Expression of miR-18a and miR-210 in Normal Breast Tissue as Candidate Biomarkers of Breast Cancer Risk

Breast cancer is a disease driven by progressive genetic abnormalities involving mutations in oncogenes and tumor suppressor genes as well as other chromosomal abnormalities (1). More recent research has established that it may also be driven by epigenetic alterations, including DNA methylation, histone modifications, and control of gene expression by noncoding RNAs such as microRNAs (miRNA; refs. 2, 3). Understanding the molecular mechanisms involved in breast cancer initiation and progression will provide strategies to identify new diagnostic and prognostic markers while enabling better prevention and treatment.

miRNAs are a group of noncoding RNAs generally 18 to 25 nucleotide long that can block mRNA translation, affect mRNA stability, and regulate at least 30% of all human protein-coding genes by targeting their 3′- untranslated region (UTR) sequences (4). Studies showing up- or downregulation of miRNAs in human cancers indicate that miRNAs can function as classical tumor suppressors or oncogenes and are associated with the initiation of tumorigenesis and tumor suppressor gene networks (e.g., p53; refs. 5, 6). In breast cancer, growing evidence shows loss of tumor suppressor miRNAs results in enhanced expression of target oncogenes (7), whereas increased expression of oncogenic miRNAs (known as Oncomirs) can repress target tumor suppressor genes (8). The role of miRNAs in breast cancer progression and biology demonstrates that the expression of specific miRNAs correlates with features such as estrogen receptor (ER) and progesterone receptor (PR) expression (9, 10) and other key features such as cell proliferation, cell cycle, invasion, and metastasis (11–13).

miRNAs are therefore candidate biomarkers because of their crucial role in the regulation of large gene networks. Their expression in normal high-risk breast tissue has never been evaluated and may provide information about early dysregulation events in miRNA expression that are important in the high-risk breast. Thus, we conducted a carefully designed study where miRNA expression profiles of ER-positive tumors and their matched contralateral unaffected breast (CUB) samples were compared with those of ER-negative tumors and their matched CUB samples and both groups were compared with normal reduction mammoplasty samples. All 3 groups of subjects (ER-positive cases, ER-negative cases, and reduction mammoplasty controls) were matched for age and race. Our design allowed us to evaluate miRNA expression across the spectrum of normal breast tissue from reduction mammoplasty samples, high-risk non-atypical breast tissue from the contralateral breasts of patients with breast cancer, and matched breast cancer samples to seek potential miRNA biomarkers as candidate early drivers of tumor initiation and progression.

We queried the medical records of patients treated at the Lynn Sage Comprehensive Breast Center of Prentice Women's Hospital of Northwestern University, following Northwestern University IRB approval (protocol NU09B1). Women with unilateral breast cancer who opted for contralateral risk-reducing mastectomy were identified. The patients with ER-positive breast cancer were matched for age (±2 years), race, and menopausal status to those with ER-negative breast cancer. Similarly matched reduction mammoplasty controls were also obtained. Archived formalin-fixed, paraffin-embedded (FFPE) tissue was retrieved and blocks that contained sufficient samples of tumor, non-atypical CUB tissue and reduction mammoplasty tissue were selected. This process yielded 75 samples from 15 matched triplets of 45 women with adequate FFPE material for study (Table 1, discovery set with 15 with ER-positive primary, 15 with ER-negative primary, each with their respective CUBs, and 15 reduction mammoplasty controls). The ER status designation was based on immunohistochemical assays performed for routine clinical care. The histology of the primary tumor and the CUB tissue was reviewed by a breast pathologist (M.E. Sullivan) who selected areas for further processing. Known carriers of BRCA1 and BRCA2 mutations were excluded.

