Methylated genes in breast cancer

Abstract

Background: Hundreds of hypermethylated genes have been described in breast cancer, yet the nature and contribution of these genes in their methylated state to overall risk and prognosis is under-characterized in non-sporadic breast cancers. We therefore compared associations of DNA methylation with tumor stage, hormone/growth receptor status, and clinical outcomes in a familial breast cancer cohort. Because few previous methylation studies have considered the oncogenic or tumor suppressor properties of their gene sets, this functional status was included as part of our correlative analysis. Results: We found methylation of oncogenes was associated with better prognostic indicators, whereas tumor suppressor gene methylation was associated with a more severe phenotype in women that were either HER2+ or lymph node positive at diagnosis, and/or tended to recur or develop distant metastases. For example, the methylation of the tumor suppressor gene APC was strongly associated with a specific subset of tumors that were both ER+ and HER2+, while methylation of the TWIST oncogene was associated with breast cancers that did not metastasize. Methods: This was a retrospective, hospital-based study of n = 99 archival breast tumors derived from women with a germline genetic BRCA1 or BRCA2 mutation and/or familial breast cancer history. DNA methylation was quantified from formalin fixed, paraffin embedded tumors using the established protocol of quantitative multiplex-methylation specific PCR (QM-MSP). Non-parametric statistics were used to analyze candidate gene methylation in association with clinical outcomes. Conclusion: We report several novel, positive associations between percent methylation of the APC, RASSF1A, TWIST, ERα, CDH1, and Cyclin D2 genes and key variables such as tumor stage, hormone and growth receptor status, and a history of recurrent or metastatic disease. Our data suggest the potential utility of parsing gene methylation by functional status and breast tumor subtype.

See commentary:

Using methylation analysis to assess tumor heterogeneity in familial breast cancer

This article is referred to by:

Using methylation analysis to assess tumor heterogeneity in familial breast cancer

Introduction

We compared associations of DNA methylation with tumor features and clinical variables in n = 99 women with breast cancer. Hundreds of hypermethylated genes have been described in breast cancer, yet the nature and contribution of these genes in their methylated state to overall risk and prognosis is under-characterized in non-sporadic breast cancers. The inherent heterogeneity of breast cancer and its myriad subtypes pose a challenge in any research undertaking. In an effort to limit heterogeneity, we narrowed the inclusion criteria for this study to an at risk cohort of women with the rationale that distinct epigenetic factors common to those with familial breast cancers and/or known BRCA1 or BRCA2 genetic mutations, are at differing frequencies than in sporadic breast cancers.1–4 Recent studies suggest that specific gene methylation patterns are associated with distinct histopathological categories, raising the possibility that such patterns may have prognostic value.4–6 This study sought to further refine these categories by testing associations of candidate gene methylation with tumor features and clinical outcomes such as recurrence and metastasis.

How complex epigenetic processes occur, evolve and interact with each other at the DNA, RNA and chromatin level are vastly complex and not well understood. Although there is debate about the nature of these epigenetic phenomena and whether they occur in a random or targeted manner,7–11 there is nevertheless evidence that global DNA methylation pathways may “overshoot their biological targets,”11 and cause some gene methylation to occur secondary to upstream effects. Therefore, this may partly explain the occurrence of gene hypermethylation in the absence of changes in gene expression, but also underscores the potential utility of secondarily methylated genes as “surrogate” biomarkers when their presence is consistently predictive of specific clinical outcomes.

Using quantitative multiplex methylation specific PCR (QM-MSP) on a real-time platform, we quantified DNA methylation for ten genes (APC, RASSF1A, TWIST, ESR1, CDH1, Cyclin D2, BRCA1, RARβ, BRCA2 and HIN1), in n = 99 tumors derived from women with a family history of breast cancer with and without deleterious BRCA1 or BRCA2 genetic mutations. The ten genes were chosen because they were either strong biological candidate genes for breast carcinogenesis, their known or suspected involvement in various aspects of cellular regulation and development appears to be suppressed upon methylation, or because previous studies showed they were methylated in breast tumor cell lines or breast tissues. We report associations of percent methylation for six genes with distinct tumor characteristics and clinical outcomes that, depending on their tumor suppressor or oncogenic properties, may confer either a protective role or an increased risk for recurrence or metastasis.

