Hypermethylation of CDKN2A exon 2 in tumor, tumor-adjacent and tumor-distant tissues from breast cancer patients

Table 1 summarizes the characteristics of the breast cancer patients, including menopause status, histologic type, histological grading, B classification, proliferative activity (MIB-1), status of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2/neu) as well as the molecular subtype.

Breast cancer cell lines MCF-7, MDA-MB-231 and ZR-75-1 were grown in Dulbecco’s Minimal Essential Medium (DMEM), Leibovitz’s L-15 and RPMI-1640, respectively. Culture media were supplemented with 10% fetal calf serum (PAA, Austria). Cell cultures were periodically checked for mycoplasma contamination.

Genomic DNA was extracted using the QIAamp DNA Mini Kit (Qiagen, Germany) according to the manufacturer’s recommendations. The DNA concentration was determined with a Nanodrop 2000c spectrophotometer (Thermo Scientific, USA).

DNA extracted from biopsy samples and commercially available human control DNA (CpGenome Universal Methylated DNA, Millipore, USA and EpiTect Control DNA (human), unmethylated, Qiagen) were treated with sodium bisulfite using the EpiTect Fast Bisulfite Kit (Qiagen) following the manufacturer’s protocol.

Primer sequences for APC [32], BRCA1 [33], CDKN2A (promoter) [34] and PTEN [35] were taken from literature, those for CDKN2A (exon 2), ESR1, HER2/neu and LINE-1 were designed in-house. Nucleotide sequences were obtained from the National Center for Biotechnology Information (NCBI; [36]) database. The promoter region of the genes was identified using the Transcriptional Regulatory Element Database (TRED; [37]) or the Eukaryotic Promoter Database (EPD; [38]). Primers were designed with the Methyl Primer Express Software v1.0 (Applied Biosystems, USA). For each MS-HRM method, the annealing temperature (T_a) and the additional MgCl₂ concentration were optimized in-house. Sequences of the MS-HRM primers and their optimized conditions are summarized in Table 2.

Polymerase chain reaction (PCR) and MS-HRM analysis were carried out using a Rotor-Gene Q instrument (Qiagen) and the EpiTect HRM PCR Kit (Qiagen). The reaction mixture per well had a total volume of 20 μl and included 10 μl 2× EpiTect HRM PCR Master Mix (Qiagen), varying amounts of MgCl₂, forward and reverse primer, RNase-free water and 10 ng of bisulfite converted DNA. Amplification was performed under the following conditions: initial PCR activation step at 95 °C for 5 min followed by 50 cycles at 95 °C for 10 s, T_a of the respective primer set for 30 s (Table 2) and 72 °C for 10 s; denaturation step at 95 °C for 1 min followed by a hybridization step at 40 °C for 1 min. PCR was directly followed by a HRM step where the temperature was increased by 0.1 °C increments per 2 s.

MS-HRM data was evaluated with the Rotor-Gene Q Series Software 2.1.0 (Qiagen). The DNA methylation status was determined with the help of calibration curves established by analyzing DNA standards differing in their methylation status. These standards were prepared by mixing unmethylated and methylated human control DNA in different proportions. From the normalized HRM curves we calculated the average of the normalized fluorescence signal for each standard over the entire temperature. Calibration functions were established with SigmaPlot 11.0 (Systat Software Inc., USA). Limit of detection (LOD) and limit of quantification (LOQ) of the MS-HRM methods were determined by repeatedly analyzing unmethylated control DNA. We calculated the mean and the standard deviation. The LOD (signal-to-noise ratio 3) was determined by adding three times the standard deviation and the LOQ (signal-to-noise ratio 10) by adding ten times the standard deviation to the mean.

Primer sequences for ESR2 were designed in-house using the PyroMark Assay Design Software 2.0.1.15 (Qiagen). The nucleotide sequence was taken from the NCBI database, the promoter region was identified using TRED. Primer sequences are listed in Table 3.

PCR was performed in the Rotor-Gene Q instrument using the PyroMark PCR Kit (Qiagen). Each PCR reaction consisted of 12.5 μl PyroMark PCR Master Mix (2×), CoralLoad Concentrate (10×), forward and reverse primer, RNase-free water and 10 ng of bisulfite converted DNA. PCR conditions were as follows: initial activation step at 95 °C for 15 min, 50 cycles: 30 s at 94 °C, 30 s at 61 °C, 30 s at 72 °C and a final extension at 72 °C for 10 min. Pyrosequencing analyses were performed in the PyroMark Q24 Advanced instrument (Qiagen) by using PyroMark Q24 Advanced CpG Reagents (Qiagen). Sample preparation was carried out with the PyroMark Q24 Vacuum Workstation (Qiagen) as described previously [39].

