Promoter Hypermethylation of Hallmark Cancer Genes in Atypical Adenomatous Hyperplasia of the Lung

Lung cancer persists as the leading cause of death among men and women in developed countries (1). Notwithstanding improvements in therapy and diagnostic techniques, dismal survival rates have not changed over the span of 10 years (1, 2). Optimism over the decreasing incidence of squamous cell carcinomas over the last two decades has been offset by an increasing incidence of adenocarcinomas. Indeed, adenocarcinoma has now become the most common subtype of lung cancer, both in patients who smoke and in those who have never smoked (3, 4). These lung adenocarcinomas are usually discovered late in the course of the disease, when options for effective therapeutic intervention are limited. Future hopes for reducing lung cancer mortality rates may rest on a better understanding of the early steps of glandular neoplasia.

Unlike squamous cell carcinoma, relatively little is currently known about the molecular events that underlie the earliest stages of lung adenocarcinoma. Until recently, the inscrutability of early glandular neoplasia was attributed to the inability to recognize and study the precursor lesion from which lung adenocarcinomas arise. This obstacle has been partly overcome with a growing awareness of the true nature of atypical adenomatous hyperplasia (AAH; ref. 5). AAH is a microscopic proliferation of atypical pneumocytes that is regarded as a form of glandular carcinoma in situ (6), a step in lung tumorigenesis preceding the onset of invasive tumor growth (Fig. 1). As a form of early glandular neoplasia, AAH is a useful target for studying the timing of molecular events in lung tumorigenesis including the activation of oncogenes and the inactivation of tumor suppressor genes. For example, some of the same genetic alterations that are common in advanced stages of lung adenocarcinomas have also been identified in AAH, such as mutational activation of the KRAS oncogene; loss of heterozygosity on chromosomal arms 3p, 9p, 17p, and 16q; and mutational inactivation of p53 (6). To date, the study of early glandular neoplasia has mainly focused on these more conventional mechanisms of gene alterations at the exclusion of epigenetic mechanisms.

Promoter hypermethylation of tumor suppressor genes is an epigenetic alteration that is increasingly recognized as an important component of the multistep cascade culminating in malignant transformation. More than 50% of human genes have CpG islands, which span the 5′ region of the transcriptional start site and the first exon (7). The methylation of targeted CpG islands prevents the transcription of tumor suppressor genes, and the inactivation of these genes drives, in part, the progressive stages of tumorigenesis. In some well-studied models of tumorigenesis such as the colorectal, promoter hypermethylation is found to occur during the early stages of tumor development (8–10). In the lung, efforts to define the role of promoter hypermethylation in the early stages of tumorigenesis have mainly focused on squamous cell neoplasia due to the visibility and accessibility of these central squamous lesions via bronchoscopy (11–13). As for glandular neoplasia of the lung, promoter hypermethylation of tumor suppressor genes is common in overtly malignant adenocarcinomas, but the presence and extent of these alterations in early glandular neoplasia are unknown. Specifically, promoter hypermethylation of tumor suppressor genes has only been evaluated in AAH for a single gene, ASC/TMS1, which was found to be methylated for only one of 18 precursor lesions analyzed (14). Long recalcitrant to molecular analysis given its small size, the development of PCR-based amplification techniques now permits a comprehensive analysis of the minute AAH including promoter hypermethylation analysis across a spectrum of genes implicated in lung tumorigenesis. Using a multiplex nested methylation-specific PCR (MSP) approach, we assessed various stages of lung tumorigenesis for the frequency of promoter hypermethylation of genes regulating cancer hallmarks including cell cycle regulation (cyclin-dependent kinase inhibitor 2A or p16), DNA repair [O⁶-methylguanine-DNA methyltransferase gene (MGMT)], retinoic acid signaling (RARβ), apoptosis [death-associated protein kinase (DAPK)], invasion [tissue inhibitor of metalloproteinase-3 (TIMP3)], Ras signaling [Ras association domain family protein 1 isoform A (RASSF1A)], and immortalization [the catalytic subunit of telomerase (hTERT)]. This comprehensive analysis of early through late glandular neoplasia provides a more complete picture of the progression of human lung adenocarcinoma.