Results

Frequency distributions of demographic and clinical variables.

As expected in this cohort, there was an earlier average age of breast cancer onset of 44.8 years as compared to the reported mean (54 years) and median (61 years) age of onset in the general population.12,13 Hormone receptor IHC tumor staining and lymph node status of the n = 99 breast cancer cases were similar to the general population; slightly over two thirds were estrogen/progesterone receptor positive, and approximately two thirds were lymph node and HER2 negative at diagnosis (Table 2). These tumor type distributions by histology have been shown to correlate highly with breast cancer subtypes by gene expression profiling.14 Frequency distributions for the remaining variables used in the statistical analyses are provided in Table 2. Table 3 shows the number of tumors of n = 99 total with methylation equal to or above the 5% methylation threshold. Additionally, the average percent methylation per tumor as well as the range of percent methylation values are provided in Table 3.

Associations of percent methylation with clinical variables.

To examine the contribution of % methylation for each gene to a selected set of histological and clinical variables, box plots of median percent methylation, based on the log of percent methylation + 1 or log (% M + 1), were constructed by tumor stage, recurrent, metastatic or bilateral breast disease, age at diagnosis, BRCA mutation status, and by ER, HER2 and lymph node status (Fig. 1A–J). Percent methylation varies widely between genes and often has a non-normal distribution, with many values skewed toward zero for multiple genes in a particular assay. The standard statistical correction for such skewed distribution is to log transform the data. Accordingly, we log transformed methylation values for each gene in our study with a zero correction factor of log (% M + 1). Because log of zero is undefined, adding 1 was necessary to obtain finite values.

For each gene in the study, differences in percent methylation (as measured by the median log (% M + 1), between these variables were tested using the Wilcoxon rank sum test (Fig. 1A–J). The following variables were excluded from the analysis as none of the ten genes were significantly methylated by age (<45, ≥45 yo), bilateral breast disease (yes, no), BRCA1 or BRCA2 germ-line mutation status (positive/negative: unknown test results and variants of unknown significance were not included in the model), or progesterone receptor (PR) status. In summary, Figures 1A–J are box plot comparisons of ER, HER2, lymph node status, tumor stage and recurrent or metastatic disease variables by percent methylation for the following ten genes: ERα, TWIST, Cyclin D2, CDH1, APC, RASSF1A, HIN1, RARβ, BRCA1 and BRCA2, respectively.

Primer selection for quantitative methylation assays.

QM-MSP was carried out on DNA extracted from formalin fixed breast tumors using the primer and probe sequences listed in Table 1. Given the phenomenon of field cancerization effects reported in histologically “normal” tissues adjacent to tumors, and the observation that even microdissected tumors will have some percentage of normal cells embedded therein,15,16 these primer and probe sets were previously tested in normal breast parenchyma derived from healthy women without a breast cancer history and showed very low to no methylation.17–19

The classical approach in DNA methylation studies has been to consider gene hypermethylation when linked to loss of gene expression as a modified version of Knudson's Two Hit Hypothesis.20 However, not all promoter gene methylation is associated with silenced or decreased gene expression.21–23 Moreover, gene hypermethylation in the absence of changes in expression, may nevertheless be useful as a biomarker if consistently found to be predictive of clinical outcomes such recurrence, metastasis and survival. For example, only two studies have examined the relationship between TWIST promoter hypermethylation and TWIST RNA and protein, and did not find any alteration in gene expression.24,25 Although the limited archival tumor tissues in our sample set did not allow gene expression to be assayed in tandem with methylation, with the exception of TWIST, methylation has consistently been associated with transcriptional silencing of the remaining gene loci covered by our primers.24,26–35

Discussion

The main purpose of this study was to explore the predictive contributions of DNA methylation of selected genes, together with clinical and tumor features, on breast cancer recurrence and progression. Of the percent methylation for the ten genes studied in tumor tissues, six genes; ERα, TWIST, Cyclin D2, CDH1, APC and RASSF1A (Fig. 1A–F) were preferentially methylated for at least one of the histopathological variables tested. No significant associations were found for the remaining four genes: HIN1, RARβ, BRCA1 or BRCA2 (Fig. 1G–J).

