medical Center, Seoul, Korea, from April 1994 to September 2003 participated in this study. The surgically removed tissue samples were collected from these patients after obtaining appropriate Institutional Review Board permission and after written informed consent was obtained from all patients. Postoperative follow-up was scheduled at 1 month, 2 months, and every 3 months during the first 2 years after surgery and every 6 months thereafter or more frequently if needed. Whenever patients did not keep the postoperative follow-up schedule, specialized nurses called patients and checked their health status. Information regarding smoking habits, demographics, and lifestyle factors were obtained by using an interviewer-administered questionnaire. Neoadjuvant chemotherapy consisted of paclitaxel and cisplatin for patients with Stage IIIA NSCLC. Adjuvant chemotherapy for patients with Stage IIIA disease was limited to those who had lymph node involvement. Patients with Stage IIIB and IV NSCLC were treated with chemotherapy or radiotherapy. The 305 patients with tissue samples available ranged in age from 29 years to 83 years. One hundred eighty-three patients had Stage I disease, 70 patients had Stage II disease, 44 patients had Stage III disease, and 8 patients had Stage IV disease.
Genomic DNA Isolation from Paraffin Blocks
For genomic DNA extraction from paraffin blocks, formalin fixed, paraffin wax embedded tissues were cut into 10-μm sections. Before DNA extraction, the sections were placed on slides and stained with hematoxylin and eosin to evaluate the admixture of tumorous and nontumorous tissues. Areas that corresponded to tumor or to surrounding normal lung tissue were microdissected separately. Microdissected tissues were collected in 15-mL centrifuge tubes and deparaffinized overnight at 63 °C in xylene. After centrifugation at full speed for 5 minutes, the supernatant was removed. Ethanol was added to the pellet to remove residual xylene and was then removed by centrifugation. After ethanol evaporation, the tissue pellet was resuspended in lysis buffer ATL (DNeasy Tissue kit; Qiagen, Valencia, CA), and genomic DNA was isolated according to the manufacturer's instructions.
One microgram of genomic DNA from a paraffin block was denatured by incubating it with 0.3M NaOH for 10 minutes at 37 °C. The denatured DNA was then treated with 3 M sodium bisulfite, pH 5.0, and 10 mM hydroquinone, both freshly prepared. The sample was mixed gently, overlaid with mineral oil to cover the surface of the aqueous phase, and incubated in the dark at 55 °C for 16 hours. The DNA was recovered from under the mineral-oil layer after snap freezing at − 70 °C for 10 minutes and removing the unfrozen oil. The bisulfite-modified DNA was then purified using Wizard™ DNA purification resin according to the manufacturer's recommendations (Promega Corp., Madison, WI), desulphonated with 3 M NaOH, precipitated with ethanol, and dissolved in 20 μL of 5 mM Tris, pH 8.0.
Methylation-Specific Polymerase Chain Reaction
The methylation status of the H-cadherin gene was determined by methylation-specific polymerase chain reaction (PCR) (MSP) (Fig. 1A), as reported by Herman et al.,17 using primers specific for the DNA that were either methylated or unmethylated at the promoter region of H-cadherin gene. The primers used for unmethylated H-cadherin were 5′-TTGTGGGGTTGTTTTTTGT-3′ (sense) and 5′-AACTTTTCATTCATACACACA-3′ (antisense), and the primers for methylated H-cadherin were 5′-TCGCGGGGTTCGTTTTTCGC-3′ (sense) and 5′-GACGTTTTCATTCATACACGCG-3′ (antisense). The PCR mixture contained 1 × PCR buffer (50 mM KCl; 67 mM Tris, pH 8.7; 1.5 mM MgCl2), deoxynucleotide triphosphates (1.25 mM each), primers (300 ng of each per reaction), 2.5 units of Taq polymerase, and bisulfite-modified DNA (50 ng of DNA from paraffin blocks). PCR amplification consisted of 1 cycle at 95 °C for 5 minutes, 35 amplification cycles (at 95 °C for 30 sec, 60 °C for 1 min, and 72 °C for 30 sec), and a final extension at 72 °C for 10 minutes. DNA from the peripheral blood lymphocytes from healthy individuals were used as negative controls for the methylation-specific assays. Lymphocyte DNA from healthy volunteers was treated with Sss1 methyltransferase (New England BioLabs, Beverly, MA), then treated with bisulfite, and used as a positive control for methylated alleles. PCR products were observed on 2% agarose gels that were stained with ethidium bromide.