Patients and lesion classification. All clinical samples were obtained as part of the protocol approved by the institutional review board from the Johns Hopkins University. We evaluated 121 formalin-fixed and paraffin-embedded tissues from 16 patients. Such patients were selected for having multiple AAH lesions to allow comparisons among lesions for each patient. The difference in the number of AAH lesions between each patient directly represents the normal variation of the occurrence of such lesions in patients with primary tumor, which is typically found when meticulous specimen sampling is done.

These samples included 56 AAHs, 19 synchronous primary lung carcinomas, and 46 histologically normal lung samples from the lung parenchyma adjacent to the AAHs. For patients 1, 3, 8, and 12, multiple parts of the same tumor were analyzed. The 121 samples were taken from 16 lung resections identified through a histologic review of all lung resections done at The Johns Hopkins Hospital between 1995 and 2002. The AAHs were identified using the criteria of Nakanishi (15). Specifically, AAHs were identified as a growth of cytologically atypical cuboidal to columnar cells along the alveolar septa in the absence of significant inflammation and fibrosis of the surrounding lung parenchyma (Fig. 1). Furthermore, AAHs were not included if they were contiguous with or directly adjacent to a primary lung tumor. The AAHs were further subclassified as low grade or high grade by one of us (W.H.W.) without knowledge of methylation status using the grading guidelines provided by Kitamura et al. (16, 17). AAHs classified as low grade were characterized by round to cuboidal cells with low cellular density; small cell size; minimal variation in nuclear size, shape, and chromaticity; and minimal thickening of the alveolar septa (Fig. 1B). By comparison, AAHs classified as high grade had increased cellular density, larger cell size, greater variation in cell size and shape, and mild fibrosis of the alveolar septa (Fig. 1C). AAHs tend to be microscopic lesions (<5 mm), and only those cases in which sufficient cells remained for DNA isolation were included in this study. Normal lung (also referred to as normal-adjacent) was defined as a 3- to 5-mm area of peripheral lung parenchyma adjacent to an AAH that did not show any microscopic abnormalities (Fig. 1A). For each component (normal lung, AAH, and adenocarcinoma), the formalin-fixed and paraffin-embedded tissue block was retrieved for subsequent macrodissection. Smoking histories were obtained from a review of the medical records of the patient (Supplementary Table S2). Six patients were current smokers, five were former smokers, one was a never smoker, and four had unknown smoking history.

Manual macrodissection/DNA extraction. Slides were deparaffinized by incubation at 60°C followed by xylene treatment. Slides were rehydrated in ethanol bath (95%, 90%, 75%, 50% ethanol) and finally washed with autoclaved water. Areas of normal, low-grade AAH, high-grade AAH, and adenocarcinoma were identified by one of us (W.H.W.) on an adjacent H&E slide.

Areas of interests were rehydrated with 2 μL of lysis buffer (50 mmol/L Tris, 50 mmol/L EDTA, 2% SDS, 10 mg/mL protein kinase; Sigma-Aldrich), scraped off the slide with a scalpel, and transferred into an Eppendorf tube containing 50 μL of lysis buffer. To prevent sample contamination, normal lung was sampled first followed by AAH and tumor areas. Tissue specimens and cell pellets were resuspended in lysis buffer and incubated overnight at 60°C. The next day, DNA was extracted with phenol/chloroform/isoamyl alcohol (25:24:1), ethanol precipitated, and resuspended in 50 to 200 μL of Tris-Cl buffer (pH 8.0). DNA was quantified with the NanoDrop ND-1000 Spectrophotometer (Nanodrop Technologies).

Bisulfite treatment. Bisulfite modification was carried out as previously described (18). Briefly, DNA was denatured with NaOH (2 mol/L final concentration) for 10 min at 37°C. Sodium bisulfite (3 mol/L; pH 5.0) and hydroquinone (0.05 mol/L) were added and the mixture was incubated at 50°C for 16 h. The solution was then desalted with the Wizard DNA cleanup kit (Promega Corporation) and the bisulfite conversion was completed with the addition of 3 mol/L NaOH. Finally, the modified DNA was ethanol precipitated and resuspended in 20 μL of water.