Depending on multiple factors including epigenetic regulation, downstream signaling and tissue type, these genes may possess oncogenic, tumor suppressor properties or both. For example, ERα (estrogen receptor alpha, ESR1) is a transcription factor that can act as both an oncogene and a tumor suppressor gene by its role in modifying expression of a variety of genes in a tissue specific manner as a consequence of the binding of estrogens and estrogen receptor modulators.36,37 ERα expression is silenced through methylation in a variety of tissues including lung, colon and breast and its methylation predicts hormone receptor status of breast tissues.6 Therefore, it follows that if the tumorigenic functions of ERα are modified via epigenetic pathways, ERα methylation would manifest as a protective factor. Supporting this hypothesis was our finding that ERα methylation was associated with less severe histological features. It was significantly more methylated in those breast cancers which did not recur (p = 0.03) and trended toward tumors that did not metastasize (p = 0.06). These results complement the work of Kim et al. (2004) who also found ERα methylation was not associated with more severe histological indicators such as node positive or high grade tumors.38

Similarly, the bHLH transcription factor TWIST is considered an oncogene due to its role in inhibiting apoptosis, blocking expression of other tumor control genes in response to DNA damage and promoting epithelial-mesenchymal transition.39–41 Also supporting the protective hypothesis for oncogene methylation, was our finding that TWIST was preferentially methylated in lymph node negative tumors (p = 0.06), and in breast cancer cases which did not metastasize (p = 0.006), with a trend (p = 0.1) towards very low to no methylation in high grade stage III tumors. Likewise, Cyclin D2 (CCND2), is most often considered an oncogene that acts as a crucial cell cycle regulator involved in cell growth and malignant processes.42–44 Several previous studies have shown Cyclin D2 to be methylated in invasive breast carcinoma,42–44 and we found Cyclin D2 methylation was associated (p = 0.03) with non-metastatic breast cancers. As with the other genes in our set with the exception of TWIST, hypermethylation of Cyclin D2 has consistently been linked to decreased expression.

Because CDH1 (E-cadherin), mediates normal cell to cell adhesion in epithelial cells,45 its loss of expression, either though epigenetic or other mechanisms, has been associated with more severe clinical outcomes such as lymph node invasion and metastasis.46–48 Interestingly, neither this study nor a study of sporadic breast cancer,49 found CDH1 methylation in association with lymph node metastases. Although we found CDH1 methylation was associated with HER2⁺ (p = 0.03) tumors, only two tumors in this cohort had methylation values that exceeded the 5% threshold (Table 3) and therefore this result must be viewed with caution.

As opposed to oncogene methylation, we found methylation of the tumor suppressor genes, APC and RASSF1A, was associated with more prognostically severe clinical outcomes. For example, the tumor suppressor APC (adenomatous polyposis coli), can be lost through genetic and epigenetic mechanisms, but in breast cancer, methylation frequency increases with tumor stage and size, and has been associated with a poor prognosis.50–52 Likewise, we found APC methylation was associated with ER⁺ tumors (p = 0.009), generally considered to have a better prognosis than ER-tumors, but also with more aggressive HER2⁺ tumors (p = 0.005), suggesting that APC methylation merits further study as a possible marker to identify a subclass of ER⁺ positive tumors that behave more aggressively and have poorer clinical outcomes.

