Despite the role of DNMT3B in de novo DNA methylation, a correlation between DNMT3B expression and promoter DNA methylation has not being established in tumors. We recently reported ΔDNMT3B, a subfamily of DNMT3B, with seven variants, as the predominant expression forms in non–small cell lung cancer (NSCLC). We hypothesized that expression of the ΔDNMT3B variants plays a role in promoter methylation formation during lung tumorigenesis. Expression of seven ΔDNMT3B variants was measured in 119 NSCLCs and the corresponding normal lungs using reverse transcription-PCR. The expression patterns of ΔDNMT3B variants were analyzed with the status of p16 and RASSF1A promoter methylation in the tumors as well as in patients' clinical variables, including outcomes. Expression of ΔDNMT3B variants was detected in 94 of 119 (80%) tumors but in only 22 (18%) of the corresponding normal lungs (P < 0.0001). ΔDNMT3B1, ΔDNMT3B2, and ΔDNMT3B4 were the most frequently detected transcripts in the tumors (62%, 76%, and 46%, respectively). The expression of ΔDNMT3B variants was associated with p16 and RASSF1A promoter methylation in the tumors, but the strongest association was between ΔDNMT3B4 and RASSF1A. Forty-two of 46 (91%) tumors with RASSF1A promoter methylation expressed ΔDNMT3B4 compared with only 13 of 73 (18%) tumors without the promoter methylation (P < 0.0001). Strong associations were also observed between expression of the variants in the tumors and in patients' clinical outcomes. Expression of ΔDNMT3B variants is common in NSCLC and may play an important role in the development of promoter methylation. (Cancer Res 2006; 66(17): 8361-6)
- promoter methylation
- alternative splicing
DNA methylation plays an important role in regulation of gene expression, genomic imprinting, and X-chromosome inactivation ( 1– 5). During tumorigenesis, promoter hypermethylation is a major mechanism to inactivate tumor suppressor genes with CpG-rich promoters ( 6). In non–small cell lung cancer (NSCLC), promoter methylation has been frequently detected in several tumor suppressor genes, such as p16 and RASSF1A ( 7– 10). Although some tumors tend to have more promoters methylated than others, each tumor has distinct profiles of methylated promoters, suggesting a complex mechanism in controlling the de novo methylation process. Understanding the process is important to develop strategies for lung cancer prevention, molecular classification, and targeted therapy.
DNMT3B, an important member of DNMT3 family, is a de novo methyltransferase, which adds the first methyl group to the cytosine of unmethylated DNA ( 1, 11, 12). It has been shown that DNMT3B is overexpressed in transformed cells and in multiple types of primary tumors, including NSCLC ( 13– 15). However, the correlation between the expression levels of DNMT3B and promoter methylation status in human cancers has been weak ( 15– 17), suggesting that other key factors are involved in regulation of the promoter methylation. We recently identified a new subfamily of DNMT3B, termed ΔDNMT3B, due to its lack of part of the NH2-terminal sequence ( 18). We showed that ΔDNMT3B is the major transcript of DNMT3B in NSCLC and has at least seven transcriptional variants resulting from alternative splicing ( 18). The purpose of the study was to analyze the expression patterns of ΔDNMT3B variants in primary NSCLC and to determine their potential relationship with methylation status of tumor suppressor genes commonly altered in NSCLC. To determine biological significance of the molecules, the relationship between expression of the ΔDNMT3B variants and patients' clinical outcomes was also analyzed.
Materials and Methods
Patients and specimens. One hundred and nineteen primary tumor samples and their corresponding nonmalignant lung tissues were obtained from patients with stages I to IIIa NSCLC (all stages are pathology stages). All the patients were treated by surgery with curative intend, except those with stage IIIa tumors who might also received postoperative radiation therapy and adjuvant chemotherapy in M. D. Anderson Cancer Center from 1995 to 2000. Samples were immediately frozen and stored at −80°C until analysis. The selection of these patients was based on the availability of archived fresh tumor and corresponding normal lung tissues for the investigators. The clinical information and follow-up data were based on chart review and reports from tumor registry service. Informed consent for the use of residual resected tissues for research was obtained from all the patients enrolled in the study. The study was reviewed and approved by the institution's Surveillance Committee to use the tissues and clinical information. The patients ranged in age from 32 to 84 years (median, 64 years). There were 40 patients with stage I disease, 30 stage II, and 49 stage IIIa. Histologic subtypes include 60 adenocarcinomas, 49 squamous cell carcinoma (SCC), and 7 large cell carcinoma ( Table 1 ). None of the patients had received chemotherapy or radiation treatment before surgery. The median follow-up time was 50.96 months.
