Main

Mutations in tumor suppressor genes, such as BRCA1 and p53, are known to be important in breast cancer development. Epigenetic events, such as DNA methylation, may also play an important role. DNA methylation, the addition of a methyl group to the cytosine residue of CpG dinucleotides within gene promoters, has been implicated in cancer development because many tumor suppressor genes are silenced by DNA methylation (1). Genes found to be methylated in breast cancers include the following: p16, cyclin D2, BRCA1, ER, PR, 14ā€“3-3Ļƒ, E-cadherin, TIMP-3, GSTP1, RASSF1A, Twist, and RARĪ²2 (2, 3, 4, 5, 6, 7, 8, 9). The reported frequency of methylation of these genes in breast carcinoma ranges from 15% for BRCA1 to >90% for 14ā€“3-3Ļƒ.

RARĪ²2 is one of the nuclear receptors for the active derivative of vitamin A, retinoid acid (RA), which is critical for normal development and differentiation. Studies have shown that RARĪ²2 also mediates the anticancer effect of RA (10, 11, 12). Down-regulation of RARĪ²2 has been observed in a number of different types of human malignancy, including breast carcinoma (10, 13, 14, 15). In these cases, DNA methylation is responsible for the decreased transcription of RARĪ²2 more often than RARĪ²2 mutation (14, 16). The methylation of RARĪ² also confers RA resistance by facilitating the deacetylation of histones in the promoter region of RARĪ² (17). RASSF1A, the human Ras association domain family 1A gene, has also been implicated in multiple cancers, including lung, breast, and bladder carcinoma (18, 19, 20, 21, 22). Although its exact role in tumorigenesis is not certain, exogeneous expression of RASSF1A in tissue culture cells reduces in vitro colony formation and in vivo tumorigenecity (19). Increased expression of cyclin D2, a D-type cyclin, has been associated with proliferation of tumor cells and poor prognosis in gastric carcinoma (23). However, down-regulation of cyclin D2 expression was noted in breast cancer cell lines, as well as primary tumors, in comparison to normal breast epithelial or stromal cells (4). This paradoxical down-regulation of cyclin D2 suggests other functions for this molecule, possibly in apoptosis (4).

Although methylation of individual tumor suppressor genes has been demonstrated in breast tumor development, there has not been a detailed characterization of the methylation profiles of breast carcinoma subtypes. Also, the utility of gene methylation profiling in diagnostic breast cytopathology has only begun to be addressed (24). Therefore, we generated methylation profiles of RARĪ²2, RASSF1A, and cyclin D2 from a spectrum of breast lesions ranging from benign to invasive to assess the clinical utility of such profiling in surgical and cytological specimens and to characterize different histologic subtypes of breast carcinoma.

MATERIALS AND METHODS

Case Selection

Cases were selected from Johns Hopkins Hospital archives after obtaining IRB approval. Thirty-six benign lesions (intraductal papillomas [n = 32], radial scar/sclerosing adenosis [n = 4]), 21 carcinomas in situ (ductal carcinoma in situ [n = 20], lobular carcinoma in situ [n = 1]), and 45 invasive carcinomas (ductal [n = 15], lobular [n = 15], tubular [n = 12], medullary [n = 1], papillary [n = 1], and carcinosarcoma [n = 1]) were included in the study. Eight cases of invasive carcinoma with corresponding FNA cytology and surgical specimens were also analyzed. An additional 17 indeterminate cytology specimens were analyzed, with each having a cytological diagnosis of ā€œsuspicious for malignancy,ā€ ā€œmalignancy cannot be excluded,ā€ or ā€œatypia.ā€