The tumor suppressor RASSF1A (human RAS effector homolog), has been found to be epigenetically inactivated in lung, ovary, bladder, kidney and breast tumor tissue,53 and is methylated in approximately 60–70% of breast cancers.34,54 Complementing previous studies which show RASSF1A methylation confers a poorer prognosis,55–57 was our finding methylation in 62 of n = 99 tumors (Table 3), and significantly higher methylation with increasing tumor stage from TIS to stage III (p = 0.09, p = 0.06 and p = 0.01, respectively), with a trend (p = 0.08) towards HER2⁺ tumors and in women who were lymph node positive at diagnosis (p = 0.006). These findings replicate a previous study finding of higher RASSF1A methylation in lymph node and HER2⁺ tumors.46

Methylation of the remaining genes (HIN1, RARb, BRCA1, BRCA2), were all negative for associations with clinical and tumor features. Negative findings for the tumor suppressor genes BRCA1 and BRCA2 were not surprising given the characteristically low frequencies of BRCA2 tumor methylation in our study (Table 3) and others.58,59 Moreover, data have shown lower percent BRCA1 methylation or varied methylation levels of other candidate genes in women with BRCA germline genetic mutations,3,60 as well as varied gene methylation levels in familial breast “BRCAx” cancers negative for deleterious BRCA1 or BRCA2 genetic mutations.1,61

Both tumor heterogeneity and varied methylation findings in BRCAx or familial breast cancers have been previously documented.1,62–64 Indeed, even in our familial breast cancer cohort, there was heterogeneity by tumor type, age of onset and BRCA mutation status. For example, the majority of women (56%), had not been tested for heritable mutations in the BRCA1 or BRCA2 genes. Of the women that were tested, 22% were negative for a BRCA1 or BRCA2 genetic mutation, 12% were either BRCA1 or BRCA2 positive, 5% had a mutation or variant of unknown significance (VUS) and 5% were tested, but their BRCA test result was not reported in any of the three clinical databases abstracted for this study (Table 2). Also, 30% of the women with a confirmed deleterious BRCA genetic mutation did not have a strong family history, but rather only had either one first or second degree relative with breast cancer (data not shown).

This retrospective study was conducted using archival breast tumors and medical records abstraction dating from 1986–2004. The amount of missing data (Table 2) for HER2 IHC status reflects those tumors that were archived prior to the discovery, development or common use of the HER2 antibody in clinical pathology practice. Therefore, both negative and positive findings for associations between gene methylation and HER status must be considered in this context. Due to the finite amount of nucleic acids obtained from limited archival specimens, we were not able to perform gene expression profiling in tandem with methylation analyses to determine intrinsic (e.g., luminal A, luminal B, HER2-enriched, basal-like and normal-like),65 breast cancer subtypes. However, the study strengths include our selection of a non-sporadic breast cancer cohort and the highly quantitative QM-MSP method that allowed us to detect and set thresholds for low levels of methylation that might otherwise test differently on qualitative gel-based approaches.

In summary, although our methods include the classical approach of considering gene hypermethylation linked to loss of gene expression, we also support the merits of evaluating hypermethylated biomarkers in the context of purported oncogene or tumor suppressor function, regardless of their effects on gene expression. Also, the parsing of analyses by consideration of oncogene or tumor suppressor function, allows the exploration of a protective or risk factor role when analyzed in the context of important clinical outcomes such as recurrence or metastasis.

Taken as a whole, these findings illustrate the existence of specific patterns of gene methylation in association with distinct combinations of histopathological features. Such results emphasize the potential utility of characterizing methylation patterns in order to better define breast tumors such as the luminal, HER2, normal like and basal like subtypes characterized by gene expression profiling. Future studies of DNA methylation and breast cancer will assay gene expression in tandem with other clinical correlates and outcomes.

Materials and Methods

Population and sample procurement.

This archival, retrospective hospital-based study tested associations of tumor DNA methylation with clinical variables and tumor features in a cohort of n = 99 women with breast cancer. Inclusion criteria for at risk status were having a known deleterious BRCA1 or BRCA2 germline gene mutation, early age of breast cancer onset of 40 years or younger and/or a moderate to strong family history of breast cancer depending on the number of first (FDR), second (SDR) and third (TDR) degree relatives with breast cancer and their respective ages of cancer onset. Approximately one half of the cases had a strong family history ranging from three to eight members with breast cancer. Cohort demographics included 61% White, 20% Ashkenazi, 15% Black, 3% Middle-eastern and 1% Hispanic women. Archival, formalin fixed, paraffin embedded tumors (FFPEs) and clinical data for all women in the study were collected subsequent to HIPAA and IRB approval.