Multiplex nested MSP. Nested primers and MSP primers (Supplementary Table S1) were synthesized by Integrated DNA Technologies (Integrated DNA Technologies). Four microliters of bisulfite-treated DNA form paraffin-embedded samples were used in a 25-μL reaction tube containing 10× PCR buffer, 10 mmol/L deoxynucleotide triphosphates, 0.5 unit of JumpStart REDTaq DNA polymerase, and a mixture of forward and reverse nested primers for each of the seven genes analyzed in this study. Amplification cycles were as follows: one cycle of 95°C for 5 min, followed by 35 cycles of 95°C for 30 s, 55°C for 30 s, 72°C for 30 s, and a final extension step of 72°C for 5 min.

Each of the genes that were amplified by multiplex nested MSP for each sample was then analyzed by MSP. Specifically, each amplification reaction was diluted 1:500 and used as template in the second reaction using the appropriate MSP primers and the following amplification cycles: one cycle of 95°C for 5 min, followed by 35 cycles of 95°C for 30 s, annealing temperature (Supplementary Table S1) for 30 s, 72°C for 30 s, and a final extension step of 72°C for 5 min. Ten microliters of each amplification reaction were loaded and run on a 2% agarose gel containing GelStar. A normal human bronchoepithelial cell line (NHBE), an HCT-116 colorectal cancer cell line that has been genetically depleted for DNA methyl transferases 1 and 3b (DKO), and normal lymphocytes were used as unmethylated controls, whereas in vitro methylated DNA, DLD-1, and H69 were used as methylated controls.

Statistical analysis. Comparison of continuous and dichotomous variables was done with Student's t test and χ² test or Fisher's exact test, respectively. Tests for trend across histologic groups were achieved using Pearson correlation coefficient or Wilcoxon rank-sum test. Results were considered significant at P < 0.05 (two sided). We estimated the relative risk of exposures by calculating odds ratios (OR) and 95% confidence intervals using univariate logistic regression. Receiver operator characteristic curve analysis was done comparing different histologic groups using the number of methylated genes to determine the Wilcoxon estimate of area under the curve. All analyses were accomplished using the STATA statistical software.

Previous studies have established a causal link between promoter hypermethylation of p16, TIMP3, DAPK, MGMT, RARβ, RASSF1A, and hTERT and their transcriptional silencing in lung cancer (12, 19–25). We therefore set out to determine the frequency of promoter hypermethylation for each gene in lung adenocarcinoma and to determine the timing of these events in the initiation of adenocarcinoma through examination of AAH lesions. To do so, we obtained from each patient with invasive adenocarcinoma of the lung, multiple normal-adjacent lung specimens, and AAH lesions for analysis of promoter region methylation for these seven genes using multiplex nested MSP. This approach allows the analysis of the methylation status of the promoter of multiple genes using small amounts of fragmented genomic DNA obtained from these microscopic formalin-fixed lesions. Specifically, nested MSP is more sensitive than the conventional MSP procedures (26) in that conventional MSP is commonly used for methylation analysis of cell lines or fresh primary tissues, whereas nested MSP is particularly well suited for sensitive detection in archived specimens. Examples of the primary analysis of these tissues are shown in Fig. 2 for DAPK, p16, RASSF1A, hTERT, and MGMT. Each gene examined was frequently methylated in invasive cancers, as expected from previous data published for most genes in adenocarcinoma (no studies have examined hTERT hypermethylation in primary lung cancers), allowing an ability to track these changes back to normal-appearing tissues (27–29).

