Implications of Methylation Patterns of Cancer Genes in Salivary Gland Tumors

Salivary gland neoplasms are composed of histopathologically and clinically diverse entities of disputed histogenesis and unpredictable behavior (1–3). Efforts to identify biomarkers that assist in diagnosing these tumors and in explaining their evolution and progression have been unrewarding (4–6). The underlying reasons for such slow progress in understanding these tumors include rarity, failure to grow cell lines, clinical and methodologic interstudy differences, and the lack of a phenotypic progression model of their evolution. However, recent molecular genetic studies have provided important information on common chromosomal loci and genes alterations in a spectrum of the tumors (7–15). These data are critical in advancing our understanding of their biology and identifying new targets of potential therapeutic applications.

Epigenetic alterations may also play a role in the development of these tumors. A particular advantage of these alterations is that, unlike genetic alterations, they are reversible and can be restored in vitro and in vivo by using demethylating agents in cell lines and tumors (16–19). Methylation is one of the epigenetic modifications that play an important role in the transcriptional inactivation of tumor suppressor genes in human cancer (20). Several methylation pattern studies of cancer-related genes have shown a correlation with the biological behavior of different types of tumors and evidence for potential use in future molecular diagnostic testing (21–23). Recent methylation studies of salivary gland tumors, however, have been limited in scope and types of tumors and genes analyzed, rendering the evaluation of the role of methylation in salivary gland tumorigenesis tentative (14, 15, 20, 24, 25). Simultaneous analyses of the methylation status of multiple tumor suppressor genes are important for assessing their differential association in tumorigenesis, biological behavior, and future targeted therapy (23, 26, 27).

We investigated the methylation status and protein expression of four tumor suppressor genes, death-associated protein kinase (DAPK), O⁶-methylguanine-DNA methyltransferase (MGMT), retinoic acid receptor β2 (RARβ2), and RAS-associated domain family protein 1A (RASSF1), in the largest cohort study of benign and malignant salivary neoplasms, to date, to determine the incidence, role, and clinical significance of methylation in these tumors.

Tissue samples. Fresh tissues from 29 normal salivary glands, 23 benign, and 79 primary malignant salivary gland tumors representing the phenotypic spectrum of these entities harvested at The University of Texas M.D. Anderson Cancer Center (Houston, Texas) from 1992 to 2004 formed the materials for this study. Thirteen of the normal samples were paired with tumor samples (six benign and seven malignant). Benign neoplasms were composed of 2 myoepitheliomas, 12 pleomorphic adenomas, and 9 Warthin's tumors. The 79 malignant salivary gland tumors consisted of 14 acinic cell carcinomas (ACC), 26 adenoid cystic carcinomas (ADC), 18 mucoepidermoid carcinomas (MEC), and 21 salivary duct carcinomas (SDC) with further histologic classification provided in Table 4. The specimens were harvested by one pathologist (A.E-N.) and were evaluated for quality and tissue representation by frozen tissue section review and stored at −80°C until use. All neoplastic specimens contained at least 85% tumor cells.

Genes. The DAPK, MGMT, RARβ2, and RASSF1 suppressor genes were selected because of their high frequency of methylation reported in other types of solid tumors and in our own experience of these genes (15, 26, 28).

Methylation-specific PCR. Genomic DNA was extracted from fresh-frozen tissues using DNAzol (Molecular Research Center, Cincinnati, OH) according to the manufacturer's instructions. Extracted DNA samples were treated with bisulfite by using the CpGenome DNA modification kit (Chemicon International, Temecula, CA) following the manufacturer's instructions. Bisulfite-treated DNA was then amplified by PCR using well-designed and previously verified primer sets for unmethylated and methylated sequences of the selected CpG islands in each gene.

In the design of these primers, we included multiple cytosine residues flanking the targeted CpG islands, which were converted to uracil by bisulfate, to avoid amplification of undigested DNA (29). PCR was done in a final volume of 25 μL containing template DNA, 10× PCR buffer, 25 mmol/L MgCl₂, 0.25 mmol/L deoxynucleotide triphosphate, 50 pmol/μL each primer, and 1 unit of AmpliTaq Gold (Applied Biosystems, Branchburg, NJ). PCR amplification was initiated at 94°C for 10 min, cycled 35 times at 94°C for 45 s, followed by the specific annealing temperature for each gene (Table 1) for 45 s, and then at 72°C for 1 min, with a final extension at 72°C for 5 min. The PCR products were then electrophoresed on 2% low-melt MetaPhor agarose gel (Cambrex Bioscience Rockland, Inc., Rockland, ME) and visualized with ethidium bromide.