H&E slides from archival FFPE tissue blocks for each case were reviewed by a diagnostic breast pathologist (R. Vang, MD), who provided the breast cancer diagnosis, tumor type and grade. During slide review, the percentage of tumor epithelium relative to stroma/adjacent histologically normal tissue was estimated, with values of tumor epithelium achieving as much as 85% per sample. Once the presence of tumor was confirmed by H&E, a 5 µm section was cut from each tumor block and used for DNA extraction. Due to the unavoidable and variable admixture of each tumor with adjacent histologically “normal” epithelium and stroma, we previously assayed methylation in normal breast tissues as a comparison, and found very low to no methylation for the 10 loci in this study.17–19

DNA extraction from tumors, cell lines and human sperm.

FFPE sections were de-paraffinized in xylene for 20 minutes, scraped from the slide and extracted in 30 µl TNES (10 mM Tris, pH 8.0, 150 mM NaCl, 2 mM EDTA, 0.5% SDS) containing 40 µg of proteinase K for 5 h at 52°C. After Proteinase K treatment, cells were heat inactivated at 99°C for 10 min and centrifuged at 16,000x g for 10 min. 13.5 of the 30 µl supernatant was used directly as a source of DNA for sodium bisulfite (NaBi) treatment, performed according to Fackler et al. (2004).56 NaBi treatment converts non-methylated cytosine residues to uracil (later replicated as thymidine during PCR), whereas methylated cytosines remain unchanged. Such conversion of the original sequence allows distinct primer sets to be made that are specific to unmethylated and methylated CG sequences respectively for each sample. DNA was extracted from cell lines with phenol-chloroform,66 and from human semen using the PUREGENE DNA Purification Kit (Gentra Systems, Minneapolis, MN), according to the manufacturer's instructions and stored at −20°C.

Positive and negative methylation controls.

Breast carcinoma cell line MDA-MB231 was obtained from American Type Culture Collection (www.atcc.org, Manassas, VA) and cultured as directed. Human sperm was obtained from a healthy volunteer. Many genes are highly methylated in MDA-MB231 cells and therefore this cell line is used as a positive methylation control both before and after treatment with a DNA methyltransferase. Human sperm DNA (HSD) has been shown to have a lower degree of genomic methylation as compared to other somatic cells,67 and has therefore been used as a negative, unmethylated reference control in methylation studies of tumor control genes.68 HSD treated with SssI methyltransferase (New England Biolabs Inc., Beverly, MA) was also used as 100% methylation positive control for the 10 candidate genes assayed in this study. Positive methylation controls: 2 µg of either MDA-MB231 or HSD was incubated with S-adenosyl-L-methionine and 1 unit of SssI methyltransferase (New England Biolabs, Beverly, MA) for 1.5 h at 37°C according to manufacturers instructions prior to sodium bisulfite treatment.

Quantitative multiplex—methylation specific polymerase chain reaction.

Quantitative methylation analysis was performed on sodium bisulfite treated tumor DNAs and controls with the QM-MSP real-time method using “protocol C” as described in Swift-Scanlan et al. (2006).19 Both methylated and unmethylated targets for a single gene were amplified simultaneously in the same well using FAM and VIC probes, respectively on the ABI 7900 real-time platform (Applied Biosystems, Foster City, CA).19 The following controls were included on each 96 well real-time plate: (1) a 100% methylation control using MDA-MB-231 DNA or human sperm DNA treated with SssI methyltransferase; (2) human sperm DNA methylation negative controls, (3) nontemplate negative reaction controls using water as a template in the multiplex reaction, (4) internal sample controls of multiplex reactions (randomly chosen from cases and controls), repeated on a different day and on a different reaction plate and (5) internal standard controls in the form of an 80,000 methylated copy number and 1% methylation control standard run on each 96 well real-time reaction plate. Primer and probe sequences are listed in Table 1.

Statistical analysis.