DNA Extraction and Sodium Bisulfite Treatment

The procedures were carried out according to protocols described elsewhere (25). Briefly, paraffin-embedded tissue blocks were retrieved, and for each surgical pathology specimen, one section was prepared for H&E stain and morphologic evaluation, and four sections (10 Ī¼m thickness) were cut for DNA extraction. Paraffin was removed by incubating in 500 Ī¼L of xylene for 30 min at room temperature, twice, followed by two washes with 1 mL of 95% ethanol. The deparaffinized tissue sections were then allowed to dry and were incubated in 200 Ī¼L of 1 Ɨ TNES buffer (10 mm Tris, pH 8.0; 150 mm NaCl; 2 mm EDTA; 0.5% SDS; with 0.5 mg/mL proteinase K, at 55Ā° C overnight. DNA was purified using a DNA Wizard Clean-up Kit according to the manufacturer's protocol (Promega, Madison, WI) and quantified. For archival cytological specimens, coverslips were removed from either air-dried, Diff-Quickā€“stained slides or from 95% ethanol-fixed, Papanicolaou-stained slides by immersion in xylene, then allowed to dry at room temperature. Next, 150 Ī¼L of 1 Ɨ TNES buffer was added to the dried surface of slides, and cells were removed by scraping the surface with a blade and transferring the material into an Eppendorf tube. Proteinase K was added to a concentration of 0.5 mg/mL and incubated at 55Ā° C overnight. The supernatant was used for sodium bisulfite treatment and methylation-specific PCR (MSP) analysis according to published procedures (25). Briefly, 50 Ī¼L of DNA (ā‰¤1 Ī¼g) was denatured by adding 5.5 Ī¼L of 2 m NaOH for 10 min at 37Ā° C. Next, 30 Ī¼L of 10 mm hydroquinone (Sigma) and 520 Ī¼L of 3 m sodium bisulfite (Sigma) at pH 5, both prepared fresh, were added. Samples were then layered with mineral oil and incubated at 50Ā° C overnight. Modified DNA was then purified using the DNA Wizard Clean-Up Kit (Promega) according to the manufacturerā€™s protocol and eluted with 50 Ī¼L of water. Chemical modification was completed by treating DNA with 5.5 Ī¼L of 3 m NaOH and incubating for 5 min at room temperature. DNA was precipitated with ethanol and resuspended in 20 Ī¼L of water and stored at āˆ’20Ā° C until use.

Methylation-Specific PCR

The RARĪ²2 and cyclin D2 sequences of primers used in this study have been reported previously (24; Invitrogen, Carlsbad, CA). The RASSF1A primer sequences were as follows: unmethylated forward, 5ā€²-GGTTGTATTTGGTTGGAGTG-3ā€²; unmethylated reverse, 5ā€²-CTACAAACCTTTACACACAACA-3ā€²; methylated forward, 5ā€²-AGCGAAGTACGGGTTTAATC-3ā€²; and methylated reverse, 5ā€²-GCACCACGTATACGTAACG-3ā€². The PCR mixture contained 1 Ɨ MSP buffer (17 mm ammonium sulfate; 67 mm Tris, pH 8.8; 67 mm MgCl2; and 10 mm 2-mercaptoethanol), dNTPs (each at 1.25 mm), primers (100 ng/mL), and sodium bisulfite-modified DNA (1.5 Ī¼L) in a final volume of 50 Ī¼L. Reactions were hot-started at 95Ā° C for 5 min before the addition of 10 Ī¼L of 1:10 Taq polymerase (RedTaq; Sigma, St. Louis, MO) or were performed without hot start when using Jump-start Red-Taq (Sigma). Amplification was carried out in a Hybaid Omnigene thermal cycler for 36 cycles (30 s at 95Ā° C, 30 s at 56Ā° C, and 45 s at 72Ā° C), followed by a 5-min extension at 72Ā° C. A negative control (water only) and positive controls were included for each set of MSP. DNA from the breast cancer cell line MDA-MB-231, was used as a positive control for methylated products, and DNA from normal white blood cells or the EBV-transformed lymphoblast cell line, NLBL1, was used as a positive control for unmethylated reactions. Each PCR reaction was then subjected to electrophoresis on a 2% agarose gel and stained with ethidium bromide to visualize PCR products.

Statistical Analysis

All p-values are based on Pearson Chi-square test except the ones with asterisks, which are based on a Fisherā€™s Exact test (because of sparse data).

RESULTS

Promoter Methylation Frequency in Benign Breast Lesions, In Situ Carcinomas, and Invasive Carcinomas

We analyzed the methylation status of the RARĪ²2, RASSF1A, and cyclin D2 gene promoters in 36 archival samples of benign breast lesions, predominantly intraductal papillomas (Fig. 1). At least one of the three promoters was methylated in 42% of the cases (cumulative methylation index, CMI). This was predominantly due to methylation of RARĪ²2 and RASSF1A, which displayed a much higher methylation frequency (33% and 34%, respectively) than cyclin D2 (<3%). There was no difference between the mean ages of the patients in the groups with methylated genes versus unmethylated genes (55.6 y and 57.7 y, respectively). We next examined the methylation frequency of the same gene promoters in a group of 21 in situ breast carcinomas, predominantly ductal carcinoma in situ (Fig. 1). This in situ carcinoma group had a significantly higher frequency of promoter methylation than the benign lesions (CMI of 76% and 42%, respectively, P = .03). The methylation frequency for the individual genes, RARĪ²2, RASSF1A, and cyclin D2, was 60%, 62%, and 57%, respectively, among these in situ carcinomas. There was a significant difference in the methylation frequency of cyclin D2 between benign and in situ carcinomas (57% versus 3%, P < .001*). There was also a trend toward more frequent methylation of the other two genes among in situ carcinomas, relative to benign lesions, but these differences were not statistically significant (RARĪ²2: 60% versus 33%, P = .06; RASSF1A: 62% versus 34%, P = .05).