Immunohistochemical Analyses of Ki-67
Formalin fixed and paraffin embedded tissues in 5-μm-thick sections were deparaffinized in xylene and rehydrated through a series of alcohols. Endogenous peroxidase activity was blocked with 5% H2O2 in methanol for 10 minutes. For antigen retrieval, the sections were treated in 10 mmol/L citrate buffer, pH 6.0, for 10 minutes at 120 °C in an autoclave, and nonspecific reactions were blocked with 5% horse serum in phosphate buffered saline. The sections were then incubated with mouse monoclonal anti-Ki-67 monoclonal antibody (DAKO; clone MIB-1) overnight at 4 °C and then treated with biotinylated antimouse immunoglobulin G, Vectastain Elite avidin-biotin complex reagent (Vector Laboratories, Burlingame, CA), and 3.3′-diaminobenzidine tetrahydrochloride as the chromogen, respectively. All sections were counterstained with hematoxylin. For negative controls, primary antibody was omitted from the reaction sequence. Cytoplasmic reactivity was disregarded for scoring Ki-67 staining (Fig. 1B, C), and only nuclear staining above the levels of any cytoplasmic background was considered evidence of the expression of Ki-67. The fraction of Ki-67-positive cells (the Ki-67 labeling index) was defined as the percentage of cells that showed positive Ki-67 nuclear staining as a fraction of total resting cells.
Continuous variables were tested for normality using the Shapiro–Wilkinson test. The Wilcoxon rank-sum test (or t test) and the Fisher exact test (or the chi-square test) were used for the univariate analysis of continuous and categorical variables, respectively. Multivariate logistic regression was conducted to determine the correlation between H-cadherin methylation and covariates that were identified as statistically significant by univariate analysis and to calculate odds ratios (ORs). Covariates with P values < 0.25 in the univariate analysis were subjected to multivariate analysis. The effect of H-cadherin methylation on the time to death was evaluated using the Kaplan–Meier method, and differences between two groups were compared using the log-rank test. Cox proportional hazards regression analysis was used to estimate the hazard ratios of independent survival factors after controlling for potential confounding factors. All statistical analyses were 2-sided with a 5% Type I error rate.
The association between H-cadherin methylation in the 305 tumor samples and the clinicopathologic features of the patients are listed in Table 1. H-cadherin methylation was detected in 130 of 305 samples (43%). The mean age of patients was 58 years, and this was similar in patients with and without H-cadherin methylation. H-cadherin methylation occurred in 85 of 218 men (39%) and in 45 of 87 women (52%), and this difference was statistically significant (P = 0.04). Hypermethylation of the H-cadherin gene was not associated with smoking (P = 0.32).
