Changes in CpG Islands Promoter Methylation Patterns during Ductal Breast Carcinoma Progression

Materials and Methods

Cases Selection

Pathologic samples from 76 patients [median age, 48 y; interquartile range (IQR), 44-60] and 20 normal breast tissues distant from tumor (NBDT; median age, 56 y; IQR, 50-63) were obtained as paraffin-embedded samples or fresh-frozen specimens from the Department of Pathology “L.Armanni,” University of Naples, Italy; the Department of Pathology, IRCCS “Casa Sollievo della Sofferenza,” San Giovanni Rotondo, Italy; and the Department of Pathology, IRCCS Oncology Institute, Bari, Italy. Prior written and informed consent was obtained from each patient in accordance with institutional guidelines.

Promoter methylation status was initially determined on 52 patients for whom synchronous preinvasive and invasive lesions and/or normal breast tissue adjacent to tumor (NBAT) were available (Table 1) and 20 NBDT (training set). Additional pure preinvasive and invasive lesions from 24 patients were also tested, and results from this analysis were merged with data from the training set to constitute the validation set. In three cases multiple synchronous lesions were available for each patient (two ADH were analyzed for patient BP23, three DCIS and two IDC were available for patient BP18, and two DCIS were tested for patient BP22). Multiple lesions were located on separate paraffin blocks or at >0.5 cm in distance one from another. Clinicopathologic characteristics of the 61 patients affected by invasive breast cancer (45 synchronous lesions and 16 pure IDC) are shown in Supplementary Table S1.

DNA Extraction and Sodium Bisulphite Modification

Sections (5-μm thick) were cut from paraffin blocks or fresh-frozen specimens, H&E staining was performed, and two pathologists (A. Apicella and F. Zito) reviewed each slide to identify the areas of ADH, DCIS, and IDC, and to exclude cancer cell contamination in normal breast tissues (NBT). Sample specimens were manually dissected under a microscope from 12-μm-thick sections to enrich for areas containing ADH, DCIS, or IDC. The 5-μm-thick H&E-stained sections evaluated and marked by pathologist were used as reference slides. For each specimen, 6 to 70 H&E-stained sections were dissected depending on the dimension of the lesion. Tissues were subsequently placed in xylene and genomic DNA was extracted as described previously (18).

DNA extracted from tumors was subjected to bisulphite treatment, as described previously with minor modification (19). Briefly, 1 μg genomic DNA from each sample was denatured with NaOH (final concentration, 0.2 mol/L) in a total volume of 20 μL at 50°C for 20 min. The denatured DNA was diluted by adding 500 μL of a freshly prepared solution of 10 mmol/L hydroquinone and 3 mol/L sodium bisulphite, and incubated at 70°C for 3 h. Bisulphite-modified DNA was purified using a Wizard DNA Clean-Up System (Promega), treated with 0.3 mol/L NaOH at room temperature for 10 min, precipitated with ethanol, resuspended in 120 μL of LoTE [2.5 mmol/L EDTA, 10 mmol/L Tris-HCl (pH 8)], and stored at −80°C.

Quantitative Methyl Specific PCR

Bisulphite-modified DNA was used as template for fluorescence-based real-time PCR using a relative quantification method with standard curve to determine methylation levels in pathologic samples and controls (19, 20). In this system two amplification primers as in conventional methyl specific-PCR and a dual-labeled fluorogenic hybridization probe are used (21). One fluorescent dye serves as reporter (FAM) and its emission spectra are quenched by a second fluorescent dye (TAMRA). Primers and probe of the genes of interest were designed to specifically amplify the bisulphate-modified region containing the putative methylated CpGs, whereas primers and probe of a reference gene (ACTB) specifically amplify a bisulphite-modified region not containing CpGs (sequences and annealing temperatures are provided in Supplementary Table S2). Leukocyte DNA from a healthy subject was methylated in vitro with excess SssI methyltransferase (New England Biolabs Inc.) to generate completely methylated DNA. Serial dilutions (90-0.009 ng) of this DNA were used to construct a calibration curve for the ACTB gene and for the genes of interest and were run parallel with each analysis. The calibration curves for the reference and the target genes were construed by plotting the log [control in vitro methylated leucocyte DNA amount] to the Ct values and calculating the equation of the curve (y = mx+q where m is the slope and q the intercept). The quantity of the unknown samples is then calculated as follows: log [unknown DNA amount] = (Ct values - intercept)/slope (Applied Biosystem User Bulletin 2). These procedures are automatically done on 7900 Sequence detection system by the SDS 2.21 software. The relative level of methylated DNA for each gene in each sample is determined as a ratio of the amount of the target gene to the amount of ACTB (reference gene) and then multiplied by 1,000 for easier tabulation (average value of triplicates of gene of interest/average value of triplicates of ACTB × 1,000).

