Synchronous Alterations of Wnt and Epidermal Growth Factor Receptor Signaling Pathways through Aberrant Methylation and Mutation in Non–Small Cell Lung Cancer

The Wingless-type (Wnt) signaling cascade is a major regulator of embryonic development. Activation of the Wnt signal transduction pathway results in an increase in the cytoplasmic pool of β-catenin, which is subsequently translocated to the cell nucleus where it interacts with members of the T-cell factor/lymphocyte enhancer factor family. These downstream gene targets include developmental regulators and other genes involved in cell proliferation and cancer progression (1, 2). Aberrant activation of the Wnt signal transduction pathway has been closely linked to tumorigenesis in a variety of human tumors (3).

Two functional classes of extracellular Wnt antagonists have been identified (4). The first class includes the secreted frizzled-related protein (sFRP) family (sFRP1-sFRP5) and Wnt inhibitory factor-1 (Wif-1). These proteins inhibit Wnt signaling by directly binding to Wnt molecules (4). The second class is the Dickkopf (Dkk) family of proteins (Dkk1-Dkk4) that inhibit Wnt signaling by binding to the LRP5/LRP6 component of the Wnt receptor complex. These antagonists control Wnt signaling in normal conditions. Down-regulation of Wnt antagonists, such as sFRP and Dkk, through aberrant methylation results in anomalous Wnt signaling and has been identified in several human malignancies, including non–small cell lung cancer (NSCLC; refs. 5–10). Although aberrant methylation of a Wnt antagonist gene (sFRP-1 or Wif-1) has been reported in lung cancer (6, 8, 11, 12), the methylation status of these genes in NSCLC has not been comprehensively analyzed. In addition, a detailed mechanism for β-catenin transport and accumulation in the nucleus as a late step of Wnt signaling has not yet been elucidated for many types of cancer (13, 14).

Dysregulation of the epidermal growth factor receptor (EGFR) signaling pathway has also been implicated in a variety of cancers. NSCLCs bearing an EGFR mutation have generated considerable interest because such mutations are associated with an increased sensitivity to gefitinib, an EGFR inhibitor (15, 16). We have also shown the clinical importance of EGFR and KRAS mutations in resected NSCLCs (17, 18).

Primary NSCLCs were first examined for alteration of the Wnt signaling pathway, including aberrant methylation of Wnt antagonists and abnormal expression of β-catenin. Next, because these two pathways have critical roles in the pathogenesis of NSCLC through cellular proliferation and transformation, the results of aberrant Wnt signaling were correlated with mutations in EGFR signaling.

Patients. Surgically resected samples were obtained from 238 unselected patients with NSCLC at the Chiba University Hospital (Chiba, Japan) from 1995 to 2000. This study was approved by the Institutional Review Board and written informed consent was obtained from all participants. All patients received curative intent surgery, but none had received any treatment before resection. Resected samples were immediately frozen and stored at −80°C until used. Each sample was used for methylation and mutation assays, whereas 91 of the 238 cases were analyzed by immunohistochemistry.

DNA extraction and methylation-specific PCR. Genomic DNA was obtained from primary tumors and nonmalignant tissues by digestion with proteinase K (Life Technologies, Inc.) followed by phenol/chloroform (1:1) extraction (19). The DNA was treated with sodium bisulfite as described previously (20). PCR amplification of sFRP-1, sFRP-2, sFRP-5, Wif-1, and Dkk-3 gene targets was done using bisulfite-treated DNA as the template and specific primer sequences for the methylated and unmethylated forms of the genes (5). DNA methylation patterns in the CpG island of these genes were determined with methylation-specific PCR as reported by Herman et al. (21). Bisulfite-treated universal methylated DNA (Chemicon) was used as a positive control for the methylated alleles. DNAs from lymphocytes (n = 14) of healthy nonsmoking volunteers were used as negative controls for methylation-specific assays. Nine microliters of each PCR product were loaded on 2% agarose gels stained with ethidium bromide. Results were confirmed by repeating the bisulfite treatment and methylation-specific PCR for all samples.

Mutation assay. Sequences of the first four exons (18–21) of the EGFR tyrosine kinase domains and exon 2 of KRAS were analyzed (17). All PCR products were incubated with exonuclease I and shrimp alkaline phosphatase (Amersham Biosciences Corp.) and sequenced directly using a dye terminator cycle sequencing kit from Applied Biosystems (Perkin-Elmer Corp.). All sequence variants were confirmed by independent PCR amplifications and sequenced in both directions.

Immunohistochemistry. Immunostaining was done on 5-μm-thick sections using β-catenin mouse monoclonal antibody (clone 14; BD Transduction Laboratories) at a 1:400 dilution for 12 h. The slides were prepared with antigen retrieval using citrate buffer [10 mmol/L (pH 6.0)] before incubation with primary antibody. Nonimmune serum was used instead of the primary antibody in negative controls. 3,3′-Diaminobenzidine (Sigma-Aldrich) was used as the chromogen, with hematoxylin as the counterstain. β-Catenin expression was detected at the plasma membrane of normal bronchial epithelial cells, gland cells, and pneumocytes, which served as internal controls. Cytoplasmic and/or nuclear staining was regarded as positive. The positive cells were counted and divided into three categories (<25%, 25-75%, or >75%). In addition, the mean staining intensity in each specimen was categorized as 1, 2, or 3 (weak, moderate, or strong). The total immunostaining score was then divided into low, moderate, and high scores. This scoring method has been widely used to evaluate the results of immunohistochemical staining for β-catenin (10, 22, 23).