For the validation set, we recruited patients diagnosed with unilateral breast cancer proceeding to contralateral prophylactic mastectomy at the Prentice Women's Hospital of Northwestern Medicine under an IRB approved protocol (NU11B04), with exclusions for neoadjuvant treatment, prior endocrine therapy or pregnancy/lactation during the prior 2 years. Again, known BRCA1/2 mutation carriers were excluded. A group of patients with reduction mammoplasty were also recruited as standard risk controls. FFPE tissue samples of tumor and CUB from 40 bilateral mastectomy cases (20 ER-positive and 20 ER-negative) were used in this portion of the study. FFPE samples from 20 healthy patients with reduction mammoplasty were used as controls. The ER-positive cases, ER-negative cases, and controls were matched by age, race, and menopausal status (Table 1, validation set 1).

Because risk biomarkers are only valuable if present in benign breast biopsy (BBB) prior to the onset of breast cancer, we assembled a second validation set, consisting of BBB from case–control pairs under an IRB-approved protocol (NU09B6). The cases were women undergoing benign breast biopsy at least 1 year prior to breast cancer diagnosis. The controls were the benign biopsy samples obtained from women who remain cancer free to the time of enrollment into the study, with similar or greater follow-up duration, and matched with cases by age (±2 years), race, and year of biopsy (±1 year). A section of all BBB blocks from cases and controls were reviewed by a pathologist to identify areas of matching histology and classified as “non-proliferative” (class1), “proliferative” (class 2), or “atypical proliferation” (class 3). Non-atypical Terminal duct lobular units (areas of hyperplasia of the usual type and nonproliferative breast epithelium) were marked for laser capture microdissection (LCM). Atypical hyperplasia, if encountered, was subjected to LCM and stored for future analyses. Forty BBB cases (20-ER positive cases, 20 ER-negative cases) and 40 BBB controls were used in this study (Table 1, validation set 2). Overall, 255 samples from 185 patients (among them, 70 patients provided both CUB and tumor samples) were used for discovery and validation in this study.

Ten-micrometer-thick sections of FFPE tissues were prepared and mounted on PEN-membrane glass slides (Molecular Devices) and then dewaxed, dehydrated, and stained using the LCM Staining Kit as described by the manufacturer (Ambion) prior to LCM. Areas of invasive tumor and normal ductal epithelium were outlined under microscopic observation. Outlined areas were then microdissected using Veritas microdissection system with the CapSure macro LCM caps (Molecular Devices). The microdissected tissues were then transferred to an LCM cap and tissue RNA was extracted using RecoverAll total nucleic acid isolation kit as recommended by the manufacturer (Ambion). Total RNA quantity and quality were evaluated using Nanodrop ND-1000 and Agilent 2100 Bioanalyzer.

Reverse transcription (RT) reactions were performed using TaqMan microRNA Reverse Transcription Kit according to the manufacturer's instructions (Thermo Fisher Scientific) in a 7900HT Fast Real-Time PCR system as described previously (14, 15). TaqMan Low Density Arrays (TLDA, Thermo Fisher Scientific) were used to quantify the expression levels of 754 mature miRNAs and housekeeping miRNAs (MammU6, RNU44, and RNU48). The probes covered all the miRNA that were known at the time when the study was initiated. Reactions were run in a 384-well TLDA block at 94.5°C for 10 minutes, followed by 40 cycles at 97°C for 30 seconds and 60°C for 1 minute. A customized TLDA card representing a subset of 7 target miRNAs (miR-18a, miR-210, miR-214*, miR-124, miR-193a-3p, miR-485-3p, and miR-671-3p) and 3 housekeeping miRNAs (MAMMU6, RNU44, and RNU48) in duplicates was used to validate the differentially expressed miRNAs that were identified in the discovery set. TaqMan gene expression assays of miR-18a and miR-210 and three housekeeping miRNAs (MAMMU6, RNU44, and RNU48) were used to validate the miRNAs in benign biopsy samples.

