Purpose: In an effort to additionally determine the global patterns of CpG island hypermethylation in sporadic breast cancer, we searched for aberrant promoter methylation at 10 gene loci in 54 primary breast cancer and 10 breast benign lesions.
Experimental Design: Genomic DNA sodium bisulfate converted from benign and malignant tissues was used as template in methyl-specific PCR for BRCA1, p16, ESR1, GSTP1, TRβ1, RARβ2, HIC1, APC, CCND2, and CDH1 genes.
Results: The majority of the breast cancer (85%) showed aberrant methylation in at least 1 of the loci tested with half of them displaying 3 or more methylated genes. The highest frequency of aberrant promoter methylation was found for HIC1 (48%) followed by ESR1 (46%), and CDH1 (39%). Similar methylation frequencies were detected for breast benign lesions with the exception of the CDH1 gene (P = 0.02). The analysis of methylation distribution indicates a statistically significant association between methylation of the ESR1 promoter, and methylation at CDH1, TRβ1, GSTP1, and CCND2 loci (P < 0.03). Methylated status of the BRCA1 promoter was inversely correlated with methylation at the RARβ2 locus (P < 0.03).
Conclusions: Our results suggest a nonrandom distribution for promoter hypermethylation in sporadic breast cancer, with tumor subsets characterized by aberrant methylation of specific cancer-related genes. These breast cancer subgroups may represent separate biological entities with potential differences in sensitivity to therapy, occurrence of metastasis, and overall prognosis.
Breast cancer is the most common malignancy in women and represents 18% of all female cancers. The incidence of breast cancer increases with age, and in Western countries the disease is the single most common cause of death among women aged 40–50. The natural history of breast cancer is characterized by a long duration and marked heterogeneity within and between patients. Tumors with similar histopathological appearance can follow significantly different clinical courses and show different responses to therapy. The presence of hormone receptor identifies a subset of patients responsive to endocrine therapy and with better prognosis, but even in this group the clinical outcome can be heterogeneous (1) .
Silencing of cancer-associated genes by methylation of CpG islands located at the 5′ end of many genes is thus far the main epigenetic modification identified in tumors. The distribution of the CpG dinucleotide in human genome is generally underrepresented with the exception of genomic regions that usually contain gene promoters or first exons and are almost always unmethylated in adult tissues. Hypermethylation of the CpG islands is associated with delayed replication, condensed chromatin, and inhibition of transcription initiation. A deregulation of DNA methylation is observed in cancer cells that leads to genome-wide hypomethylation and hypermethylation of CpG island associated to cancer-related genes (2) .
The global pattern of methylation event in different tumor-types was analyzed with two different approaches. Costello et al. (3) used restriction landmarks genomic scanning to determine the methylation profile of CpG islands interspersed in the genome, whereas Esteller et al. (4) used a candidate gene approach. In both cases they found that hypermethylation of some of the CpG islands is shared by multiple tumor types, whereas others are methylated in a tumor type-specific manner (3 , 4) . These data suggest that aberrant patterns of methylation are not random and that intra- and intertumor type heterogeneity may display distinguishable methylation subtype.
We have used a candidate gene approach to investigate the methylation profile of primary breast cancer and benign breast lesions. We chose genes that belong to a critical pathway either involved in breast cancer tumorigenesis or methylated in other tumor types. ESR1, RARβ2, and TRβ1 belong to the nuclear receptor superfamily and are ligand-mediated transcription factors (5) . The protein product of the APC gene is an important component of the wingless-type mouse mammary tumor virus integration site signaling pathway, which inactivates β-catenin (6) . CDH1 encodes for a cell surface molecule with a main role in maintaining cell-cell adhesion in epithelial tissues (7) . p16 and CCND2 are important cell cycle checkpoints (8) . GSTP1 is implicated in the detoxification pathway of xenobiotics and chemotherapeutic agents (9) , whereas BRCA1 is critically involved in cellular response to DNA damage and in the maintenance of genome integrity during DNA synthesis (10) . Finally, HIC1 is a transcription factor with an important role in embryonic development (11) . For each of these individual genes it has been reported previously that promoter CpG islands are methylated in cancer cells and expression of the gene is silenced (4) . For the RARβ2 and the ESR1 genes the silencing could be partially relieved in vitro by demethylation of the promoter region after treatment with 5-deoxy-azacytidine (12, 13, 14) .