Fluorogenic PCR was set up in 384-well plates on a 7900 Sequence detector (Applied Biosystems). Each plate included patient DNA samples, positive (in vitro methylated leukocyte DNA) and negative (normal leukocyte DNA or DNA from a known unmethylated cell line) controls, and multiple water blanks. Amplification reactions were carried out in a volume of 20 μL that contained 3 μL bisulphite-modified DNA, 600 nmol/L forward and reverse primers, 200 nmol/L probe, 5 U of Platinum Taq polymerase (Invitrogen), 200 μmol/L each of dATP, dCTP, dGTP, dTTP, and 5.5 mmol/L MgCl₂. Amplifications were carried out using the following conditions: one step at 95°C for 3 min, 50 cycles at 95°C for 15 s, and 60°C to 62°C for 1 min. Supplementary Table S3 lists the nine genes whose promoters were examined, their proposed functions, and the tumors in which these promoters were shown to be hypermethylated.

Statistical Analysis

For each gene the median and IQR of methylation ratios were determined for each of the study groups. Nonparametric tests were used to analyze data samples because the Kolmogorov-Smirnov and the Shapiro tests led to rejection of the Gaussian distribution assumption for the variables analyzed. For each of the genes cutoff values were determined by drawing the Receiver Operating Characteristic (ROC) curve using methylation levels in normal breast tissues adjacent and distant from tumors (n = 32) and pathologic samples (ADH, DCIS, and IDC; n = 138). The area under the ROC curve, computed numerically and tested for statistically significance, was assumed as a measure of goodness of the test. Cutoff values were chosen among ROC curves points in order to optimize specificity while fixing a good value of sensitivity. Methylation frequencies were then calculated for each of the genes as the percentage of samples showing methylation levels equal or above the cutoff values.

For the analysis of methylation frequencies and levels, paired samples represented the training data set, whereas the extended data sample represented the validation data set. Specific statistical models were implemented, because for the paired samples not all possible lesions were present in each patient. Methylation levels were analyzed by using a hierarchical mixed model with unstructured covariance matrix and robust residuals' estimation to account for nonsymmetric distribution (22). In a similar way methylation frequencies above cutoff values in the training data set were analyzed by using a Generalized Estimating Equations logistic model (23). Such type of modeling overcomes missing data values. In both approaches, the methylation levels and frequencies were treated as repeated measures, in the different tissues/lesions, within each subject. P values reported in Table 2 refer to the statistical significance of the hierarchical mixed model's coefficients and Generalized Estimating Equations logistic model's coefficients, respectively.

In the validation set the within-gene differences that were likely to exist between groups were checked using the Kruskal-Wallis test. We then applied the Dunn test for multiple comparisons to determine among which groups the previously found differences live. Such test allowed us to merge the sample groups among which nonstatistical difference was discovered. Therefore, it was possible to carry out the comparison between preinvasive versus invasive and between NBT and pathological samples (Mann-Whitney test). Differences in methylation frequencies in the validation set were assessed using the χ² test. Because our validation set was not an independent data sample, P values of the Kruskall-Wallis, Dunn, Mann-Whitney, and χ² tests were not computed asymptotically but by adopting a permutation approach with 10,000 resamplings (24). At the first step the test statistics were computed, then at each of the 10,000 resamplings, the methylation data were randomly relabeled and the test statistics were computed for the new permutated dataset. At the final step, empirical P values were derived. These are shown on Table 3. Sensitivity and specificity to distinguish between normal and pathologic breast tissues were estimated with their 95% confidence intervals on the basis of cutoff values. Statistical analysis was done using R-project software and all tests were two-sided with significance at P ≤ 0.05.

Discussion

We investigated the pattern of methylation in a series of preinvasive and invasive breast lesions. Due to the limited amount of DNA in preinvasive lesions we were not able to test all the relevant genes that show significant alterations in breast cancer. Thus, we chose to analyze those genes that were more likely to be involved in progression from preinvasive to invasive phenotype (Supplementary Table S3). Of the nine genes tested, only APC, CDH1, and CTNNB1 showed a different methylation pattern in normal breast tissues and pathologic samples. In synchronous lesions NBAT, ADH, and DCIS showed similar methylation patterns for APC and CTNNB1, whereas statistically significant differences were seen in ADH and DCIS as compared with NBAT for CDH1. Similar levels of CDH1 methylation were detected for ADH as compared with DCIS (Table 2). Although these results may be due to the panel of genes analyzed, they are in agreement with immunophenotypic and genetic data and suggest that atypical hyperplasia should be considered as a well-differentiated or simply small in situ carcinoma (4).