For example, we found that DAPK was methylated in 39%, p16 in 50%, MGMT in 78%, RARβ in 88%, RASSF1A in 59%, and hTERT in 56% of invasive adenocarcinoma (Fig. 3; Supplementary Table S3). However, DAPK methylation was not detected in the normal lung (normal-adjacent) specimens as shown by the detection of an unmethylated band in these samples (Fig. 2A). One of the two AAH lesions, AAH 36, as well as the adenocarcinoma sample 11 of patient 11, showed evidence of promoter hypermethylation for this gene, as seen by the presence of both an unmethylated and a methylated band in these samples. For patient 13, DAPK methylation was restricted to the adenocarcinoma sample and was not seen in normal lung or AAH lesions. Methylation analysis showing intra- and inter-patient heterogeneity is also illustrated for p16, RASSF1A, hTERT, and MGMT (Fig. 2B-E, respectively). Interestingly, we found that normal-adjacent and AAH lesions that came from different slides but from the same patient showed distinct methylation patterns, therefore suggesting that multiple “epigenetic fields” may exist within the lung of patients with adenocarcinoma (Fig. 2E).

The frequency of promoter hypermethylation was associated with advancing stages of tumorigenesis for each individual gene evaluated (Fig. 3; Supplementary Table S3). Specifically, the frequency of methylation progressively increased with each step from normal-adjacent lung to low-grade AAH to high-grade AAH to adenocarcinoma (P_trend ≤ 0.01).

Although the progressive accumulation of these epigenetic events was a consistent finding, the precise timing of individual promoter hypermethylation was not uniform but varied by gene locus. MGMT, RARβ, RASSF1A, and hTERT were often methylated in AAH as well as the adjacent normal lung, suggesting that these genes may be targeted during the earliest stages of lung tumorigenesis. For example, the frequency of methylation for MGMT increased from 21% in normal lung to 32% in low-grade AAH, 58% in high-grade AAH, and 78% in adenocarcinomas. Similarly, hTERT promoter hypermethylation was found in 10% of normal-adjacent lung, 14% of low-grade AAH, 42% of high-grade AAH, and 56% of adenocarcinomas. Promoter methylation targeting p16 and DAPK seems to occur at a later stage of tumorigenesis. For p16, promoter methylation was rarely detected in normal-adjacent lung and was never found in low-grade AAH, but it was detected in high-grade AAH (30%) and in adenocarcinomas (50%). DAPK promoter hypermethylation was not detected in normal-adjacent lung, but it was increasingly detected in low-grade AAH (10%), high-grade AAH (22%), and adenocarcinoma samples (39%). TIMP3 promoter hypermethylation was only found in adenocarcinomas and was infrequent in these tumors, suggesting that it is a later event (Fig. 3; Supplementary Table S3). In many instances, this increasing frequency of promoter hypermethylation from one histologic step to the next was statistically significant. For example, promoter hypermethylation was more likely to be encountered in high-grade AAH than in low-grade AAH for p16 (P = 0.002), MGMT (P = 0.053), and hTERT (P ≤ 0.030; Supplementary Table S3).

The total number of genes methylated also increased in direct relationship to advancing histologic grade (P_trend ≤ 0.001; Fig. 4; Supplementary Table S4). The mean number of methylated genes increased from 0.67 in normal lung to 1.30 in low-grade AAH, 2.67 in high-grade AAH, and 3.72 in adenocarcinomas (Fig. 4A). The trend observed for progressive promoter hypermethylation with advancing histologic grade was also consistently observed in individual patients with multifocal AAH and synchronous lung adenocarcinomas (see Supplementary Table S2 for complete information). Examples of the methylation profile for patients 1, 3, and 12 show the accumulation of methylation from normal lung to overt adenocarcinoma with respect to individual genes and total number of methylated genes (Table 1). In patient 1, the mean number of methylated genes was 1.1 in the normal lung, 3.0 in the AAHs, and 5.5 in the adenocarcinoma. In normal lung from patient 3, the mean number of methylated genes was 0.7 in the normal lung, 1.5 in the AAHs, and 2.5 in the adenocarcinoma. For patient 12, the mean number of methylated genes was 1.3 in the normal lung, 2.7 in the AAHs, and 3.5 in the adenocarcinoma.