Differences in age among ER-positive cases, ER-negative cases, and controls were examined using ANOVA and differences in the distribution of race, menopausal status, invasive cancer, and tumor grade were examined using χ² tests. For miRNA expression, values of C_t were generated for 768 unique probes (including the internal controls) on 2 separate TLDA (A and B) mapping to 671 unique miRNAs and respective controls. Probes were retained for analyses if at least 25 of the total 75 samples had a signal. We selected MAMMU6 as an internal control for normalization and the mean of 4 MAMMU6 control probes were used for TLDA-specific background subtraction. The resultant ΔC_t values were quartile-normalized across all TLDA arrays, with separate normalizations performed for the A and B TLDA arrays. Linear models were used to conduct pairwise comparisons of expression values with Empirical Bayes variance correction functionality (16). False Discovery Rate (FDR) correction (17) was applied in a nested F-test strategy and 5 specific tests of contrast were evaluated: ER⁺ tumors versus reduction mammoplasty control, ER⁺ CUB versus reduction mammoplasty control, ER⁻ tumors versus reduction mammoplasty control, ER⁻ CUB versus reduction mammoplasty control. The potential target miRNAs were defined as those for which miRNAs were differentially expressed in both tumor and CUB tissue relative to control for each ER subtype with an estimated fold change (calculated by raising 2 to the power of the normalized ΔC_t values) of at least 3 and adjusted P values of less than 0.05. All statistical analyses were performed using R (2.14.0) packages from the Bioconductor project (release 2.9, http://www.bioconductor.org).

For the validation of the individual miRNAs in the customized TLDA and TaqMan expression assays, the means of the 3 housekeeping miRNAs (MammU6, RNU44, and RNU48) were used to normalize the expression of each target miRNA. ANOVA was performed to examine differential expression among groups and pairwise comparisons were conducted using a Sidak adjustment for multiple comparison. Logistic regression analysis and receiver operating characteristic (ROC) analysis were performed using the expression of miR-18a and miR-210 to establish models discriminating BBB cases from controls. Tertiles were constructed on the basis of ΔC_t values for miR-18a and miR-210 and then a combined variable was created, reflecting expression in the lowest tertile for both miRs, highest tertile for both, and any other combination as a middle group. ROC analysis was performed using this combined variable.

Among the 45 subjects in the discovery set, the 3 groups (ER-positive cancer, ER-negative cancer, and reduction mammoplasty controls) were comparable with regard to age, race, and menopausal status. There was no significant difference in these parameters by ANOVA tests (Table 1, discovery set). In cancer cases, the majority of the index cancer was invasive ductal carcinoma (IDC; 80% in the ER-positive group, 100% in the ER-negative group).

The first validation set consisted of tumor, CUB, and reduction mammoplasty samples from 60 subjects that were independent of the discovery set and comparable in terms of age, race, and menopausal status across the 3 groups (Table 1, CUB validation set). In the cancer cases, 60% of ER-positive cases and 70% of ER-negative cases were IDC.

In the benign biopsy validation sample set, 20 ER-positive cases and 20 ER-negative cases were matched on age, race, and menopausal status, which were also matched on the same parameters to 40 controls (Table 1, BBB validation set). In the cases, 75% of ER-positive cases and 85% of ER-negative cases had subsequent IDC. There was no significant difference in tumor size and tumor grade among the 3 sample sets as shown in Supplementary Table S1.

For the initial set of comparisons of miRNA expression profiles, a total of 490 probes mapping to 435 unique miRNAs were retained after excluding those with more than 2/3 undetectable calls. Among the detectable miRNA, 15 miRNAs were significantly increased in ER⁺ or ER⁻ tumor and 20 miRNAs were significantly decreased in ER⁺ or ER⁻ tumor when compared with the normal tissues from reduction mammoplasty samples (Supplementary Table S2). As the major goal was to find the miRNAs that were differentially expressed not only in tumor but also high-risk benign tissues, we identified 7 miRNAs that were differentially expressed among the 5 groups of samples by comparing ER-positive tumors and matching CUBs, ER-negative tumors and matching CUBs, to the reduction mammoplasty controls (Fig. 1). Among them, miR-18a, miR-210, and miR-214* were expressed at significantly higher levels in ER-positive tumors and ER-negative tumors than in reduction mammoplasty samples. Their expression in the matching CUB samples displayed an intermediate level of expression, which was lower than the expression in tumors and significantly higher than in reduction mammoplasty samples. In addition, for miR-18a and miR-210, differences between tumors and the matching CUBs were also significant. The other 4 miRNAs (miR-124, miR-193a-3p, miR-485-3p, and miR-671-3p) were significantly higher in reduction mammoplasty controls than in tumors and their matching CUBs. There were no significant difference between tumors and CUBs.