|Absent (n = 175)||Present (n = 130)||P value|
|Agea||58 ± 11||59 ± 12||0.91|
|Pack yrsa||59 ± 36||60 ± 41||0.82|
|Squamous cell carcinoma||93||49|
A statistically significant association was found between H-cadherin methylation and pathologic stage. H-cadherin methylation occurred in 81 of 183 patients (44%) with Stage I disease, in 16 of 70 patients (23%) with Stage II disease, in 26 of 44 patients (59%) with Stage III disease, and in 7 of 8 patients (88%) with Stage IV disease (P = 0.001; Fisher exact test) (Fig. 2A). H-cadherin methylation occurred at a lower frequency in Stage II than in the other stages. In the current study, the main difference between Stages I and II was the extent of regional lymph node involvement, because we did not have any patients with T3N0M0 disease. Therefore, we also investigated the association between H-cadherin methylation and the tumor, lymph node, and metastasis (TNM) staging variables separately (Fig. 2). H-cadherin methylation was detected in 31% of T1 tumors, in 42% of T2 tumors, in 53% of T3 tumors, and in 75% of T4 tumors (P = 0.005) (Fig. 2B), and it was associated significantly with distant metastasis (42% of patients with M0 status; 88% of patients with M1 status; P =0.02) (Fig. 2D). However, H-cadherin methylation was associated inversely with regional lymph node involvement (47% of patients with N0 status vs. 34% of patients with N1–N3 status; P = 0.03) (Fig. 2C). Moreover, H-cadherin methylation occurred more frequently in adenocarcinoma (49%) than in squamous cell carcinoma (35%; P = 0.03), consistent with previous findings.3, 15 No significant correlation was found between H-cadherin methylation and recurrence or differentiation.
Tumor Growth and H-Cadherin Methylation
We examined the correlations between H-cadherin methylation and tumor size and the Ki-67 index. The average tumor size was 3.71 cm ± 1.8 cm and ranged from 0.6 cm to 14.8 cm. No correlation was found between H-cadherin methylation and tumor size (P = 0.89): The average greatest dimensions of tumors with and without H-cadherin methylation were 3.78 cm and 3.65 cm, respectively. The correlation also was examined according to the degree of differentiation, because the tumor growth rate generally is correlated with the level of differentiation; however, no significant difference between tumor size and H-cadherin methylation was found in well differentiated, moderately differentiated, or poorly differentiated tumors (data not shown). Moreover, H-cadherin methylation was not associated with cell proliferation, as determined using Ki-67 antigen as a proliferation marker (Figs. 1B, 3). The Ki-67 indexes for 305 NSCLC patients with and without H-cadherin methylation were similar (23% vs. 27%, respectively; P = 0.53). Ki-67 staining patterns differed strikingly according to histology, thus, data were stratified by histologic subtypes for further analysis. Among the patients with adenocarcinoma, the Ki-67 indexes were similar for those with and without H-cadherin methylation (19% vs. 20%, respectively; P = 0.79), and the corresponding Ki-67 indexes in patients with squamous cell carcinoma were 32% and 29% (P = 0.45).
Multivariate Logistic Regression Analysis
Multivariate logistic regression was performed to control for the potential confounding effects of variables, such as age and histology. The coefficient for variable age was not statistically significant, but age was considered a biologically important variable; thus, it was included in the multivariate analysis for further parsimonious model building. In the multivariate analysis (Table 2), the three individual TNM variables were used first instead of the variable “pathologic stage” to understand the possible causes of the lower prevalence of H-cadherin methylation in Stage II than in other stages, as determined in the univariate analysis. Six variables (age, gender, TNM classification, and histology) were considered as the main effect model. On the basis of the main effect model, two-way or three-way interaction was considered. The final model was selected based on deviance. Whether the final model fit the data well was tested by a Hosmer and Lemeshow goodness-of-fit test. No evidence of lack of fitness was observed for the final model (P = 0.34).
|Variables||Odds ratio||95% CI||P valueb|
|Lymph node status|
Women had a greater prevalence of H-cadherin methylation than men in the univariate analysis, and this correlation also was significant in the multivariate analysis (OR = 2.08; 95% CI, 1.10–3.97; P = 0.03). H-cadherin methylation occurred with 2.47 times greater frequency in adenocarcinoma than in squamous cell carcinoma (95% CI, 1.32–5.28; P = 0.01). Patients who had T2 tumors had a 2.71 times increased risk of H-cadherin methylation compared with patients who had T1 tumors (95% CI, 1.21–6.09; P = 0.01) (Table 2). Increased risks of H-cadherin methylation were found in patients with T3 tumors (OR = 3.78; 95% CI, 1.05–13.59; P = 0.04) and T4 tumors (OR = 8.75; 95% CI, 2.14–35.77; P = 0.002) compared with patients who had T1 tumors. H-cadherin methylation had a preventive effect on lymph node metastasis (OR = 0.51; 95% CI, 0.28–0.95; P = 0.03). Patients who had H-cadherin methylation had a 9.83 times greater risk of distant metastasis than patients without H-cadherin methylation (95% CI, 1.18–82.16; P = 0.03). When the four variables (age, gender, disease stage, and tumor histology) were considered together as a main effect model, we obtained similar results. Patients who had Stage II NSCLC had a lower risk of H-cadherin methylation compared with patients who had Stage I NSCLC (OR = 0.41; 95% CI, 0.22–0.75; P = 0.004). However, H-cadherin methylation occurred with 1.71-fold greater prevalence in patients who had Stage III disease (95% CI, 0.87–3.33; P = 0.09) and with 8.89-fold greater prevalence in patients who had Stage IV disease (95% CI, 1.04–75.89; P = 0.03) compared with patients who had Stage I disease.