We employed non-parametric analysis to examine differences in percent methylation (% M) for each tumor by clinical outcomes and tumor features. Box plots of percent methylation (as measured by median log [% M + 1]), were constructed by estrogen receptor (ER) and HER2 immunohistochemical (IHC) tumor stain results, nodal status at diagnosis, tumor stage (TNM system), and history of breast cancer recurrence or metastatic disease. A boxplot is a graphical display of the how the data are distributed and includes the median center, spread and skewness of the methylation values. The vertical height of the box extends from the 25–75^th percentile methylation values (interquartile range or IQR), with the median (50th percentile) denoted by the horizontal line within the box. The whiskers of the plot extend to the data points that are between −1.5 and 1.5*IQR. Individual points beyond the whiskers are data points that are considered outliers. Differences in percent methylation between variables were tested with Wilcoxon rank sum tests. It was observed prior to statistical testing that % M, and cumulative % M values were skewed toward zero. Therefore, these data were log transformed prior to analysis. Genetic, epigenetic and histopathological variables are described in Table 2. Percent methylation for each gene was calculated as the total (methylated copies of gene/[unmethylated + methylated copies of gene] × 100).

Abbreviations

QM-MSP	=	quantitative multiplex methylation specific PCR
IHC stain	=	immunohistochemical stain
% M	=	percent methylation

Figures and Tables

Figure 1 Percent methylation* by ER, HER2, lymph node status, tumor stage and history of recurrent or metastatic disease. *Percent methylation is based on the log of percent methylation + 1 or log (% M + 1). Percent ERα (A), TWIST (B), CyclinD2 (C), CDH1 (D), AP C (E), RASS F1A (F), HIN1 (G), RARβ (H), BRCA1 (I), BRCA2 (J) methylation by ER, HER2, lymph node status, tumor stage and history of recurrent or metastatic disease.