We performed validation of the 7 target miRNAs using an independent sample set consisting of 20 ER⁺ tumors and their matching CUBs, 20 ER⁻ tumors and their matching CUBs, and 20 reduction mammoplasty controls. Among the 7 target miRNAs, the expression of miR-18a and miR-210 was significantly higher in tumor (regardless ER status) and matching CUB samples when compared with normal controls (Fig. 2). The difference between tumors and CUBs was not significant, although the expression level in CUBs was intermediate between tumor and controls, suggesting that miR-18a and miR-210 were potential biomarkers not only for tumor but also for high-risk breasts. The other 5 target miRNAs (miR-214*, miR-124, miR-193a-3p, miR-485-3p, and miR-671-3p) did not show significant difference when tumor and CUB were compared with the reduction mammoplasty controls (Supplementary Fig. S1) thus were not validated in the independent samples. We also evaluated ER expression in these samples and found that there was no significant difference between CUBs and normal controls (Supplementary Fig. S2A). As expected, the expression of ER was significantly increased in ER-positive tumors and decreased in ER-negative tumors.

We then stratified the CUB sample set on the basis of occurrence of noninvasive cancer [9 ductal carcinoma in situ (DCIS) cases] and invasive cancer (including 26 IDCs and 5 invasive lobular carcinomas) regardless of ER status. The expression of miR-18a and miR-210 was significantly higher in invasive tumor than in all other sample types; these included DCIS (P < 0.05 for miR-18a, P < 0.01 for miR-210) and CUBs (P < 0.01 for both miRs) as well as reduction mammoplasty controls (P < 0.001 for both miRs; Fig. 3). The expression of both miRs in DCIS and in CUBs was similar and displayed an intermediate position between invasive tumors and reduction mammoplasty controls. Expression was significantly higher in DCIS and CUBs than in reduction mammoplasty controls (P < 0.001 for miR-18a in CUBs vs. reduction mammoplasty samples and P < 0.01 for miR-210 in CUBs vs. reduction mammoplasty samples). Overall, the expression of miR-18a and miR-210 showed an increasing trend from normal breast to high-risk breast (CUBs and DCIS) and to invasive cancer. We then correlated the 7 target miRNAs to each other and to hormone receptors within each sample. There was a strong correlation between miR-18a and miR-210 in all the samples and in different subgroups (R > 0.9, P < 0.001, Supplementary Table S3). The correlation among 5 miRNAs was stronger in CUBs (especially in CUBs of ER-positive cases) than in tumors. The negative correlation between miR-18a and ER was only significant in ER-positive tumors (R = −0.394, P = 0.036).

We assayed the expression of miR-18a and miR-210 in benign breast biopsy samples from women, who later developed cancer (20 cases who developed ER-positive cancer and 20 cases who developed ER-negative cancer) and from controls who did not develop subsequent cancer (40 controls). The results revealed that the expression of miR-18a and miR-210 was significantly increased in benign biopsy samples from cases (both ER-positive and ER-negative cases) compared with the controls (Fig. 4). There was no significant difference in the case BBB samples, by ER status of the subsequent cancer. Overall the differences between all cases (40) and all controls (40) were significant (P = 0.0021 for miR-18a, P = 0.00025 for miR-210). The distribution of miR expression varied across tertiles when both miRs were in the lowest tertile (N = 14, including 4 cases and 10 controls), both were in the highest tertile (N = 16, including 13 cases and 3 controls), and any other combination as a middle group (N = 50, including 23 cases and 27 controls). ROC analysis using this combined variable discriminated cases from controls with AUC = 0.64 [OR, 3.20; 95% confidence interval (CI), 1.17–8.95; P = 0.023]. BBB samples from ER-positive cases displayed strong and significant correlations between miR-18a and miR-210 and between miR-18a and the hormone receptors (ER, PR, AR, and PRLR). miR-18a also displayed strong significant correlations with miR-210 in BBB from ER-negative cases and in control BBB (Supplementary Table S4).