Data were stratified by disease stage for the survival analysis, because stage is a recognized independent risk factor in patients with NSCLC. Kaplan–Meier survival curves demonstrated that the overall survival of patients with Stage I NSCLC (Fig. 4A) and Stage II NSCLC (Fig. 4B) did not differ significantly for patients with and without H-cadherin methylation. The patients who had unmethylated H-cadherin in Stage I had a slightly better median disease-free survival compared with patients who had methylated H-cadherin (55 mos vs. 40 mos, respectively), but the difference was not statistically significant (P = 0.51). The median survival of patients with and without H-cadherin methylation in Stage II was 35 months and 50 months, respectively (P = 0.46). Age, gender, smoking status, grade of differentiation, pathologic stage, and histology were chosen as covariates for a parsimonious model in a Cox proportional-hazards regression analysis. Age was log transformed, because a log transformation made the assumption of linearity reasonable versus dummy or continuous variables. It was observed that patients who had H-cadherin methylation had a 1.27 times greater hazard than patients who were without H-cadherin methylation, but this difference was not statistically significant (95% CI, 0.76–2.47; P = 0.27). Even when we analyzed the survival data according to TNM classification instead of variable stage, no association was found between H-cadherin methylation and hazard (data not shown), suggesting that H-cadherin methylation may not be an independent prognostic factor.
In an initial step in the study of the clinicopathologic significance of H-cadherin methylation in human NSCLC, we previously studied the methylation status of the H-cadherin gene in bronchial lavage samples from disease-free individuals and from small numbers of patients with NSCLC (n = 85 patients). We found that H-cadherin methylation occurred in 4 of 127 bronchial lavage specimens (3%) from the disease-free individuals studied.18 The mechanism of H-cadherin methylation in disease-free individuals is not clear, but it may reflect a latent period before clinical tumor detection or a high-risk status, which suggests that H-cadherin methylation may occur at the early stage of lung carcinogenesis. To understand the role of H-cadherin methylation further as an early detection marker, we studied H-cadherin methylation in a larger population (n = 305 patients). In the current study, H-cadherin methylation occurred in 81 of 183 patients (44%) with Stage I disease, suggesting that H-cadherin methylation may be a candidate molecular marker for the early detection of NSCLC.
The prevalence of H-cadherin methylation, as determined in the current study, is consistent with previous findings by other groups.9, 10, 12, 15 It was observed that the prevalence of H-cadherin methylation was increased significantly in patients with NSCLC in an advanced pathologic stage, suggesting that H-cadherin is associated with tumor progression. It is noteworthy that a preventive effect was observed for H-cadherin methylation on lymph node metastasis (OR = 0.51; 95% CI, 0.28–0.95; P = 0.03), suggesting that unmethylated H-cadherin may confer a selective advantage to lymph node metastasis. However, this finding is not consistent with the results of Takeuchi and Ohtsuki,6 who reported that the loss of H-cadherin in primary lung tumors was not associated with the presence of lymph node metastasis. Unlike patients with lymph node metastasis, most patients with distant metastasis in the current study showed aberrant methylation of the H-cadherin gene, suggesting that H-cadherin may be a negative regulator of hematogenous metastasis. Metastasis may depend in part on how different the primary tumors are from surrounding tissue and how close the site of a primary tumor is to lymph or blood vessels. In addition, the activation states of certain genes are an important consideration in terms of invasive tumor cells gaining access to a blood or lymph vessel. E-cadherin expression in NSCLC was common among those patients without lymph node metastasis (N0) (63%) and then declined from 46% in patients with N1 status to 8% in patients with N2 status, suggesting that the reduced expression of E-cadherin is correlated significantly with increased lymphogenous metastasis.19 On the basis of these observations, it seems likely that hematogenous and lymphogenous metastasis may be regulated differently by the cadherin family members and that H-cadherin may be involved in tumor metastasis in a complex way.