Our analysis indicates that aberrant promoter methylation does not occur randomly in breast cancer. Indeed, cancer-related genes are targeted in a specific fashion with a direct correlation among ESR1, CDH1, TRβ1, GSTP1, and CCND2 promoter methylation and an inverse correlation between RARβ2 and BRCA1 methylation. Similar patterns were also detected in breast benign lesions, suggesting that promoter hypermethylation of specific genes is an early event in breast cancer development.
MATERIALS AND METHODS
Specimens and DNA Extraction.
Fifty-four invasive primary breast cancer tumors (44 ductal and 10 lobular) were collected at the Area of Pathology, University Campus BioMedico (Rome, Italy), at the Clinical Experimental Oncology Laboratory, National Cancer Institute (Bari, Italy), and at the Department of Pathology The Johns Hopkins University, (Baltimore, MD) with Institutional Review Board approval. Thirty-four of the samples were from fresh-frozen specimens, whereas the remaining 20 samples were paraffin embedded. Ten benign breast lesions were obtained as paraffin-embedded slides from the Department of Pathology University “Federico II” (Naples, Italy). The histological type and grade of the tumors were classified according to the World Health Organization criteria. Fresh-frozen samples were carefully dissected on a cryostat so that the tumor samples contained at least 70% of neoplastic cells (15) . Paraffin sections were processed as described previously (16) .
Genomic DNA sodium bisulfate conversion of the unmethylated cytosine residue to thymidine was performed as described previously (17) . The bisulfate converted DNA was used as a template for methyl-specific PCR using primers specific for either the methylated or the modified unmethylated sequences (18) . Primer sequences for ESR1 (19) , p16 (18) , GSTP1 (20) , BRCA1 (21) , RARβ2 (22) , TRβ1 (23) , HIC1 (24) , APC (6) , CCDN2 (25) , and CDH1 (18) were as described previously. PCR reactions were carried out in a total volume of 40 μl, containing 3 μl of modified DNA, 300 ng of each of the primers, 4.5 μl of PCR buffer (16) , 1.25 mm deoxynucleotide triphosphate (Life Technologies, Inc., Rockville, MD), and 0.3 μl PlatinumTaq DNA polymerase (Life Technologies, Inc.; Ref. 19 ). PCR conditions were as follows: 35 cycles at 95° for 1 min, 54°C-64°C depending on primer pairs for 1 min, and 72°C for 1 min. For each PCR reaction CpGenome Universal Methylated DNA (Serologicals Corp., Norcross, GA) was used as positive control. About half of the reaction volume was run on Tris-borate EDTA 10% Criterion Gels (Bio-Rad Laboratories Inc., Hercules, CA) stained with ethidium bromide and visualized at UV light.
Frequencies with which other loci were methylated when a particular locus was either methylated or unmethylated were compared by the Mann Whitney U test as reported previously (26) . Comparison of methylation frequencies between two loci was performed using the Fisher exact test. Mean values were compared using the ANOVA. Kaplan-Meier analysis was used to evaluate differences in survival between patients groups. Differences were considered statistically significant for P < 0.05.
Aberrant Promoter Methylation in Breast Cancer and Benign Lesions.
We searched for aberrant promoter methylation at 10 gene loci (ESR1, BRCA1, TRβ1 p16, HIC1, RARβ2, CCND2, APC, GSTP1, and CDH1) in 54 primary breast cancers and 10 benign lesions (Fig. 1)⇓ . Aberrant promoter methylation for at least 1 gene was found in 46 of the 54 (85%) breast cancers and in 7 of the 10 (70%) benign lesions.
In malignant tumors, all of the gene loci tested showed methylation in significant proportion of the samples (Table 1)⇓ . The highest frequencies of aberrant promoter methylation were found for HIC1, ESR1, CDH1, TRβ1, APC, and RARβ2 loci, methylated in 48%, 46%, 39%, 28%, 28%, and 20% of the cases, respectively. Lower frequencies were detected for p16, BRCA1, GSTP1, and CCND2 (18%, 17%, 13%, and 11% respectively; Table 1⇓ ). The analysis of benign lesions showed minor differences in promoter hypermethylation frequencies for each of the individual genes as compared with malignant lesions (Table 2)⇓ . The only exception was the CDH1 gene, for which the difference in frequency between benign (0%) and malignant (39%) tumors was statistically significant (P < 0.02).