The increasing frequency of methylation with histologic change suggested that methylation at individual loci or the number of changes may be useful as a diagnostic adjunct to histology. Thus, we determined the ORs for the presence of methylation as determinants of histologic categories (Table 2). The small number of methylation events for TIMP3 did not allow this calculation, but for all other genes OR ranged from 4.35 to 45.54. For example, the presence of promoter hypermethylation of p16 and RARβ had an OR of 11.38 and 9.50, respectively, for the sample being an adenocarcinoma versus the other histologic categories, and each gene had an OR of >4.64, all being statistically significant (Table 2, left). Increasing numbers of genes hypermethylated were also associated with increased ORs for the sample being an invasive tumor, with hypermethylation of four cancer genes having an OR of ∼19. However, whereas the ORs of hypermethylation were strongly able to predict that a sample was an invasive tumor, there was an even stronger association of hypermethylation in predicting tumor or high-grade dysplasia versus normal or low-grade dysplasia (Table 2, right). Specifically, samples methylated for p16 were 45 times more likely to be adenocarcinomas or high-grade AAH lesions than low-grade AAH or normal lung specimens (P < 0.001). All other genes, with the exception of TIMP3, which showed a low frequency of methylation in all samples, showed statistically significant OR, ranging from 4.35 for RASSF1A to 9.79 for DAPK (P < 0.001), and once again, the increasing number of methylation events also showed large and statistically significant ORs (Table 2). In addition, the number of methylated genes produced significant receiver operator characteristic curves, with area under the receiver operator characteristic curves ranging from 0.815 to 0.875 (Fig. 4B-D), showing the sensitivity and specificity of detection of the number of methylated genes.

In this study, we provide evidence for the role of promoter hypermethylation of hallmark cancer genes in the development of lung adenocarcinoma by means of multiplex nested MSP. Furthermore, the determination of the frequency of promoter hypermethylation across the full spectrum of glandular neoplasia including precursor lesions helps delineate the timing and sequence of these alterations, allowing distinction between early and late events. Such nonquantitative technique allows for the sensitive assessment of methylation status of multiple promoters in a two-step PCR reaction without being confounded by tissue heterogeneity that might affect quantitative analyses. It is relatively easy to optimize and requires minimal amount of starting material, which is important when working with fixed tissues.

Promoter hypermethylation of genes affecting DNA repair (MGMT), retinoic acid signaling (RARβ), Ras signaling (RASSF1A), and immortalization (hTERT) were early events that preceded alteration of apoptosis (DAPK), cell cycle control (p16), and invasion (TIMP3). Indeed, promoter hypermethylation of MGMT, RARβ, RASSF1A, and hTERT could be detected in lung parenchyma without any morphologic changes. Clearly, promoter hypermethylation is not a haphazard consequence of an unstable genome during advanced tumorigenesis. Instead, epigenetic inactivation of certain hallmark cancer genes may occur early and set a context for the accumulation of additional epigenetic and genetic events that, in turn, drive morphologic and biological progression toward malignancy. As one example, we found that MGMT promoter hypermethylation is twice as likely to occur in high-grade AAH than in low-grade AAH, a finding which parallels the increasing frequency of p53 mutations in higher-grade AAHs and supports the concept that MGMT inactivation promotes p53 mutations in lung cancer (30, 31).

Recent studies have identified two regions at the hTERT promoter that may represent binding sites for an activator and a repressor (25, 32). The activator site has been implicated at the −150/+150 promoter region, and the repressor site is believed to be located 3′ of that region (32). The area that we analyzed is located 5′ from the −150/+150 region and therefore is not a readout of active gene expression (25). Instead, this region may correspond to the site of a second repressor (32). Hence, promoter methylation of this region could, through blocking of repressive regulation, increase hTERT expression. This would then correlate with the observed progressive increase in the frequency of hTERT expression in AAH as previously observed (33). A recent study has suggested a link between cigarette smoke and increase of telomerase activity in human bronchial cells (34). We hypothesize that cigarette smoke may mediate this mechanism by the localized methylation of both repressor regions.