Although there is a clear distinction between metastasis-promoting and growth-transforming genes, there is increasing evidence that some genes, such as integrins, mediate signals that affect both processes. Several groups have suggested the involvement of H-cadherin in the regulation of tumor growth and progression. It has been reported that the introduction and overexpression of H-cadherin in human breast carcinoma cells (MDAMB435) markedly inhibit tumor growth and invasiveness.7, 8 Zhong et al.11 also reported that the loss of H-cadherin expression is associated with tumorigenicity in nude mice transplanted with NSCLC tumors and that it is more prevalent in larger local tumors (5 of 76 T1 tumors [29%] vs. 9 of 15 T2 tumors [60%]; P = 0.15). The mechanism by which H-cadherin functions as cell growth regulator is not clear; however, Huang et al.20 recently reported that T-cadherin (also designated H-cadherin) regulates cell growth by inducing p21CIP1/WAF1 expression and G2-phase arrest. T-cadherin overexpression results in the suppression of C6 glioma cell growth by inducing G2-phase arrest, and the growth arrest mediated by T-cadherin is associated with p21CIP1/WAF1 expression but not with p27Kip1 expression. Zhong et al.21 also reported that H-cadherin is involved in contact inhibition by inducing p21CIP1/WAF1 expression in Chinese hamster ovarian cells. These observations suggest that H-cadherin may be involved in two processes, i.e., tumor growth and progression. However, in the current study, the prevalence of H-cadherin methylation was not associated with tumor size. H-cadherin methylation occurred in 38% of patients who had tumors that measured < 1.5 cm in greatest dimension, in 42% of patients who had tumors that measured < 3.0 cm and ≥ 1.5 cm, in 37% of patients who had tumors that measured < 4.5 cm and ≥ 3.0 cm, and in 43% of patients who had tumors that measured ≥ 4.5 cm, suggesting that H-cadherin may not be involved in tumor growth regulation in NSCLC (P = 0.89). The average dimensions of tumors with and without H-cadherin methylation also were similar (3.78 cm vs. 3.65 cm, respectively). In addition, the Ki-67 proliferation index did not show a correlation with H-cadherin methylation in adenocarcinoma or squamous cell carcinoma. Accordingly, further study is needed to investigate the role of H-cadherin in cell proliferation and tumor growth.
No association was found between patient survival and H-cadherin methylation, consistent with the findings of others.22, 23 It has been established that reduced E-cadherin expression is associated with an unfavorable prognosis in NSCLC: Patients with negative E-cadherin expression had a hazard ratio of 2.41 compared with patients who showed strong E-cadherin expression.19 Given these observations, E-cadherin and H-cadherin may have different effects on the prognosis of patients with NSCLC. The current study was limited severely by a short follow-up of patients and by the small numbers of patients, especially patients with Stage IV disease (n = 8 patients). Large-scale testing is warranted to investigate further the role of H-cadherin in lymphogeneous and hematogenous metastasis. In conclusion, the current results suggest that H-cadherin methylation is associated with disease progression in NSCLC but not with patient survival.
The authors thank Jin-Hyuk Kim for his critical reading of the article, Eunkyung Kim for assistance with data collection and management, and Hoon Suh for sample collection.