Statistical analysis. The Fisher's exact test and Mann-Whitney U test were applied to assess the association between categorical variables. Overall survival curves were calculated with the Kaplan-Meier method and compared by the log-rank test. The Cox proportional hazards regression model was used for multivariate analyses.

The overall extent of methylation for the panel of genes was examined by comparing the methylation index (MI) for each case [MI = (total number of methylated genes) / (total number of analyzed genes)] and determining the mean for the different groups. Two MI values were calculated. The first was the MI for all genes examined and the second was the MI for three sFRP genes (sFRP-1, sFRP-2, and sFRP-5). The cases were also divided based on the presence or absence of sFRP gene methylation and the presence or absence of any other gene methylation (9). Statistical significance was defined as a P value of <0.05. All P values were two tailed.

Methylation profile of Wnt antagonists in NSCLC. Examples of the bands obtained by methylation-specific PCR are illustrated in Fig. 1, and detailed data on the frequency of aberrant methylation are summarized in Table 1. The number of aberrant methylation events in malignant lung tissue was 81 (34%) for sFRP-1, 123 (52%) for sFRP-2, 78 (33%) for sFRP-5, 66 (28%) for Wif-1, and 32 (13%) for Dkk-3. In addition, 175 nonmalignant lung tissues corresponding to these tumors were also examined. The number of aberrant methylations in nonmalignant lung tissue was 6 (3%) for sFRP-1, 13 (7%) for sFRP-2, 14 (8%) for sFRP-5, 2 (1%) for Wif-1, and 2 (1%) for Dkk-3. Methylation of all five genes was a tumor-specific event (P < 0.0001) when compared with corresponding nonmalignant tissues.

The results of these methylation tests were then correlated with clinical factors. The frequency of sFRP-2 methylation was higher in females (P = 0.0067), nonsmokers (P = 0.0038), and adenocarcinoma cases (P = 0.0015). The frequency of any sFRP gene methylation was higher in females (P = 0.0322) and nonsmokers (P = 0.0299), whereas the frequency of any other gene methylation was higher in females (P = 0.0255). The MI for sFRP genes was 1.40 ± 1.01 (mean ± SD) in nonsmokers (n = 67) and 1.11 ± 1.04 in smokers (n = 171). The sFRP MI was significantly higher in nonsmokers than in smokers (P = 0.036).

Dkk-3 methylation was significantly associated with a poor prognosis in adenocarcinoma cases as estimated using the log-rank test (P = 0.0092; Fig. 2A). Cox proportional hazards regression analysis was done to determine whether Dkk-3 methylation is an independent prognostic factor (Table 2). Adenocarcinoma cases with Dkk-3 methylation had a significantly poorer prognosis than those without Dkk-3 methylation (hazard ratio, 3.72; P = 0.0005).

There were 65, 65, 50, 26, 21, and 11cases with MIs of 0, 0.2, 0.4, 0.6, 0.8, and 1, respectively. Methylation frequency for all genes was correlated with MIs, excluding that of the gene currently under examination (all genes, P < 0.0001). These findings imply that methylation and silencing of the Wnt pathway tend to occur frequently and synchronously.

Mutation profile of EGFR signaling in NSCLC. Among the 238 patients, 60 carried EGFR mutations and 15 had KRAS mutations. Details of these mutations are shown in Table 1. EGFR mutations were more prevalent in females (P < 0.0001), nonsmokers (P < 0.0001), and adenocarcinoma cases (P < 0.0001). KRAS mutations were present in 1 squamous cell carcinoma and in 14 adenocarcinomas (P = 0.006). EGFR and KRAS mutations were mutually exclusive.

Protein expression of β-catenin. Typical immunostaining patterns for β-catenin in NSCLC are shown in Fig. 3. Using the criteria outlined in Materials and Methods, high, moderate, and low expression scores were found in 20 (22%), 51 (56%), and 20 (22%) tumors. More than 20% of nuclei were positive in all high-score cases. This high-score group is important because the nuclear entry of β-catenin is a key step in gene activation by the β-catenin/T-cell factor/lymphocyte enhancer factor transcriptional complex (22). However, β-catenin expression was not correlated with various clinicopathologic factors, including prognosis.