For control, DNA extracted from normal salivary gland tissues was treated with SssI methylase to completely convert unmethylated to methylated DNA. To qualitatively assess methylation, we used a stepwise dilution (90-5%) of SssI-treated, control DNA with untreated DNA as a reference. On the basis of the dilution results, a distinct methylated band of at least 10% of the ratio between the methylated and the unmethylated bands was used as a cutoff for scoring (data not shown). The extent of methylation was recorded as the percentage of the relative ratio of the methylated to the unmethylated bands.

Western blotting. The protein content of 42 malignant salivary gland neoplasms (14 SDC, 9 MEC, 13 ADC, and 6 ACC) was determined by Western blotting for protein expression of the MGMT, RARβ2, DAPK, and RASSF1 genes. Approximately 200 mg of each frozen tissue specimen were thawed and homogenized in Triton X-100 lysis buffer (20 mmol/L HEPES, 150 mmol/L NaCl, 1% Triton X-100, 0.1% deoxycholate, 2 mmol/L EDTA, 2 mmol/L sodium vanadate, and protease inhibitor cocktail; Sigma-Aldrich, St. Louis, MO). Samples were heat denatured at 95°C for 3 to 5 min and were fractionated by SDS-PAGE under standard conditions (120 V for 1.5 h). Fractionated proteins were transferred to a nitrocellulose membrane and blocked with 5% skim milk in TBS-T [20 mmol/L Tris, 150 mmol/L NaCl, 0.1% Tween 20 (pH 7.6)] for 1 h followed by incubation with protein-specific antibodies and luminescence buffer. Protein expression was qualitatively determined as a ratio of the intensity of specific protein to those of β-actin, with low expression defined as <10% and high expression defined as >10%.

Pyrosequencing. To validate the methylation-specific PCR results, five tumors representing different methylation levels were selected and their DNA was bisulfite treated. A different set of DNA primers were designed (Table 2) to include all sequence variations within the targeted PCR product for pyrosequencing (30).

Clinicopathologic characteristics. Of the 79 patients with malignant salivary gland neoplasms, 35 were female and 44 were male who ranged in age from 8 to 90 years (mean, 56.9 years) at the time of diagnosis. Tumors most often arose from the parotid gland [in 62 of 79 (78.5%) patients], with eight tumors arising in the tongue, five in the submandibular gland, two in the maxilla, and two in the larynx. The tumors ranged from 1.3 to 15 cm in diameter (mean, 3.8 cm), and 53.2% (42 of 79) of patients developed lymph node or distant metastases. All patients were treated primarily with surgery and 62.0% (49 of 79) also received adjuvant radiotherapy. Only 11 patients received chemotherapy. At the last follow-up (mean follow-up, 51 months; range, 3-204 months), 43 patients (54.4%) were alive.

Methylation. No methylation of any of the genes was detected in any of the 29 normal salivary tissue specimens. In contrast, methylation in at least one gene was detected in 9 of the 23 (39.1%) benign salivary gland tumors; 3 of the 12 (25.0%) pleomorphic adenomas and 6 of the 9 (66.7%) Warthin's tumors. No methylation was detected in the two myoepitheliomas. The most frequently methylated genes in benign tumors were MGMT and DAPK (17.4% each), with Warthin's tumors accounting for seven of the eight methylation events of these two genes (P = 0.02; Fig. 1; Table 3). Pyrosequencing results validated the methylation-specific PCR in all five tumors selected.

Methylation content ranged from 20% to 77% by pyrrosequencing for cases with methylation-specific PCR methylated bands of 10% to 90% (data not shown).