To check identity and purity, PCR products randomly selected were loaded onto a 2% agarose gel in 1× TAE buffer, stained with GelRed (Biotium, USA) and visualized with an UVT-20 M transilluminator (Herolab, Germany).

In MS-HRM analysis, bisulfite-treated DNA from breast tissue samples was analyzed in duplicates on at least two different days. In PSQ analysis, the bisulfite treated DNA was subjected to at least two PCR reactions. Each PCR product was subsequently sequenced once.

Methylation levels were treated either as categorical variables (MS-HRM methods: < LOD, < LOQ or ≥ LOQ; PSQ method: < LOQ or ≥ LOQ) or as continuous variables. If the methylation status was treated as continuous variable, a methylation status < LOD or < LOQ was substituted with a default value, namely half the LOD or half the LOQ, respectively, as proposed previously [40].

For principal component analysis (PCA), Qlucore’s Omics Explorer V3.2 (64bit) was used. The continuous methylation data was loaded into the software and each variable was normalized to mean 0 and variance 1. Ordering of the samples was based on the total variance captured by each component.

MS-HRM assays for determining the methylation status in the promoter regions of ESR1 and HER2/neu, exon 2 of CDKN2A and LINE-1 were developed and optimized in-house. The same holds for the PSQ method allowing the determination of the methylation status of the ESR2 promoter. MS-HRM assays for the promoters of APC, BRCA1, CDKN2A and PTEN were taken from literature. Primer sequences and experimental conditions are summarized in Tables 2 and 3. Figure 1 indicates the position of the CpGs targeted by the assays in relation to the respective transcription start site (TSS).

Figure 2 shows representative normalized melting curves of DNA standards obtained with the MS-HRM assays for ESR1 (a) and CDKN2A exon 2 (c). The corresponding calibration curves are shown in Fig. 2b and d, respectively. Each of the MS-HRM assays applied in the present study was validated with regard to limit of detection (LOD; S/N = 3), limit of quantification (LOQ; S/N = 10) and inter-day repeatability. LOD and LOQ were determined by repeatedly analyzing bisulfite-treated, unmethylated control DNA, the inter-day repeatability by analyzing mixtures of bisulfite-treated methylated and unmethylated control DNA on different days. With the exception of the assay for CDKN2A exon 2, the MS-HRM assays showed a PCR bias towards methylated alleles in PCR amplification, resulting in very low LODs and LOQs in the range from 0.1 to 1.0% and 0.5 to 3.3%, respectively (Table 2). LODs and LOQs of the MS-HRM assays for CDKN2A exon 2 and LINE-1 were 3.5% and 10.7% and 11.7% and 22.3%, respectively. The calibration curves (Fig. 2b and d) demonstrate the high inter-day repeatability of the assays. Figure 2e shows a representative pyrogram obtained with the PSQ assay for ESR2. By repeatedly amplifying unmethylated control DNA and subjecting the amplicons to pyrosequencing, the LOQ was determined to be 5% which is in line with the LOQ given by the manufacturer of the pyrosequencing instrument used.

DNA extracts from breast tissues of four healthy women were used to determine base levels of DNA methylation in non-cancerous breast tissues. In all breast tissue samples, BRCA1 and HER2/neu were unmethylated (methylation status < LOD), whereas ESR2 showed methylation levels between LOD and LOQ. PTEN was slightly methylated (methylation status < LOQ) in three breast tissue samples. Methylation levels slightly ≥ LOQ were only obtained for APC in one and for ESR1 in three samples. In case of CDKN2A exon 2, we analyzed breast tissue samples from seven healthy women. From five women, samples of left and right breast were available. In six healthy women, the methylation status of CDKN2A exon 2 was found to be < LOQ, in one woman it was even < LOD. The global methylation status was in the range from 88.4 to 90.7%. No differences were found between the tissue samples originating from the same women, neither in the gene-specific nor in the global methylation status.