Novel relationships between the Wnt and EGFR signaling pathways. The Wnt and EGFR signaling pathways (Fig. 4) were examined for relationships among genetic/epigenetic factors and protein expression. Tumors with high β-catenin expression were frequently present in tissues with EGFR mutants (12 of 23 versus 8 of 68; P = 0.0002). KRAS mutations were frequently present in tumors with aberrantly methylated sFRP-1 (10 of 81 versus 5 of 157; P = 0.0096), sFRP-2 (12 of 123 versus 3 of 115; P = 0.031), sFRP-5 (9 of 78 versus 6 of 160; P = 0.042), and Dkk-3 (5 of 32 versus 10 of 206; P = 0.036). KRAS mutations were also prevalent in tumors with any methylated sFRP gene (15 of 162 versus 0 of 76; P = 0.0033) and any other methylated gene (15 of 173 versus 0 of 65; P = 0.0132). The MI for sFRP genes was significantly higher (2.07 ± 1.03) in cases with a KRAS mutation than in those without (1.13 ± 1.02; P = 0.0009). The MI for all genes was also significantly higher (2.67 ± 1.35; n = 15) in cases with a KRAS mutation than in the absence of the mutation (1.53 ± 1.43; n = 223; P = 0.003).

The effects of the methylation status of the sFRP and Wnt antagonist genes on the association between EGFR mutation and overall patient survival were evaluated. In tumors with any sFRP or Wnt antagonist gene methylated, or in tumors having no sFRP gene methylated, the EGFR mutation had no effect on patient survival. However, in patients with tumors that lacked methylated Wnt antagonists, the EGFR mutation was significantly associated with a good prognosis, as estimated using the log-rank test (P = 0.0287; Fig. 2B) and Cox proportional hazards regression analysis (hazard ratio, 0.27; P = 0.041; Table 2).

Cross-talk between Wnt and EGFR has been identified in some tumors (24–28). There are two interpretations of the correlation of the EGFR and Wnt pathways. EGF and EGFR are able to induce β-catenin signaling in human epidermoid carcinoma cells A431 (24). On the other hand, Wnt transactivates EGFR in other tumors (25, 26). In our data, there was a significant positive correlation between activated EGFR mutations and nuclear accumulation of β-catenin in primary NSCLCs. In addition, there was no correlation between EGFR protein expression and nuclear accumulation of β-catenin in this series.⁴

We found that EGFR mutations were significantly associated with a good prognosis in patients that had tumors with unmethylated Wnt antagonist genes. The data about the prognostic role of EGFR mutations are inconsistent and confusing, with some reports indicating that EGFR mutations are associated with good survival, whereas no prognostic association was seen in other reports. In this study, patient prognosis with unmethylated Wnt antagonists was split according to the particular EGFR mutation. NSCLCs with an EGFR mutation but unmethylated Wnt antagonists may have a smaller number of epigenetic changes and exhibit fewer invasive behaviors.

This study also revealed a significant positive correlation between KRAS gene mutations and aberrant methylation of Wnt antagonists in primary NSCLCs. We showed significant association in three ways: via analysis of the methylation of single genes, any or no gene methylation, and calculation of the MI. In previous reports, KRAS seemed to up-regulate vascular endothelial growth factor through the Wnt pathway in a phosphatidylinositol 3-kinase–dependent manner in colon cancer cells (27). In addition, activated KRAS is known to induce tyrosine phosphorylation of β-catenin, leading to its release from E-cadherin at the adherens junction in intestinal tumor cells (28). In contrast to these mechanisms, our data showed that there may be an interaction between KRAS mutations and aberrant methylation of Wnt antagonists in NSCLCs, although how this synchronous alteration functions is unclear.

Recently, a molecular targeting approach for various types of cancer has been widely used to maximize therapeutic benefits and minimize undesirable side effects. Treatment strategy is based on the specific alteration, such as overexpression, mutation, or amplification. Gefitinib, an EGFR tyrosine kinase inhibitor, is used in NSCLC. Because many NSCLCs showed EGFR overexpression (18), this drug was first developed as an EGFR inhibitor. However, EGFR overexpression was not a good predictive marker for response to gefitinib (29). Cases with an EGFR mutation or EGFR amplification are now known to be good candidates (15, 16, 30), but the exact selective sensitivity of NSCLC to this drug is still unknown. A similar situation for better predictive markers in molecular targeting therapy is also observed in other tumors (31). We found two novel synchronous alterations between the Wnt and EGFR signaling pathways that may have some effect on chemosensitivity of molecular targeting agents, such as EGFR inhibitors. Further studies will be needed to establish a correlation between these alterations and chemosensitivity.

In conclusion, we have revealed that (a) aberrant methylation of Wnt antagonist genes is common in NSCLCs; (b) females, nonsmokers, and adenocarcinomas have higher levels of sFRP-2 methylation; (c) Dkk-3 methylation is significantly associated with a poor prognosis in lung adenocarcinomas; (d) there is a positive correlation between activated EGFR mutations and nuclear accumulation of β-catenin; (e) KRAS mutations and aberrant methylation of Wnt antagonists are positively correlated; and (f) EGFR mutations are significantly associated with a good prognosis in tumors having unmethylated Wnt antagonist genes. The results of the methylation and mutation analyses will contribute to a better understanding of the cross-talk between the Wnt and EGFR signaling pathways and may help foster development of chemotherapeutic treatments in NSCLC.