RNA extraction and reverse transcription-PCR. Total RNA from tissue samples was extracted by using Tri-Reagent according to the manufacturer's instruction. Approximately 1 to 2 μg of total RNA from each sample were used to conduct reverse transcription reaction in a 20 μL volume by using SuperScript II RNase H-reverse transcriptase (Life Technologies, Inc., Gaithersburg, MD). The PCR was carried out in a 12.5 μL volume containing 0.5 μL reverse transcription product, 1.5 mmol/L deoxynucleotide triphosphate (dNTP), 7% DMSO, 6.7 μmol/L MgCl2, 16.6 mmol/L (NH4)SO4, 67 mmol/L Tris, 10 mmol/L B-mercaptoethanol, 6.7 μmol/L EDTA, 0.5 μmol/L of both the sense and antisense primers, and 0.625 unit of HotStar Taq DNA polymerase (Life Technologies). Amplification was done with an initial denaturing step at 95°C for 15 minutes followed by 40 cycles of 95°C for 30 seconds, 58°C for 1 minute, and 72°C for 1 minute in a thermal cycler with a last extension step of 72°C for 10 minutes. PCR products were separated on 2.5% agarose gels and visualized after staining with ethidium bromide. The expression of specific ΔDNMT3B variants was determined by using specific primer sets corresponding to unique sequences of individual ΔDNMT3B variants as reported previously ( 18). Any expression observed for a given variant was considered expression positive for both normal and tumor tissues.
DNA extraction and methylation-specific PCR. Frozen tissues were homogenized, and genomic DNA was extracted by digestion of homogenized tissues in buffer containing 50 mmol/L Tris-HCl (pH 8.0), 1% SDS, and 0.5 mg/mL proteinase K at 42°C for 36 hours. The digested products were purified with phenol-chloroform twice. DNA was then precipitated using the ethanol precipitation method and recovered in distilled DNase-free water. For methylation-specific PCR, 1 μg of genomic DNA from each tissue sample was used in the initial step of chemical modification. Briefly, DNA was denatured by NaOH and treated with sodium bisulfite (Sigma Chemical Co., St. Louis, MO). After purification with the use of Wizard DNA purification resin (Promega Corp., Madison, WI), the DNA was treated again with NaOH. After precipitation, DNA was recovered in water and was ready for PCR with the use of specific primers for either the methylated or the unmethylated p16 or RASSF1A promoter as reported previously ( 10). PCR was carried out in 25 μL containing ∼100 ng of modified DNA, 3% DMSO, all four dNTPs (each at 200 μmol/L), 1.5 mmol/L MgCl2, 0.4 μmol/L PCR primers, and 1.25 units of Taq DNA polymerase (Life Technologies). DNA was amplified for 35 cycles at 95°C for 30 seconds, 60°C for 60 seconds, and 70°C for 60 seconds followed by a 5-minute extension at 70°C in a temperature cycler (Hybaid; Omnigene, Woodbridge, NJ) in 500 μL plastic tubes. PCR products were separated on 2.5% agarose gels and visualized after staining with ethidium bromide. For each DNA sample, primer sets for methylated DNA and unmethylated DNA were used for analysis. CpGenome universal methylated DNA (Chemicon International, Temecula, CA) was used as positive controls, and water replacing for DNA was used as blank controls. The hypermethylation status was determined by visualizing a 150-bp PCR product for p16 and a 169-bp PCR product for RASSF1A with the respective methylation-specific primer sets.
Statistical analysis. The χ2 test or Fisher's exact test was used to test the association between categorical variables. Overall survival, disease-specific survival (i.e., survival rates among people who died of lung cancer–related causes specifically), and disease-free survival (i.e., recurrence, metastasis, or cancer death is considered as an event) were analyzed. Survival probability was estimated using the Kaplan-Meier method. The log-rank test was used to compare patients' survival time between or among groups. Cox regression was used to model the risks of biological variables on survival time, with adjustment for clinical and histopathologic variables (age, sex, tumor histology subgroup, tumor size, smoking status, and adjuvant treatment). All statistical tests are two sided, and P < 0.05 was considered statistically significant.
Expression of ΔDNMT3B variants in primary NSCLC and the corresponding normal lungs. Expression of ΔDNMT3B variants was detected in 94 of 119 (80%) NSCLCs but in only 22 (20%) of the corresponding normal lungs. The difference of the expression rates between the tumor tissues and normal tissues was statistically significant (P < 0.0001). In the tumor tissues, the most frequently expressed variant was ΔDNMT3B2 (91 or 76%) followed by ΔDNMT3B1 (74 or 62%) and ΔDNMT3B4 (55 or 46%) in the 119 primary NSCLCs. Expression of the other variants was less frequent; 32 (27%), 21 (18%), 19 (16%), and 3 (2.5%) for ΔDNMT3B6, ΔDNMT3B5, ΔDNMT3B7, and ΔDNMT3B3, respectively. Expression of the later group of ΔDNMT3B variants was not detected in any of the normal lung tissues.