FIGURE 1
figure 1

Frequency of gene methylation in benign breast lesions, in situ carcinomas, and invasive carcinomas. The percentage of cases that contain methylated RARĪ²2, RASSF1A, and cyclin D2 genes is indicated. The percentage of cases that contain at least one methylated gene from this three-gene panel (cumulative methylation index, CMI) is also indicated.

Full size image

The methylation frequency of the same gene promoters was then analyzed in a group of 45 invasive breast carcinomas (Fig. 1). This group included the following subtypes of invasive carcinoma: ductal carcinoma, lobular carcinoma, tubular carcinoma, and several less common carcinomas. The cumulative methylation index for these invasive carcinomas was 96%. The individual genes (RARĪ²2, RASSF1A, and cyclin D2) were methylated in 64%, 64%, and 71% of the cases, respectively. The cumulative methylation index of this group is significantly higher than that of the benign group (96% versus 42%, P < .0001) or the in situ carcinoma group (96% versus 76%, P = 0.01*). There was no significant difference in the methylation profiles among the subtypes of invasive carcinoma using this three-gene panel (P > 0.05; Fig. 2).

FIGURE 2
figure 2

Gene methylation profiles of histologic subtypes of invasive breast carcinomas. Each row represents MSP results from an individual case. A filled box indicates methylation of that gene, whereas an open box indicates no methylation; ni indicates cases that were not informative because of the lack of a product upon MSP.

Full size image

High Concordance of Methylation Profiles between Corresponding Surgical and Cytological Specimens

Eight cases of invasive carcinoma with corresponding cytology and surgical specimens were analyzed to determine the feasibility of MSP analysis on archival cytological smears and the concordance of methylation profiles between the two types of specimens. Preliminary studies in our laboratory using air-dried, Diff-Quikā€“stained tissue culture cell smears and ethanol-fixed, Papanicolaou-stained smears indicated that the two cytological preparations yielded equally informative MSP results (data not shown), and thus cytological archival material was suitable for this analysis. We then compared the methylation profiles from archival cytological specimens with those of the corresponding surgical specimens. The concordance of RARĪ²2, RASSF1A, and cyclin D2 methylation was 63%, 83%, and 100%, respectively, between cytology and surgical specimens (Fig. 3). The cyclin D2 methylation profile matched exactly between surgical and cytology specimens, whereas the RASSF1A MSP results differed in one case. RARĪ²2 displayed more discordance in MSP results between cytologic and surgical specimens, possibly because of sampling differences between specimens or differences in the sensitivity of the MSP assays for different genes.

FIGURE 3
figure 3

Correlation of methylation profiles between corresponding surgical and cytologic biopsies. Each row represents MSP results from an individual patient. The Cyto column contains results from FNA samples, and the SP column contains the results from the corresponding surgical pathology biopsies. A filled box indicates methylation of that gene, whereas an open box indicates no methylation; ni indicates cases that were not informative because of the lack of a product upon MSP.

Full size image

Methylation Profiles of Atypical Breast FNA Biopsies

To test the utility of MSP to identify malignancy in atypical breast FNA biopsies, we examined the methylation profiles of 17 atypical breast FNA samples and compared these results to the ultimate histologic diagnosis. These histologic diagnoses included 12 cases of malignancy (3 DCIS and 9 invasive carcinomas) and 5 benign cases. Of the 17 FNAs, 8 cases contained at least one methylated gene from our three-gene profile (Fig. 4). All of the eight FNA cases that contained at least one methylated gene were ultimately found to be malignant. Conversely, none of the benign lesions were found to contain any methylated genes in the FNA sample. However, four malignant cases did not contain any methylated genes from our three-gene panel. Consequently, the sensitivity and specificity of this assay for malignancy in these atypical breast FNAs is 67% and 100%, respectively.

FIGURE 4
figure 4

Gene methylation profiles on atypical breast FNA specimens. Each row represents MSP results from an individual FNA sample that was originally classified as atypical. Cases 1ā€“5 are FNA samples from lesions that were subsequently found to be benign, Cases 6ā€“8 are from DCIS, and Cases 9ā€“17 are from invasive carcinomas. A filled box indicates methylation of that gene, whereas an open box indicates no methylation; ni indicates cases that were not informative because of the lack of a product upon MSP.

Full size image

DISCUSSION

Gene methylation is a process in which the enzyme DNA methyltransferase adds a methyl group to the cytosine on CpG dinucleotides within the promoter of a gene. DNA methylation represses gene expression by attracting methyl domainā€“binding proteins to the region, which recruit other transcriptional repressors to the promoter. Binding of these proteins also influences the posttranslational modification of histones, which further represses transcription (reveiwed in 26).