Table 1 DNA primer and probe sequences used in steps 1 + 2 of QM-MSP

	Primer		Primer		Amplimer
Gene	abbr.	Forward primer (5′ → 3′)	abbr.	Reverse primer (3′ a 5′)	size (bp)
APC	Ext:	AAA ACC CTA TAC CCC ACT AC	Ext:	GGT TGT ATT AAT ATA GTT ATA TGT	163
APC	FM:	AAT ACG AAC CAA AAC GCT CCC	RM:	TAT GTC GGT TAC GTG CGT TTA TAT	78
APC	FU:	TAA ATA CAA ACC AAA ACA CTC CC	RU:	GTT ATA TGT TGG TTA TGT GTG TTT	79
APC	MP:	CCC GTC GAA AAC CCG CCG ATT A	UP:	TTC CCA TCA AAA ACC CAC CAA TTA AC
BRCA1	Ext:	TAT TTT GAG AGG TTG TTG TTT AG	Ext:	AAA CAT CAC TTA AAC CCC CTA T	241
BRCA1	FM:	TTT CGT GGT AAC GGA AAA GCG	RM:	CCG TCC AAA AAA TCT CAA CGA A	86
BRCA1	FU:	TGG TAA TGG AAA AGT GTG GGA A	RU:	CCC ATC CAA AAA ATC TCA ACA AA	73
BRCA1	MP:	CTC ACG CCG CGC AAT CGC AAT TT	UP:	CTC ACA CCA CAC AAT CAC AAT TTT AAT
BRCA2	Ext:	GTT GGG ATG TTT GAT AA G GAA T	Ext:	ATC ACA AAT CTA TCC CCT CAC	262
BRCA2	FM:	TTT GAT TTT CGG GTG GTG CGT	RM:	TCA AAA ACT CGC GCC ACA AAC	82
BRCA2	FU:	ATT TTT GGG TGG TGT GTG TGT T	RU:	TCA AAA ACT CAC ACC ACA AAC C	78
BRCA2	MP:	AAC CAC GTA ACG CCG TAA CGC GA	UP:	AAC CAC ATA ACA CCA TAA CAC AAC AC
CDH1	Ext:	TAA TTT TAG GTT AGA GGG TTA T	Ext:	AAC TCA CAA ATA CTT TAC AAT TC	214
CDH1	FM:	AGG GTT ATC GCG TTT ATG CGA G	RM:	ACA ACC AAT CAA CAA CGC GAA C	82
CDH1	FU:	GAG GGT TAT TGT GTT TAT GTG AG	RU:	CAC AAC CAA TCA ACA ACA CAA AC	84
CDH1	MP:	CCC AAA ACG AAA CTA ACG ACC CGC	UP:	CTC CCC AAA ACA AAA CTA ACA ACC CAC
CyclinD2	Ext:	TAT TTT TTG TAA AGA TAG TTT TGA T	Ext:	TAC AAC TTT CTA AAA AAT AAC CC	286
CyclinD2	FM:	TTT GAT TTA AGG ATG CGT TAG AGT ACG	RM:	ACT TTC TCC CTA AAA ACC GAC TAC G	83
CyclinD2	FU:	TTA AGG ATG TGT TAG AGT ATG TG	RU:	AAA CTT TCT CCC TAA AAA CCA ACT ACA AT	78
CyclinD2	MP:	AAT CCG CCA ACA CGA TCG ACC CTA	UP:	AAT CCA CCA ACA CAA TCA ACC CTA AC
ERα	Ext:	TTG GAG GTT TGG GAG TTT AG	Ext:	ACA ATA AAA CCA TCC CAA ATA C	298
ERα	FM:	AGG GCG TTC GTT TTG GGA TTG	RM:	CGC CGA CAC GCG AAC TCT AA	82
ERα	FU:	TGG TGG AGG GTG TTT GTT TTG	RU:	ATC CCA CCA ACA CAC AAA CTC	93
ERα	MP:	CGA TAA AAC CGA ACA ACC CGA CGA	UP:	CCC AAT AAA ACC AAA CAA CCC AAC AA
HIN1	Ext:	GTT TGT TAA GAG GAA GTT TT	Ext:	CAC CGA AAC ATA CAA AAC AAA ACC AC	286
HIN1	FM:	TAG GGA AGG GGG TAC GGG TTT	RM:	CGC TCA CGA CCG TAC CCT AA	74
HIN1	FU:	AA G TTT TTG AGG TTT GGG TAG GGA	RU:	ACC AAC CTC ACC CAC ACT CCT A	123
HIN1	MP:	ACT TCC TAC TAC GAC CGA CGA ACC	UP:	CAA CTT CCT ACT ACA ACC AAC AAA CC
RARβ	Ext:	GTA GGA GGG TTT ATT TTT TGT T	Ext:	AAT TAC ATT TTC CAA ACT TAC TC	239
RARβ	FM:	AGA ACG CGA GCG ATT CGA GTA G	RM:	TAC AAA AAA CCT TCC GAA TAC GTT	82
RARβ	FU:	TTG AGA ATG TGA GTG ATT TGA GTA G	RU:	TTA CAA AAA ACC TTC CAA ATA CAT TC	83
RARβ	MP:	ATC CTA CCC CGA CGA TAC CCA AAC	UP:	AAA TCC TAC CCC AAC AAT ACC CAA AC
RASSF1A	Ext:	GTT TTA TAG TTT TTG TAT TTA GG	Ext:	AAC TCA ATA AAC TCA AAC TCC C	198
RASSF1A	FM:	GCG TTG AAG TCG GGG TTC	RM:	CCC GTA CTT CGC TAA CTT TAA ACG	74
RASSF1A	FU:	GGT GTT GAA GTT GGG GTT TG	RU:	CCC ATA CTT CAC TAA CTT TAA AC	75
RASSF1A	MP:	ACA AAC GCG AAC CGA ACG AAA CCA	UP:	CTA ACA AAC ACA AAC CAA ACA AAA CCA
TWIST	Ext:	GAG ATG AGA TAT TAT TTA TTG TG	Ext:	CCT CCC AAA CCA TTC AAA AAC	273
TWIST	FM:	GTT AGG GTT CGG GGG CGT TGT T	RM:	CCG TCG CCT TCC TCC GAC GAA	78
TWIST	FU:	GGT TTG GGG GTG TTG TTT GTA TG	RU:	CCC ACC TCC TAA CCA CCC TCC	104
TWIST	MP:	AAA CGA TTT CCT TCC CCG CCG AAA	UP:	AAA CAA TTT CCT TCC CCA CCA AAA CA

QM-MSP, quantitative multiplex-methylation specific PCR. Primer abbreviations: Ext, external methylation independent primers used in the first step of QM-MSP; FM, forward methylated primers; RM, reverse methylated primers; FU, forward unmethylated primers; MP, methylated probe sequence labeled with VIC fluorophore; UP, unmethylated probe sequence labeled with FAM fluorophore.