Correlation between expression of ΔDNMT3B variants and clinicopathologic variables. We analyzed correlations between the expression of ΔDNMT3B variants and clinicopathologic variables of the patients ( Table 1). Tumors from female patients had a higher frequency of ΔDNMT3B4 expression than tumors from male patients (P = 0.03). Expression of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 was more frequent in poorly differentiated tumors than in well-differentiated or moderately differentiated tumors (P < 0.05). The expression of ΔDNMT3B7 was more frequent in stage III tumors than in stage I/II tumors (P = 0.009). No other correlation was observed ( Table 1).
Correlation between expression of ΔDNMT3B variants and promoter methylation of p16 and RASSF1A genes. Promoter methylation of p16 and RASSF1A was detected in 58 (49%) and 46 (39%) of 119 tumors, respectively. Among the 94 tumors that expressed any of the ΔDNMT3B variants, 54 (57%) had p16 promoter methylation and 46 (48%) had RASSF1A promoter methylation compared with 4 of 25 (16%) tumors without expression of any of the variants for p16 and none for RASSF1A (P < 0.0001). We then analyzed the relationship between methylation status of p16 and RASSF1A and expression of individual ΔDNMT3B variants in the tumor tissues. Promoter methylation of p16 was correlated with expression of ΔDNMT3B1, ΔDNMT3B2, ΔDNMT3B5, and ΔDNMT3B6, whereas promoter methylation of RASSF1A was correlated with all the ΔDNMT3B variants, except DNMT3B3 ( Table 2 ). The most striking correlation was between expression of ΔDNMT3B4 and promoter methylation of RASSF1A. Among the 46 tumors with promoter methylation of RASSF1A, 42 (91%) expressed ΔDNMT3B4 compared with only 13 of 73 (18%) tumors without promoter methylation of RASSF1A (P < 0.0001). In contrast, expression of ΔDNMT3B4 was not correlated with promoter methylation of p16 (P = 0.12).
Interestingly, promoter methylation of p16 was associated with promoter methylation of RASSF1A (P = 0.004). To determine independent factors correlated with promoter methylation, we did multivariate analysis. Results showed that promoter methylation of p16 and expression of ΔDNMT3B4 or ΔDNMT3B7 were independent factors for promoter methylation of RASSF1A. After adjusting for ΔDNMT3B7 expression and promoter methylation of p16 using logistic regression analysis, expression of DNMT3B4 remained its strong correlation with promoter methylation of RASSF1A (P < 0.0001), suggesting that ΔDNMT3B4 is required for promoter methylation of RASSF1A in NSCLC.
Correlation between expression of ΔDNMT3B variants and clinical outcomes. We analyzed a potential correlation between expression of ΔDNMT3B variants and patients' clinical outcomes. Because many of the patients with stage III tumors underwent postsurgery radiation therapy or chemoradiation therapy whereas patients with stage I/II tumors did not, we analyzed the two groups separately. For stage I/II tumors, patients whose tumors expressed any of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 had statistically significant poorer overall, disease-specific, and disease-free survivals than patients whose tumors had no expression of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 (P = 0.002, P < 0.001, and P < 0.001, respectively; Fig. 1 ). Because p16 promoter methylation was statistically correlated with the survivals in the patient population ( 10), we did multivariate analysis to correct the confounding factor. Both expression of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 and promoter methylation of p16 were independent factors in predicting poorer clinical outcomes (P < 0.01 for both overall and disease-specific survivals). For stage III tumors, patients whose tumors expressed any of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 variants had poorer survivals (P < 0.001; Fig. 2A-C ) similar to the observation in patients with stage I/II tumors. The difference in this group of patients was that expression of ΔDNMT3B4 was also strongly correlated with poorer survivals (P < 0.0001, P < 0.0001, and P < 0.001, for overall, disease-specific, and disease-free survivals, respectively; Fig. 2D-F). In our previous study, we found that both p16 and RASSF1A promoter methylations were correlated with survivals in the patient population ( 10). However, in the multivariate analysis, only ΔDNMT3B4 expression and p16 promoter methylation were independent factors (P < 0.001 for both overall and disease-specific survivals).