RARĪ²2 methylation has been examined previously in breast cancer cell lines and primary tumors. In these studies, the frequency of RARĪ²2 methylation ranged from 38ā€“63% (10, 16, 17) in invasive carcinomas, whereas the gene was rarely methylated in benign breast tissue. Our results are similar but suggest that RARĪ²2 methylation might be slightly more frequent in our patients. We have also found that RARĪ²2 methylation is a relatively frequent event in benign intraductal papillomas and DCIS. We also have extended this work by examining the most common histologic subtypes of invasive breast carcinomas. We did not find a statistically significant difference in the methylation frequency of RARĪ²2 in ductal, tubular, or lobular breast carcinomas, suggesting that RARĪ²2 methyation may play a similar role in the carcinogenic pathways used by these three subtypes of breast carcinomas and is likely to be an early event.

RASSF1A methylation has also been previously described in primary breast carcinomas and breast carcinoma cell lines. In general, the frequency of RASSF1A methylation in invasive breast carcinomas is high (49ā€“65%) (27, 28, 29), but it was significantly lower in one study (9%) (30). Our data agree with studies suggesting that RASSF1A methylation is a relatively frequent event in invasive breast carcinoma. Our data found somewhat different frequencies of RASSF1A methylation among three common histologic subtypes of breast carcinoma, but these differences were not statistically significant. Another study found that 42% of DCIS lesions contained a methylated RASSF1A gene (28), slightly lower than our findings. The same study found no RASSF1A methylation in normal breast. Lehman et al. (31) have found frequent methylation of RASSF1A in benign hyperplastic breast lesions and in intraductal hyperplasia but never in normal ductal epithelium. Similarly, our study found a significant frequency of RASSF1A methylation (34%) in benign breast lesions (intraductal papillomas), supporting the idea that RASSF1A methylation is an important event in abnormal breast epithelial proliferation.

Cyclin D2 methylation has been less studied in cancers. Little is known about cyclin D2ā€™s involvement in breast cancer development, other than that transcription of cyclin D2 is down-regulated, both in sporadic and familial forms of cancer cases (32). Recent studies in gastric and pancreatic cancers showed that the methylation of cyclin D2 is responsible for its down-regulation in those two types of cancer (33, 34). The down-regulation is clearly due to its promoter methylation, which has been demonstrated in tumor cell lines and microdissected tumors. One recent study by Lehmann et al. (31) found cyclin D2 methylation, predominantly in DCIS (particularly high-grade) and invasive carcinomas. Another study by Evron et al. (24) found cyclin D2 methylation in 46% of breast carcinomas and associated this methylation with decreased cyclin D2 gene expression. Our work confirms that cyclin D2 methylation is a frequent event in breast carcinoma and is rare among intraductal papillomas. There was no significant difference in the frequency of cyclin D2 methylation among three histologic subtypes of invasive carcinoma, suggesting that these subtypes share a common carcinogenic mechanism involving cyclin D2.

Accurate diagnosis of malignancy in breast core biopsies and FNA biopsies is sometimes challenging and may require ancillary studies. We have successfully obtained gene methylation profiles from breast FNA samples, suggesting that application of MSP to clinical FNA samples is feasible. We also have found a high concordance of methylation profiles between breast FNAs and subsequent resection specimens, particularly for cyclin D2. Our study and others indicate that cyclin D2 methylation may provide a useful marker for malignancy in breast lesions because of its prevalence in carcinomas and infrequency in benign breast lesions. Future studies may identify other methylated genes that would complement a diagnostic panel of markers.

Finally, studies of gene methylation in breast carcinoma may also have therapeutic implications. Numerous drugs have been shown to alter gene methylation in vitro and in vivo, raising the possibility that reexpression of methylated genes in breast cancer may be useful. It has been shown that treatment of breast cancer cell lines with 5-Aza-2ā€²-deoxycytidine (a demethylating agent) in conjunction with trichostatin (a histone deacetylase inhibitor) or all-trans-retinoic acid can induce reexpression of the estrogen receptor and RARĪ²2 (14, 35, 36). Future clinical trials may use MSP-based assays on FNA biopsy samples to identify patients who are eligible for gene reexpression therapy, based on their methylation profiles. Subsequent therapy with these agents may be monitored with repeated methylation profiles on FNA biopsies.

In short, methylation-specific PCR is a sensitive assay that is feasible for a diagnostic pathology laboratory; however, more work is clearly needed before methylation profiling becomes an adjunct to diagnostic pathology. The need to monitor new drugs that affect DNA methylation in tumors may also promote the development of clinical DNA methylation assays (26).