Table 2 Frequency distribution of clinical and tumor features for N = 99 cases

Breast diagnosis	N (%)	Percent × 100
CIS	11	11
DCIS-mINV	15	15
Invasive carcinoma	73	73
Tumor stage
TIS	11	11
T1	53	53
T2	30	30
T3	5	5
Tumor grade
I	10	10
II	71	70
III	15	15
LCIS	3	3
Bilateral breast Dx
Yes	19	19
No	80	80
Metastatic disease
Yes	17	17
No	82	82
Recurrence
Yes	24	24
No	75	75
Age of cancer onset
<30	4	4
30–39	19	19
40–49	43	43
50–59	28	28
60–69	4	4
70–79	1	1
Lymph node status
Positive	27	27
Negative	72	72
ER IHC status
Positive	60	60
Negative	28	28
ND	11	11
PR IHC status
Positive	63	63
Negative	24	24
ND	12	12
HER2 IHC status
Positive	21	21
Negative	36	36
ND	42	42
Deceased
Yes	13	13
No	86	86
BRCA mutation status
BRCA^+	11	11
UNK	7	7
VUS	6	6
NEG	26	26
NT	49	49

Breast diagnosis: CIS, carcinoma in situ; DCIS-mINV, case diagnosed with a <1.0 cm micro-invasive carcinoma, but for which only DCIS remained on the tissue block; IC, invasive carcinoma. Lymph node status: indicates if lymph nodes were positive or negative for tumor cells at initial diagnosis of breast cancer. ER, PR and HER2 status: IHC tumor staining with antibodies to the estrogen and progesterone receptors, and HER2 growth factor receptor, respectively. BRCA mutation status: genetic mutation testing of BRCA1 and BRCA2 genes; BRCA+, carrier of a known deleterious BRCA1 or BRCA2 mutation; UNK, test results unknown. By report, the individual underwent BRCA mutation testing, but the test result was not reported in any of the three Johns Hopkins clinical databases abstracted in the study. VUS, variant of unknown significance; e.g., a BRCA gene mutation was found, but no population data as yet exists to determine if the mutation is deleterious; NEG, negative, the patient did not carry a BRCA1 or BRCA2 gene mutation. NT, not tested.

Table 3 Methylation in n = 99 Breast tumors exceeding 5.0% methylation threshold

	No. of tumors	Average percent (% ×100)	Range of percent (% ×100)
Gene name	Methylated*	Tumor Methylation	Tumor Methylation
ERα	26	32	6–84
TWIST	37	38	6–93
Cyclin D2	24	44	7–93
CDH1	2	19	17–21
APC	44	41	8–91
RASSF1A	62	38	6–96
HIN1	28	62	5–95
RARb	40	37	6–93
BRCA1	23	47	7–84
BRCA2	7	33	7–89

Numer of tumors methylated: Total number of tumors that exceeded a methylation threshold of 5.0 percent for each of ten genes. Range of percent tumor methylation and percent tumor methylation: Range of percent methylation and mean methylation values, respectively for those genes that exceeded a 5.0% methylation threshold.

Acknowledgments

We gratefully acknowledge the support of a National Institutes of Health/National Cancer Institute (NIH/NCI) Specialized Program of Research Excellence (SPORE) in Breast Cancer P50CA88843 to S.S. and an NIH/NINR F31 NR008311-01A1 and American Cancer Society (ACS) DSCN-04-162-01 to T.S.S. We are also thankful for the current support to T.S.S. of a Susan G. Komen Foundation KG090180 (Swift-Scanlan), an NIH/NCRR 1KL2RR025746-01, and an NIH/NCI Breast SPORE CA058823 (Earp).