Table 3 shows the methylation status of different genes and tumor subtypes. Thirty-three of the 79 (41.8%) malignant tumors showed at least one methylated gene, with four tumors concurrently manifesting two methylated genes (Fig. 1). In the malignant tumors, 6 of the 14 (42.9%) ACCs showed methylation at the RASSF1 gene. Of the 26 ADCs, 8 (30.8%) showed methylation at one of the four genes; two at the DAPK, one each at the MGMT and RARβ2, and four at the RASSF1 gene. Six of the 18 (33.3%) MECs showed methylation in at least one gene (MGMT or RARβ2), with one at both the RASSF1 and the RARβ2 genes. Of the 21 SDCs, 13 (61.9%) showed methylation at one, two, or three genes. Methylation in 10 of the 13 (76.9%) occurred at the RASSF1 gene; four SDC tumors showed concurrent methylation at the RASSF1 and RARβ2 genes with one of these also at the DAPK gene. The most frequently methylated genes in malignant tumors were RASSF1 (26.6%) and RARβ2 (12.6%). Methylation status was also correlated with histologic grade in the malignant tumors (Table 4). High-grade phenotypes (dedifferentiated ACC, solid or dedifferentiated ADC, high-grade MEC and SDC) were more often methylated than low-grade phenotypes (21 of 35 versus 12 of 32 respectively; P = 0.006).

Within malignant tumors, a highly significant difference in the degree of methylation at the RASSF1 gene, in SDCs (P < 0.001) and ACCs (P < 0.001), compared with the methylation at the remaining genes was found. In all five tumors (two ADCs and three SDCs) with complete methylation (100%), this occurred at the RASSF1 gene.

Protein expression.Table 5 presents the correlation between protein analysis and gene methylation in different tumor types. Of the 42 malignant tumors analyzed for expression of the proteins encoded by the four genes tested, 22 showed methylation of at least one of these four genes. The DAPK protein was detected in only 5 of the 42 malignant tumors; no DAPK protein was identified in any SDCs, MECs, or ACCs; and DAPK protein was faintly detected in only 5 of 13 ADCs. MGMT, RARβ2, and RASSF1 proteins were detected in 27 or 28 of the 42 malignant tumors (Table 5).

Of the four tumors with methylation at the DAPK gene, Western blotting showed no DAPK protein. Similarly, no MGMT protein expression was detected in the two MECs with methylated MGMT. However, MGMT protein was not detected in 13 additional malignant tumors without methylation at the MGMT gene. In contrast, 16 of 21 tumors with methylation at the RARβ2 and RASSF1 genes showed detectable RARβ2 and RASSF1 proteins. In six of these cases, however, protein expression was low and not significantly expressed. We conclude from this that the protein expression of methylated and unmethylated genes varied among individual genes (Fig. 2), although the absence of protein in 11 of 27 methylated genes was noted.

Clinical correlation.Table 6 correlates methylation status to the different clinicopathologic factors of salivary gland carcinomas. A significant statistical association was noted between tumor methylation status and age of patient at diagnosis (P = 0.05). Although no significant statistical correlation was found between methylation pattern and tumor metastasis, 52.4% (22 of 42) of metastatic tumors showed methylation of at least one gene compared with 29.7% (11 of 37) of nonmetastatic tumors that showed methylation (P = 0.07).

Our study showed, for the first time, nonrandom and differential tumor suppressor gene methylation patterns between benign and malignant salivary gland tumors. Benign neoplasms showed a greater frequency of methylation at the DAPK and the MGMT genes, whereas malignant tumors showed frequent methylation at the RARβ2 and RASSF1 genes, suggesting a different role for these genes in the biological development of these tumors. These findings may have potential diagnostic implications, especially in fine-needle aspiration studies of these tumors. We also observed, notwithstanding the qualitative nature of the methylation analysis, that benign neoplasms manifested a relatively lower degree (10-40%) of methylation at specific genes, in contrast to that of malignant tumors (40-100%). These differences could not be attributed to the inclusion of nonneoplastic host cells within specimens because both benign and malignant specimens contained a similar proportion of tumor elements. Thus, hemimethylation dilutional effect by host normal cells is not the likely source of these differences. We speculate that such variation may reflect a functional biological level for these genes in salivary tumorigenesis. However, the importance of our observation requires further quantitative analysis and clinical correlations.

Our results showed statistically significant differences in the frequency of methylation at individual genes in different malignant tumor subtypes. RASSF1 was the most frequently methylated gene in SDCs and in high-grade and metastatic ACCs. Similar findings have been made in high-grade mammary adenocarcinomas (31), and together these findings further support the subcellular genomic resemblance of salivary and breast ductal carcinomas (32, 33). Our data and those reported for other human malignancies provide further evidence for the role of RASSF1 gene methylation in carcinogenesis (21, 22, 34–38). The RASSF1 gene is located on the chromosome 3p21.3 locus and encodes the RAS-associated oncoproteins (21, 39). The gene encodes two major isoforms, RASSF1A and RASSF1B, by alternative splicing and the use of different promoters. The RASSF1 protein heterodimerizes with other family members to induce RAS gene activation and tumorigenesis.