Several earlier profiling studies have identified miRNAs that are differentially expressed between normal breast tissue and tumor samples or differentially expressed among different subtypes of breast cancer (18–22). These studies have helped establish miRNA signatures for breast cancer and provide insights regarding their pathogenic role but have mainly focused on the miRNA signatures in cancer. No previous studies have evaluated miRNA expression in benign breast with the goal of identifying miRNA signatures of breast cancer risk. As the first attempt in this direction, we compared the expression profiles of 754 miRNAs in 30 breast cancer samples, paired CUB samples from the same women, and normal reduction mammoplasty samples from the age of 15 years and race-matched women. From this discovery set of 75 samples from 45 subjects, we found 7 miRNAs that were differentially expressed across the 3 sample groups, with miR-18a, miR-210, and miR-214*, at higher levels in tumors compared with reduction mammoplasty samples, and miR-124, miR-193a-3p, miR-485-3p, and miR-671-3p at lower levels in tumors than in reduction mammoplasty controls. The expression in the paired CUB samples displayed an intermediate level of expression, lower than in the tumor samples, but higher than in reduction mammoplasty control samples. This was particularly true for miR-18a and miR-210, where differences between tumors and the matching CUBs were also significant. Thus the target miRNAs displayed a trend in expression from normal reduction mammoplasty to CUB, to breast cancer, suggesting that the contralateral high-risk breast tissue displays molecular changes that are present in matched tumors from the same individuals and are therefore candidate early drivers of tumor initiation and progression. We then examined the 7 target miRNAs identified in the discovery set in an independent validation set of 100 samples from 60 subjects using the same design (tumor and paired CUB samples, matched to reduction mammoplasty samples) and confirmed that the expression of miR-18a and miR-210 was significantly increased in both ER-positive and ER-negative tumors and their matching CUBs when compared with reduction mammoplasty controls.

To construct a control group that matched with the case group by age, race, and menopausal status, we recruited reduction mammoplasty women to represent the healthy women group. It is well-known that the contralateral breast of ipsilateral breast cancer patient is in high risk for the second cancer compared with the healthy women at standard risk. Women undergoing reduction mammoplasty are at standard risk of breast cancer and do not present for surgery because of a concern regarding breast cancer risk or a known breast lesion. Therefore, they are the population that is closest to standard risk and have been used as “normal healthy controls” in multiple molecular studies of breast cancer. We have evaluated the Gail risk estimates of our age-matched reduction mammoplasty controls and find that the risk of developing invasive breast cancer during the next 5-year period is 1.0% (median; range, 0.4%–1.8%) and the lifetime risk is 9.4% (median; range, 5.0%–11.9%). For the future studies, we will include samples from the Komen Tissue Bank, representing healthy women without macromastia.

We were able to further validate the increased expression of miR-18a and miR-210 in 40 cases of benign breast biopsies from women who subsequently developed breast cancer, compared with 40 BBB controls from women who have not developed breast cancer. In a ROC model using the combination of tertiles of miR-18a and miR-210 expression, we were able to successfully discriminate cases from controls (OR, 3.20; 95% CI, 1.17–8.95). The results from these 2 validation sets provide robust confirmation of the notion that the contralateral breast is an excellent high-risk model for breast cancer risk biomarker discovery and that miR-18a and miR-210 are strong candidate biomarkers of breast cancer risk, in benign breast samples.