Figure 3 shows the frequency of DNA methylation in the eight gene-specific regions investigated (seven promoter regions plus exon 2 of CDKN2A) in tumor, tumor-adjacent and tumor-distant tissues from breast cancer patients. Methylation levels obtained by MS-HRM analyses (APC, BRCA1, CDKN2A, ESR1, HER2/neu and PTEN) were divided into three subclasses (< LOD, < LOQ and ≥ LOQ), those obtained by PSQ (ESR2) into two subclasses (< LOQ and ≥ LOQ). Exon 2 of CDKN2A was the only region found to be methylated (methylation status ≥ LOD) in each (18/18) of the tumors. In contrast, the promoter of CDKN2A was only methylated in 28% (5/18). In addition to CDKN2A exon 2, the promoters of APC (83%, 15/18), ESR1 (83%, 15/18) and ESR2 (71%, 12/17) were frequently methylated in tumors. Figure 3 indicates that gene-specific methylation (methylation status ≥ LOD) was almost as frequent in tumor-adjacent and tumor-distant tissues as in tumors. However, exon 2 of CDKN2A more frequently showed a methylation status ≥ LOQ in tumor (94%, 17/18) than in tumor-adjacent (50%, 9/18) and tumor-distant (56%, 10/18) tissues.

Figure 4 shows the distribution of gene-specific and global methylation levels in tumor, tumor-adjacent and tumor-distant tissues as well as in normal tissues from healthy women. With some exceptions, the promoter regions showed rather low methylation levels. The promoters of ESR1, HER2/neu and PTEN were almost unmethylated in each of the tissue samples investigated. Methylation levels ≥ 25% were only obtained for the promoters of APC in 44% (8/18), BRCA1 and CDKN2A in 11% (2/18) and ESR2 in 12% (2/17) of the tumor tissues and in addition for ESR2 in 6% (1/18) of the tumor-distant tissues. In contrast, exon 2 of CDKN2A showed a methylation status ≥ 25% in 78% (14/18) of the tumors and 11% (2/18) of the tumor-adjacent and tumor-distant tissues.

The methylation status of CDKN2A exon 2 in tumors was significantly higher than in tumor-adjacent (p < 0.001) and tumor-distant tissues (p < 0.001) from the same patients and in normal breast tissues from healthy women (p < 0.001) (Fig. 4). In addition, the methylation status in tumor-adjacent (p = 0.002) and tumor-distant tissues (p = 0.005) was significantly higher than that in normal breast tissues from healthy women. Significant differences were also found between the methylation status of APC in tumors and that in tumor-adjacent tissues (p = 0.001), tumor-distant tissues (p = 0.001) and breast tissues from healthy women (p = 0.003) (Fig. 4). In case of ESR2, tumor-adjacent (p < 0.001) and tumor-distant tissues (p < 0.001) but not the tumors showed significantly higher methylation levels than normal breast tissues from healthy women. Tumors showed a significantly lower global methylation extent than tumor-adjacent (p = 0.046) and tumor-distant tissues (p = 0.046) and breast tissues from healthy women (p < 0.001). Significant differences were also found between the global methylation extent in tumor-adjacent (p < 0.001) and tumor-distant tissues (p < 0.001) and that in normal breast tissues.

For the following genes, a correlation was found between the methylation levels in tumors and those in tumor-adjacent tissues: CDKN2A (promoter) (r = 0.927, p < 0.001); BRCA1 (r = 0.878, p < 0.001); APC (r = 0.706, p = 0.001), ESR1 (r = 0.545, p = 0.019) and ESR2 (r = 0.543, p = 0.024). In addition, we found a correlation between the methylation levels in tumors and tumor-distant tissues for BRCA1 (r = 0.826, p < 0.001) and ESR1 (r = 0.555, p = 0.017). In case of BRCA1 (r = 0.994, p < 0.001), exon 2 of CDKN2A (r = 0.651, p = 0.003) and ESR2 (r = 0.580, p = 0.015), methylation levels in tumor-adjacent tissues correlated with those in tumor-distant tissues.

For the global methylation status, correlations were found between tumor and tumor-adjacent tissues (r = 0.737, p = 0.001), tumor and tumor-distant tissues (r = 0.679, p = 0.003) as well as tumor-adjacent and tumor-distant tissues (r = 0.778, p < 0.001).

Methylation levels of the BRCA1 promoter strongly correlated with those of the CDKN2A promoter in tumors (r = 0.990, p < 0.001), tumor-adjacent tissues (r = 0.920, p < 0.001) and tumor-distant tissues (r = 0.963, p < 0.001). In tumors, the methylation levels of ESR2 correlated with those of APC (r = 0.614, p = 0.009).