These findings indicate that the suppression of RASSF1 activity constitutes a prerequisite event, either through the RAS pathway or independently, in cell cycle arrest, motility, and apoptosis during tumorigenesis (40, 41). We contend that modification of this gene using new demethylating agents may offer a therapeutic option for the management of salivary gland carcinomas. Of particular interest in this study is the finding of methylation at the RASSF1 gene in 2 of the 12 pleomorphic adenomas analyzed, which were histologically indistinguishable from the other pleomorphic adenomas in the cohort. These tumors are commonly the primary setting for the development of SDCs (9, 10). Our finding therefore suggests a potential association of RASSF1 methylation with the malignant progression of some pleomorphic adenomas and deserves further investigation.

We also noted increased methylation at the MGMT gene in Warthin's tumors and a subset of MECs. This finding, along with the previous clinical and molecular resemblance noted between these different tumors, lends further credence to the idea of a common subcellular association in their development (13, 41, 42). Our data revealed a higher incidence of methylation at the RARβ2 gene in SDCs and high-grade MECs compared with other types of carcinomas. These findings, along with those reported for other tumor types, support the role for methylation at this gene in the development of an aggressive phenotype (22, 27, 28, 43–47). The RARβ2 gene maps to chromosome 3p24 and codes for multiple isoforms (β1, β2, 4) by alternative splicing and the use of two different promoters (6, 28, 43, 48). Among these isoforms, the loss of expression of RARβ2 has been implicated specifically in the carcinogenesis of several malignant tumors, including mammary ductal carcinoma (43–46, 48–50). This finding further supports the role of RARβ2 in the biological progression of salivary and mammary duct carcinomas, among others (51).

In our study, although no apparent association was found between protein expression and the methylation pattern of any of the genes, we did observe reduced or absent protein expression in tumors with methylated genes. However, in any given histologic phenotype, the protein level varied in tumors with methylated and unmethylated genes, where high protein in a methylated genes and low or no protein in an unmethylated genes were observed. Furthermore, the protein encoded by DAPK, one of the least frequently methylated genes in salivary malignancies, was not expressed in most of the tumors analyzed. This could be due either to the methylation of unselected CpG islands critical for translational activity or to a genetic and posttranscriptional modification of this gene. The detection of protein in some tumors with methylated genes could be related to incomplete methylation or to intratumoral clonal heterogeneity. Although host cell contamination may have contributed to these findings, it does not seem to be significant because variation in intratumoral protein level of different genes and the relative lack of correlation with the methylation status of the particular gene were noted. Our findings, however, highlight the complexity in relating protein expression with the methylation status in human solid tumors. Further quantitative assessments of these cases are under way.

Our clinicopathologic evaluation indicates that patient sex and the size and stage of the primary tumor are not associated with methylation status. However, a significantly higher incidence of methylation in tumors from older patients was found. This finding has been made in other studies and underscores the association of methylation and tumor development in older populations (52). We also observed an apparent association, although not statistically significant, between gene methylation and the incidence of metastases. Higher number of methylated genes was found in 22 of 42 (52.4%) tumors with metastasis than nonmetastatic tumors (P = 0.07). A similar association was previously reported in aggressive histologic subtypes of different tumor entities (33, 51, 53). These data support a biological association of methylation of tumor genes and aggressive clinicopathologic variables in cancer patients.

In conclusion, our study represents the first comprehensive comparison of the gene methylation profiles of four major tumor suppressor genes in salivary gland tumorigenesis. Our results indicate that the RASSF1 and RARβ2 genes are frequently implicated in SDC development and thus could serve as targets for therapy. The degree of methylation at both MGMT and DAPK genes was insignificant, indicating a minimal role for this modification in the development of salivary gland tumors. These findings suggest that an understanding of concurrently methylated genes may provide useful information for diagnosing and managing salivary gland tumors.

The authors thank Dr. Jean-Pierre Issa for his constructive suggestions and advice.