We did no focus of biopsies showing atypical hyperplasia, to keep our study results more generalizable. Although women with atypical hyperplasia are at markedly increased risk of breast cancer, they comprise under 10% of benign biopsies (23). Of all the women who develop breast cancer, at least 75% do not have a history of prior benign biopsy (24), and of those who do, most do not have atypical hyperplasia. Our long-term goal is to develop risk markers that can be applied to all benign biopsies and eventually may be extendable to women who do not have a clinical indication for breast biopsy, using methods such as random fine-needle aspiration or random core biopsy. Therefore, from a risk reduction perspective, the impact of a biomarker that applies to all women (or at least all women who undergo benign biopsy) is likely to be far greater than risk markers that apply only to women with atypical hyperplasia. In our present results, we do not see a significant difference between high-risk benign breast samples (i.e., CUB samples) and DCIS, although admittedly the number of DCIS samples is small. If this holds true in larger studies, it would imply that expression of these miRs in atypical hyperplasia is similar. We are reserving atypical hyperplasia foci during LCM, but these are rare and RNA yield is limited, adding a logistical barrier to the examination of atypical hyperplasia samples.

We matched our patient population for age and menopausal status and balanced the hormone receptor status of the tumors across all 3 sample sets. However, there are other factors that may affect miRNA expression. In the discovery and CUB validation samples sets using bilateral mastectomy and reduction mammoplasty, we excluded patients who had received neoadjuvant treatment, prior endocrine therapy, or were pregnant or lactating during the prior 2 years. In the benign biopsy validation set, information regarding hormone use at the time of benign biopsy was not available. Nevertheless, increased expression of miR-18a and miR-210 was observed, suggesting that either hormone use was infrequent, or that hormone use does not substantially influence the expression of these genes.

Our data complement a recent profiling study using deep sequencing of miRNAs in DCIS revealed that several miRNAs (including miR-210) were decreased in the transition from normal breast to DCIS, but increased in the transition from DCIS to IDC (25, 26). In this study, genes with inversely related profiles to miR-210 included BRCA1, FANCD, FANCF, PARP1, E-cadherin, and Rb1, which were all activated in DCIS and downregulated in the invasive carcinoma. In our study, we found that miR-210 and miR-18a were significantly increased in IDC compared with DCIS, which is consistent with the changes described by Volinia and colleagues in the invasive transition (25, 26). However, we did not find significant changes in DCIS when compared with normal breast or CUBs. This might be due to the smaller number of DCIS lesions in our study (20 normal control and 9 DCIS) compared with the other study (42 normal control and 17 DCIS) but may also be related to the use of laser capture of epithelial areas, so that stromal expression may have been missed. In addition, we observed that miR-210 expression in DCIS samples was more variable than in any of the other sample sets (see Fig. 3). In expanded analyses with larger sample numbers that we are currently assembling, we intend to address these issues.

The marked overexpression of miR-18a that we observed in cancer samples in this study is not surprising, as miR-18a has been found to be associated to breast cancer (20, 27). miR-18a may directly target ER (28, 29) and suppress hypoxia-inducible factor 1α activity (30) and ATM expression (31, 32). The expression of miR-18a was at higher levels in ER-negative tumors than in ER-positive tumors (20, 27). High expression of miR-18a is strongly associated with basal-like breast cancer features (high proliferation, ERα negativity, and CK5/6 positivity). Transfection of miR-18a into breast cancer cell lines (MCF7 and BT-474) suppressed ER expression (33). However, in another study, expression levels of miR-18a showed no significant differences between ER-positive and ER-negative tumors (34). Instead, pre-miR-18a levels were significantly higher in ER-positive tumors than in those that were ER-negative. This may indicate that impaired pre-miR-18a processing to miR-18a occurs in different subtypes of tumors. In our studies, we found that the expression levels of miR-18a were relatively higher in ER-negative cancer samples than in ER-positive cancer samples in both the discovery set (Fig. 1) and validation set (Fig. 2) but did not reach statistical significance, which is similar to the report from Castellano and colleagues (34). In the benign breast tissues, there was no significant difference between the CUB of ER-positive cases and the CUB of ER-negative cases but both were significantly higher than the reduction mammoplasty controls, suggesting that miR-18a acts as a breast cancer risk marker without ER subtype specificity. A recent study identified a novel circulating microRNA signature for early-stage ER-positive breast cancer. Four miRNAs (miR-18a, miR-15a, miR-107, and miR-425) were more highly expressed in women with ER-positive breast cancer than in healthy controls (35). In the latest version of miRTarBase (36), there are 28 genes (including ESR1 and ATM) that have been validated experimentally by at least 2 methods (reporter assay, Western blotting, qPCR, microarray, next-generation sequencing, or pSILAC experiments; Supplementary Table S5). Gene ontology data indicate that the gene functions are enriched for gene transcription, apoptosis regulation, and intracellular signaling cascade (Supplementary Table S6).