Other hallmark genes were targeted in later stages of tumorigenesis, suggesting that inactivation of these genes may play a role in promoting progression toward malignancy in an established neoplastic clone. For example, dysregulation of apoptosis via DAPK promoter hypermethylation contributes more robustly to the later stages of glandular neoplasia. DAPK promoter hypermethylation is not present in normal lung and is only detected in 10% of the low-grade AAHs, but it occurs in 22% of the high-grade AAHs and 39% of the adenocarcinomas. p16 was found to be methylated in ∼30% of high-grade AAH and 50% of primary cancer but was rarely found in low-grade AAH or its adjacent normal lung parenchyma. This suggests that promoter hypermethylation of the cell cycle control gene p16 occurs even later than DAPK. We therefore hypothesize that p16 inactivation may represent a critical transitional from a precursor lesion to overt adenocarcinoma. Not unexpectedly, the methlylation of TIMP3, a regulator of the extracellular matrix and an inhibitor of tumor invasion, is limited to malignant adenocarcinomas and as such was found unmethylated in the noninvasive stages of glandular neoplasia (i.e., AAH). Notably, these epigenetic alterations in glandular neoplasia of the lung do not always parallel the timing and sequence of these same alterations in squamous neoplasia of the lung. In squamous neoplasia, p16 promoter hypermethylation seems to occur during the earlier stages. It is detected in 17% of basal cell hyperplasias, 24% of squamous metaplasias, and 50% of in situ squamous cell carcinomas (12).

Likewise, the sequence and timing of promoter hypermethylation of target genes in human glandular neoplasia may not precisely parallel the process in murine models of lung tumorigenesis. In mouse models, DAPK promoter hypermethylation is detected in almost half of AAHs induced by chronic exposure to 4-methylnitrosamino-1-(3-pyridyl)-1-butanone, whereas we detected DAPK promoter hypermethylation in only 16% of the AAHs from human lungs (35, 36). Although in mouse models p16 promoter hypermethylation is an early event that precedes DAPK and RARβ methylation, we found that in human adenocarcinoma progression, it is a later hit that follows DAPK and RARβ methylation (37). Whereas murine systems have been and continue to be useful models for studying lung tumorigenesis, these differences underscore the importance of the study of human AAH as an optimal lesion for delineating the sequence of epigenetic alterations because these alterations are used for early cancer detection and high risk assessment.

Some authors have discouraged the grading of AAH, arguing that subclassification based on the severity of cytologic and architectural atypia is not consistently reproducible and of undefined clinical significance. The observations of this study support the growing recognition that AAH does not constitute a solitary step in early lung tumorigenesis, but instead encompasses a heterogeneous population of lesions representing different points along the progression toward overt lung adenocarcinoma.

Although the neoplastic nature of AAH has been examined at the genetic level, the frequency and pace of its clinical progression to overt malignancy are largely unknown. An examination of methylation of cancer-specific genes may provide insight into the risk of clinical progression of early glandular neoplasms, as suggested by the strong observed association between promoter hypermethylation of some genes, such as p16, or the number of methylation events and the degree of atypia in lung cancer patients as shown earlier. Our study identifies for the first time the sequence of hypermethylation events of key genes that contribute to the hallmark of cancer in lung tumorigenesis. The use of such genes, either individually or in combination, as epigenetic biomarkers, which could be used to identify individuals most at most risk of developing lung adenocarcinomas, will need to be further assessed. Hence, such biomarkers could in the future guide management strategies including smoking cessation, surveillance, chemoprevention, and surgical intervention. As such, detection of the epigenetic alterations in sputum samples is now feasible and a valid strategy for identifying patients at high risk for developing cancer (29, 38). A more comprehensive understanding of the timing and relevance of promoter hypermethylation could help guide the future development and implementation of these promising biomarker strategies.

We thank Dr. Rebekah L. Zinn (Department of Oncology, Johns Hopkins University, School of Medicine, Baltimore, MD) for the hTERT multiplex nested MSP primers, Dr. Mingzhou Guo for help with the MSP protocol, and Jane L. Widdowson for her continuous support during this project.