Analysis of Methylation Distribution in Malignant Tumors.
Of the 54 malignant tumors 29 (54%) were methylated at 3 or more loci (mean age ± SE, 56.56 ± 2.82); 11 tumors (20%) displayed 2 methylated loci (mean age ± SE, 60.10 ± 3.31); and 14 tumors (26%) showed 1 or 0 methylated gene (mean age ± SE, 54.21 ± 3.04; Table 1⇓ ). The analysis of methylation distribution in the 46 breast cancers with at least 1 methylated gene demonstrated a statistically significant association between methylation of the ESR1 promoter, and methylation at CDH1 (P = 0.001), TRβ1 (P = 0.02), GSTP1 (P = 0.03), and CCND2 (P = 0.007) loci (Fig. 2)⇓ . BRCA1 promoter aberrant methylation was negatively associated with methylation at the RARβ2 locus. Of the 20 breast cancers with methylation in either BRCA1 (n = 9) or RARβ2 (n = 11), none showed methylation of both loci (P = 0.03). No differences in methylation frequency were found between paraffin-embedded and fresh-frozen tissues; this excludes that the differences in methylation distribution might be due to technical artifact.
There was no apparent association between methylation distribution phenotypes and grade, stage, or lymph node status, and no correlations were found for single loci. Kaplan-Meier analysis of patients survival in the 23 cases for which follow-up data were available (mean follow up, 85.87 ± 8.21 months) showed an inverse correlation between high level of methylation and disease-free survival, but it did not reach statistical significance (data not shown).
Correlation between ESR1 Aberrant Methylation and Protein Expression by Immunohistochemistry.
ESR1 promoter methylation was studied using three primer pairs, ER3, ER5, and ER6, from Lapidus et al. (19) , covering CpG between position +225 and +529 from the transcription start site. Methylation for all three of the primer pairs was found in 18 of the 25 ESR1 methylated breast cancers. In 7 cases methylated reactions were positive for 2 (n = 5) or 1 (n = 2) of the primers. Data from estrogen receptor routine immunohistochemistry were available for 48 of the 54 breast cancers. Of these tumors, 27 were estrogen receptor positive, 5 were weakly positive, and 16 were estrogen receptor negative. ESR1 CpG island promoter methylation was found in 13 estrogen receptor-positive tumors, 4 weakly positive tumors, and 4 estrogen receptor-negative tumors (21 of the 25 ESR1 methylated tumors). Of the unmethylated tumors, 14 were estrogen receptor positive, 1 weakly expressing the receptor, and 12 were estrogen receptor negative (27 of the 29 ESR1 unmethylated tumors).
In an attempt to better understand the epigenetic events that lead to breast cancer development and progression, we have examined the methylation status of multiple loci in primary breast cancer and breast benign lesions. Methylation for at least 1 gene was found in 85% of the breast cancer and in 70% of the breast benign lesions. However, we found significant differences in methylation frequencies among the malignant tumors analyzed. Approximately half of those tumors showed >3 methylated genes, whereas the remaining tumors displayed lower level or no methylation at all (Table 1)⇓ . These differences did not correlate with age of the patient at the diagnosis suggesting that they are not due to age-related methylation changes but are expression of different levels of deregulation of the mechanisms that protect against CpG island hypermethylation.
A more detailed analysis of our result revealed that CpG promoter hypermethylation does not occur randomly in breast cancer. Nass et al. (27) have reported previously coincident methylation of the CDH1 and ESR1 genes. Our results confirm this association, but we also identified 3 other genes, GSTP1, CCND2, and TRβ1, with a statistically significant association with ESR1 promoter hypermethylation. This association was independent from the overall frequency of promoter hypermethylation suggesting that it represents a molecular feature of a subset of breast tumors. The absence of correlation for other genes frequently methylated, such as HIC1 (48%) and APC (28%; Table 1⇓ ), additionally supports this hypothesis.