Overexpression of DNMT3B, but not DNMT1 and DNMT3A, has been found common in multiple cancer types, including lung cancer ( 13, 14, 17), suggesting that DNMT3B plays an important role in the development of aberrant promoter methylation during tumorigenesis. However, the correlation between expression levels of DNMT3B and promoter methylation status was not strong in human tumors in most reports ( 15– 17). Several studies have suggested that DNMT3B alone has limited effect in promoter methylation because the maintenance of methylated promoters of tumor suppressor genes could only be effectively disrupted when both DNMT3B and DNMT1 genes were knocked out, whereas the single knockout of either DNMT3B or DNMT1 had minimal effects ( 19– 21). However, these studies did not address potential effects of individual variants of DNMT3B. A dominant-negative effect of DNMT3b4 by competing with DNMT3b3 has been suggested, which resulted in DNA hypomethylation on pericentromeric satellite regions ( 14). This result suggests a complex role of DNMT3B variants in regulation of DNA methylation formation. The identification of ΔDNMT3Bs as the predominant expressing forms of DNMT3B in lung cancer ( 18) further exemplified the complexity of the regulation in lung tumorigenesis.
In this study, we provided a comprehensive view of the expression profiles of the seven ΔDNMT3B variants in a large panel of NSCLC tumors and their corresponding normal lungs. The fact that ΔDNMT3B variants are frequently expressed in the primary NSCLC but less frequently in the corresponding normal lungs underscores the importance of these molecules in lung tumorigenesis. More importantly, our study provides first in vivo evidence to support the importance of individual ΔDNMT3B variants in the development of promoter methylation of a particular gene in lung tumorigenesis. We conclude that expression of ΔDNMT3B4 may contribute to the development of RASSF1A promoter methylation. We found that 91% of the NSCLC tumors with methylated RASSF1A promoter expressed the variant compared with only 18% of the tumors without RASSF1A promoter methylation (P < 0.0001). As a comparison, 66% of the tumors without RASSF1A promoter methylation expressed ΔDNMT3B2, although 94% of the tumors with methylated RASSF1A promoter expressed ΔDNMT3B2 ( Table 2). Interestingly, all four tumors with methylated RASSF1A but no expression of ΔDNMT3B4 expressed ΔDNMT3B2, suggesting a role of ΔDNMT3B2 in RASSF1A promoter methylation in some NSCLC tumors. To support this notion, none of the 25 tumors that expressed neither ΔDNMT3B4 nor ΔDNMT3B2 had methylated RASSF1A promoter, whereas 6 of 25 (24%) tumors carried methylated p16 promoter (P = 0.02). Because only two promoters were analyzed in this study, whether ΔDNMT3B4 is involved in promoter methylation of other genes remains to be determined.
Although the mechanism of differential promoter methylation in tumors is unclear, it is possible that the expression of DNMT variants or the relative expression levels of the variants, including those derived from DNMT3A and DNMT3B, may, at least in part, determine patterns of the promoter methylation. One possible explanation of the regulation might be a differential DNA binding through variable PWWP structure, which locates at the more proximal part of DNMT3Bs and is capable to bind DNA directly ( 22). It is also possible that different variants might have different protein-protein binding capability ( 23), resulting in different modification of chromatin structures. In either case, different promoters with distinct DNA structures, protein complexes, or protein modifications might be preferentially targeted by individual DNMT variants, resulting in a selected promoter methylation. It should be noted, however, that we only analyzed the ΔDNMT3B variants based on the alternative splicing of more proximal exons; therefore, whether the catalytic domains of ΔDNMT3B4/ΔDNMT3B2 play a role in the process is not addressed in this study.
The association between the expression of certain ΔDNMT3B variants and clinical outcomes provides further support for the role of ΔDNMT3Bs in NSCLC. In patients with either early-stage or locally advanced-stage tumors, the expression of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 correlated with poorer survivals, which was independent of other clinicopathologic variables and promoter methylation status of p16 and RASSF1A. Although the frequencies of ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 expression were relatively low in NSCLC ( Table 2), the expression was never detected in the nonmalignant lung tissues, suggesting that ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 are involved in the late-stage tumorigenesis of lungs. It is interesting to note that the predicted proteins from ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7 contain no enzymatic domain of the methyltransferase due to premature translational termination ( 18). Unfortunately, the lack of high-quality antibodies for the variants prevents us to evaluate protein expression of the variants in the tumors at this time. Whether their biological roles are inserted through their variable COOH-terminal regions remains to be determined.
In summary, we did the first comprehensive analysis to determine the expression profiles of ΔDNMT3B variants in large number of primary NSCLC tumors and their corresponding normal lung tissues. We revealed a strong correlation between the expression of ΔDNMT3B4 and RASSF1A promoter methylation, suggesting a role of the variant in regulation of promoter methylation during lung tumorigenesis. We also found that expression of certain ΔDNMT3B variants (i.e., ΔDNMT3B5, ΔDNMT3B6, and ΔDNMT3B7) was correlated with poor clinical outcomes, suggesting their role in NSCLC progression.
Grant support: Department of Defense grant DAMD17-01-1-01689-1.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received June 2, 2006.
- Revision received July 12, 2006.
- Accepted July 20, 2006.
- ©2006 American Association for Cancer Research.