In addition, correlations were found with genes the methylation levels of which we had determined in previous studies [24, 31]. The methylation status of ESR2 in tumors correlated with that of ABCB1 (r = 0.614, p = 0.009) and ABCG2 (r = 0.543, p = 0.030). The methylation status of APC in tumors was found to correlate with that of MGMT (r = 0.679, p = 0.003). In tumor adjacent tissues, the promoter methylation status of ESR2 correlated with that of MGMT (r = 0.704, p = 0.002) and in tumor-distant tissues, the methylation status of APC correlated with that of ABCB1 (r = 0.756, p < 0.001).

For none of the genes investigated in the present study, we found a correlation between the age of the healthy women and the DNA methylation status in their breast tissues (based on breast tissues from four healthy women; in case of CDKN2A exon 2, breast tissues from seven women were taken into account). In tumor-distant tissues from breast cancer patients, the methylation status of ESR1 was negatively correlated with the age of the patients at diagnosis (r = −0.673, p = 0.002).

Methylation levels of CDKN2A exon 2 in tumor tissues and also in tumor-distant tissues showed a weak but significant negative correlation with the proliferative activity of the tumor (r = −0.485, p = 0.041 and r = −0.498, p = 0.036, respectively). All tumors of the luminal A and luminal B subtype showed significantly higher (p < 0.001) methylation levels than the breast tissues from healthy women. Tumor of patient 12, belonging to the triple negative subtype, showed the lowest methylation status (methylation status < LOQ) of CDKN2A exon 2 among all tumors investigated. Since our set of breast tissue samples contained only one tumor of a triple negative subtype, we determined the methylation status of CDKN2A exon 2 in MDA-MB-231, a triple negative breast cancer cell line and compared it with the methylation status of the ER positive cell lines MCF-7 (luminal A) and ZR-75-1 (luminal B). In line with the tumor tissue from patient 12, the triple negative tumor cell line MDA-MB-231 showed a methylation status < LOQ, whereas the methylation status in MCF-7 and ZR-75-1 cells was 22 and 66%, respectively. A rather low methylation status was also found for the tumor from patient 18, belonging to the HER2/neu positive subtype.

Figure 5 shows the gene-specific methylation patterns of the individual tumor, tumor-adjacent and tumor-distant tissues from breast cancer patients and normal breast tissues from healthy women. In order to get a broader picture of the prevalence of pre-neoplastic DNA methylation changes in tumor-surrounding tissues in breast cancer patients, we included methylation levels that we had obtained for the same set of breast tissue samples in previous studies (CCND2, DAPK1, GSTP1, HIN-1, MGMT and RASSF1A: [24]; ABCB1, ABCC1 and ABCG2: [31]). Gene-specific methylation patterns obtained for tumor tissues were obviously more heterogeneous with regard to both the frequency of methylation in the regions investigated and the extent of methylation than those obtained for tumor-adjacent and tumor-distant tissues from the same patients and normal breast tissues from healthy women. This observation was confirmed by principal component analysis (Fig. 6). Due to incomplete data sets, promoters of PTEN and ABCC1 and breast cancer patients 1 and 18 were excluded in this evaluation. Figure 6 also demonstrates that the gene-specific methylation patterns of tumor, tumor-adjacent and tumor-distant tissues of patient 12 and tumor of patient 14 deviated from the patterns of the respective tissue specimens of other patients. As mentioned above, tumor of patient 12 belonged to the triple negative subtype, the tumor of patient 14 was of the luminal A subtype (ER status: moderately positive, PR and HER2/neu status: negative). However, expression of the estrogen receptor was only moderate (50%) and in consequence of chemotherapeutic treatment, expression of the estrogen receptor was found to be decreased to 10%. These facts indicate that the tumor of patient 14 was not a typical representative of the luminal A subtype.

Since hypermethylation of the promoter region of genes frequently leads to transcriptional silencing, for each of the breast tissue samples we calculated the number of promoters showing a methylation status ≥ 5%. Due to incomplete data sets, promoters of PTEN and ABCC1 and breast cancer patients 1 and 18 were again excluded. Figure 7 shows the distribution of the frequency of promoters with a methylation status ≥ 5% in tumors, tumor-adjacent and tumor-distant tissues and normal breast tissues from healthy women. In tumor tissues, up to seven (out of 14) and in tumor-adjacent and tumor-distant tissues, up to six (out of 14) promoters showed a methylation status ≥ 5%, whereas in breast tissues from healthy women at most two promoters had a methylation status ≥ 5%. Figure 7 also indicates that in tumors most frequently six, in tumor-adjacent tissues two to four, in tumor-distant tissues most frequently two and in breast tissues from healthy women most frequently only one promoter had a methylation status ≥ 5%.