Another interesting result of our molecular survey is the inverse correlation between BRCA1 and RARβ2 aberrant promoter methylation. When we reviewed the literature, this result was consistent with a study by Esteller et al. (28) that analyzed 106 sporadic breast tumors at 5 gene loci (p16, CDH1, RARβ2, GSTP1, and BRCA1). Of these tumors, 68 had at least 1 methylated gene, of which 37 displayed methylation of either BRCA1 or RARβ2. Only 3 of those 37 breast cancers showed a concomitant methylation of both gene loci (28) . When results from both studies are considered, ∼34% of the tumors show hypermethylation of either BRCA1 or RARβ2, but only 2% of them present concurrent methylation of the two loci. These data additionally support the hypothesis that CpG island aberrant promoter methylation does not occur randomly and suggest the existence of specific selection process targeting key tumor suppressor genes.
In the 10 benign lesions we found substantially the same methylation pattern as in malignant tumors suggesting that inactivation of tumor suppressor genes by methylation represents an early event in breast cancer carcinogenesis. Jeronimo et al. (29) reported similar results in a series of breast benign lesions analyzed at 5 gene loci. In our series, the only difference in methylation frequency between benign and malignant tumors was found for the CDH1 gene that has been implicated in tumor progression (27) . Aberrant promoter methylation was detected in cases of atypical ductal hyperplasia, a condition considered as premalignant, as well as in cases of fibrocystic disease. In this latter group cases with and without promoter hypermethylation were histologically indistinguishable. It would be interesting to test in retrospective studies whether fibrocystic disease patients with and without promoter hypermethylation may differ for the risk of developing malignant tumors.
Expression of the estrogen receptor protein as determined by immunohistochemistry is a predictive marker for response to hormone therapy. However, up to one third of breast carcinomas lack estrogen receptor at time of the diagnosis and a proportion of cancers that are initially estrogen receptor positive lose receptor during progression (30) . We found promoter aberrant methylation in only a subset of tumors negative for estrogen receptor by immunostaining. It is known that promoter methylation is not the only mechanisms required for gene silencing. In breast cancer cell lines, together with aberrant methylation histone deacetylation was also detected, suggesting that both mechanisms are required to reach estrogen receptor silencing (12, 13, 14) . However, there is another issue to address: the genomic organization of the ESR1 gene is much more complex than expected (31) . In the past 10 years 6 other promoters localized upstream to the first identified in 1988 (32) were found. Thus, ESR1 expression most likely is the result of the interplay between these promoters and their transcriptional regulators. On the other side, we demonstrate ESR1 hypermethylation in tumors showing ESR1 expression. Routine immunohistochemistry classifies protein expression as a percentage of positively stained tumor cells. In our series, estrogen receptor expression ranged from 40% to 95%, thus it is possible that tumors may contain subclones with some level of ESR1 promoter methylation. It remains to be determined in larger randomized case-control trials and using quantitative methylation assay whether these subclones might be responsible for the observed reduction of ERα expression with consequent acquisition of hormone resistance in breast cancer patients (30) .
Our results shed additional light on the significance of epigenetic modification in breast cancer and delineate distinct breast cancer subsets, involving aberrant methylation of specific tumor suppressor genes. Recent studies have found a correlation between CpG islands promoter hypermethylation and clinical parameters such as prognosis and response to therapy (33 , 34) . Because epigenetic modifications are potentially reversible, the identification of cancer subsets with different patterns of methylation may result in important consequence for breast cancer patient management.
Grant support: Ministero della Salute Istituto di Ricovero e Cura a Carattere Scientifico Grant (RC0302TG12) and partially supported by Italian Ministry of Health Project ICS030.1 and Project RF2003 Emilia Romagna.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Paola Parrella, Laboratorio di Terapia Genica ed Oncologia, Istituto di Ricovero e Cura a Carattere Scientifico, “Casa Sollievo della Sofferenza,” Viale Padre Pio, 71013 San Giovanni Rotondo (FG), Italy. Phone: 39-0882-416262; Fax: 39-0882-416260; E-mail:
- Received March 22, 2004.
- Revision received May 3, 2004.
- Accepted May 10, 2004.