The clinical importance of miR-210 expression has been evaluated in several meta-analyses in different cancers (37–39). miR-210 is correlated with lower overall survival, distant disease-free survival, and progression-free survival in patients with breast cancer (39). In other studies, higher expression of miR-210 significantly predicted poorer outcome, higher recurrence, and overall decreased survival rates (37, 38). Circulating miR-210 levels have been shown to correlate with trastuzumab sensitivity, tumor progression, and lymph node metastasis, suggesting that miR-210 can be used to predict and monitor the response of HER2+ breast cancer to trastuzumab (40). miR-210 is inducible by hypoxia, its targeting genes are mainly related to hypoxia and contribute to tumor initiation (41). In the miRTarBase (36), there are 69 genes that have been validated experimentally at least by 2 methods, including HIF1A and HIF3A (Supplementary Table S7). Gene Ontology data indicate that the gene functions are enriched in regulation of cell cycle, proliferation, and apoptosis (Supplementary Table S8).

Thus, both miR-18a and miR-210 have known functions that are implicated in breast cancer progression, but their role in breast cancer initiation has not so far been defined. Our study suggests that both these genes are involved in the early steps of tumorigenesis; they are overexpressed in the CUB of patients with breast cancer and in benign biopsy samples obtained prior to the development of breast cancer. Therefore, they are candidate early drivers of tumorigenesis and potential biomarkers of breast cancer risk and may allow more accurate identification of women at increased breast cancer risk than with current methods. Improved risk stratification may then result in improved acceptance of preventive medication by high-risk women, which will represent an advance in breast cancer prevention, as the reluctance of high-risk women to accept breast cancer prevention drugs is presently a lost opportunity in breast cancer prevention. We plan future studies to examine the structure and function of these genes in normal breast and breast cancer models and larger validation studies to establish their utility as biomarkers of breast cancer risk.

No potential conflicts of interest were disclosed.

Conception and design: A. Shidfar, E.F. Vanin, M.B. Soares, J. Wang, S.A. Khan

Development of methodology: F.F. Costa, E.F. Vanin, M.B. Soares, J. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Shidfar, F.F. Costa, J.M. Bischof, M.E. Sullivan, D. Ivancic, E.F. Vanin, M.B. Soares, J. Wang, S.A. Khan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F.F. Costa, D.M. Scholtens, J.M. Bischof, M.E. Sullivan, E.F. Vanin, M.B. Soares, J. Wang, S.A. Khan

Writing, review, and/or revision of the manuscript: A. Shidfar, F.F. Costa, D.M. Scholtens, M.E. Sullivan, E.F. Vanin, M.B. Soares, J. Wang, S.A. Khan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.B. Soares, J. Wang

Study supervision: M.B. Soares, J. Wang, S.A. Khan

The authors acknowledge E.F. Vanin for his tremendous help in analyzing the initial data and for his intellectual insights.

This study is supported by the Avon Foundation (Center of Excellence Award to Northwestern University) and the Susan G. Komen Foundation (grant # 12222783) to S.A. Khan.

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.