The APC promoter was methylated above the cutoff value in approximately half of preinvasive and invasive lesions. An increase in methylation levels was also seen between the two pathologic groups but the difference was not statistically significant. CDH1 was methylated in 30% of the preinvasive lesions and in 42% of the invasive tumors, and methylation levels were significantly higher in IDC as compared with preinvasive lesions.

Although APC and CDH1 methylation were previously reported in approximately 40% to 50% of breast cancers (9, 16, 17, 25-27), this is the first report of CTNNB1 methylation in breast cancer. CTNNB1 promoter methylation was recently reported in gastric and endometrial cancers (16, 17). In a series of primary metastatic gastric cancer and cell lines derived from metastasis, loss of expression of β-catenin was related to promoter methylation (16). Sequence analysis of the bisulphate-treated DNA showed heavy CpG methylation of the CTNNB1 promoter region in a cell line with absent β-catenin expression, and treatment with 5-deoxyazacytidine was able to restore β-catenin expression (16). Moreover, Whitcomb et al. (17) reported methylation at the CTNNB1 promoter region in 17% of primary endometrial tumors. In mammary tumors, loss of β-catenin expression was shown by immunohistochemistry in both lobular and ductal breast carcinomas (25, 28, 29).

Interestingly, apc, cdh1 and β-catenin play an interaction role in the maintenance of cell-to-cell adhesion, and regulation of cell extracellular matrix interactions (30, 31). The E-cadherin is mainly expressed in epithelial tissues and in normal breast prevalently in the luminal cells. In developmental animal models, E-cadherin expression is temporarily down-regulated when budding lobules invade the stroma (32). Although the cell-to-cell adhesion properties of E-cadherin reside in the extracellular domain, its function depends on the interaction with catenins and in particular β-catenin, which is responsible for the anchorage of E-cadherin to the actin cytoskeleton (31). The apc protein participates in the multiprotein complex involved in the phosphorylation and subsequent ubiquitination of β-catenin (30). Moreover, APC terminal C-region can interact with cytoskeleton proteins, contributing to cell migration, proliferation, and adhesion (30). A role in “invasion suppression” for apc, cdh1 and β-catenin would be consistent with our results showing changes in promoter methylation in the switch from preinvasive lesions to invasive breast cancer. We can hypothesize that APC and CTNNB1 methylation would be an early event during breast cancer development, correlated with abnormal proliferation of the breast epithelia. CDH1 methylation would occur later and is likely to play a more direct role in the loss of cell-to-cell adhesion and the acquisition of invasive properties by the cancer cells. cdh1 is involved in maintaining cell-to-cell adhesion and is regarded as suppressor of cellular invasion (33). Furthermore, loss of cdh1 expression in primary tumors has been associated with decreased patient survival (34). Loss of E-cadherin protein expression is most frequent for infiltrating lobular tumor types and is often a biallelic event resulting from any combination of gene promoter hypermethylation, mutation, or allelic loss, whereas ductal histology often presents with varying levels of expression (35, 36). Interestingly, Shinozaka et al. (37) also found high levels of CDH1 methylation in breast carcinomas of the ductal type. In their study methylation correlated with mRNA expression levels and was more frequent in primary tumors with a more aggressive phenotype. These data are consistent with our findings that methylation level of CDH1 is significantly higher in IDC as compared with preinvasive lesions and suggest that CDH1 methylation may represent an important step in the progression of breast carcinomas of the ductal type.

TIMP3, ESR1, and GSTP1 methylation frequency and levels were similar in normal and neoplastic breast tissues from cancer patients. Hu et al. (38) investigated, with a whole genome analysis approach, methylation patterns in epithelial and myoepithelial cells, stromal fibroblasts from normal breast tissue, and in situ and invasive breast carcinomas. The analysis showed distinct epigenetic changes in all types of cells, with some genes differentially methylated in normal and neoplastic samples and others showing the same methylation pattern in cells derived from normal and neoplastic tissues (39). Thus, our results could be explained by the presence of similar levels of methylation of TIMP3, ESR1, and GSTP1 in cells from normal and cancer specimens.

The detection of breast cancer in the early stages is a key for a successful treatment of the disease. DNA methylation has an advantage over other molecular detection methods (e.g. single gene mutation, microsatellite analysis,) because it can be detected with a very high degree of specificity even in the presence of an excess of unmethylated DNA. A number of studies have reported the ability to detect breast cancer cells by DNA methylation analysis in fine needle aspirations (nipple aspirates and ductal lavages) with variable sensitivity (40-44). We showed that the determination of methylation status of APC, CDH1, and CTNNB1 is able to distinguish between normal breast and pathologic samples with a specificity of 75% and a sensitivity of 67%. Thus, it is likely that the combined detection of methylation of APC, CDH1, and CTNNB1 could be more informative than other methylation markers in identifying cancer cells in cytologic specimens.