Protein Microarray Analysis of Mammary Epithelial Cells from Obese and Nonobese Women at High Risk for Breast Cancer: Feasibility Data

This article is featured in Highlights of This Issue, p. 413

Obesity is one of the most important known preventable causes of cancer, accounting for up to 20% of cancer deaths in women, with highest body mass index (BMI) category (BMI > 40 kg/m²) conferring higher risk (1). Previous studies have shown an association between postmenopausal breast cancer risk and excessive body weight, and this association is increased in women with a positive family history of breast cancer (2, 3). In addition, obese women have an increased risk of dying from breast cancer and an increased risk of developing distant metastasis (4, 5).

Studies investigating the molecular basis for the association between breast cancer and increased adiposity have focused on differences in circulating hormones, such as estrogen and insulin, in normal weight versus obese individuals or on how mammary adipocytes and their secreted adipokines may influence tumor cells growth and invasion (6, 7). However, few studies have focused on potential differences in the protein expression patterns of mammary epithelial cells obtained from obese versus nonobese women.

Random periareolar fine needle aspiration (RPFNA) is a research technique used to sample mammary cells from the whole breast of asymptomatic women at high risk for developing breast cancer. It has been shown to successfully predict short-term breast cancer risk in high-risk women (8) and is highly reproducible across institutions (9). The presence of atypia in a breast RPFNA confers a 5.6-fold increased short-term breast cancer risk (8).

Reverse-phase protein microarray (RPPM) is a high-throughput proteomic tool, developed to test for dysregulation of protein signaling networks in human biopsy specimens. It uses denatured lysates, so antigen retrieval, which is a limitation for tissue arrays, is not a problem. RPPM sensitivity and analytic robustness allow for very small amounts of mammary epithelial cells to be selected from RPFNA samples via laser capture microdissection and analyzed for differential protein expression patterns (10–12).

The purpose of this brief communication is to relay findings obtained from a pilot and exploratory study aimed at investigating whether mammary epithelial cells obtained from obese (BMI ≥ 30 kg/m²) and nonobese (BMI < 30 kg/m²) women at high risk for developing breast cancer display differential expression of proteins involved in cancer cell growth, survival, and metastasis.

This study was conducted among 26 women who underwent RPFNA under Institutional Review Board–approved protocol at the Duke University Medical Center (June 2008–February 2010) and who presented sufficient cells for analysis.

Women were required to have at least one of the following major risk factors for breast cancer: (a) 5-year Gail risk calculation of greater than 1.7%; (b) prior biopsy exhibiting atypical hyperplasia, lobular carcinoma in situ, or ductal carcinoma in situ; or (c) known BRCA1/2 mutation carrier. Data for sociodemographic variables, family history of breast cancer, and menopausal status were collected. Women were defined as postmenopausal if they had no menses for more than 12 months in theabsence of pregnancy and/or status postbilateral oophorectomy.

Body weights and heights were clinically measured on a platform scale with a fixed stadiometer. BMIs were calculated, and the standard cutoff point of 30 kg/m² was used to dichotomize obese from nonobese women (13).

RPFNA was performed as published previously, in accordance with methods established and validated by Fabian and colleagues (8, 14). Cells were classified qualitatively as nonproliferative, hyperplasia, or hyperplasia with atypia. The most atypical cell cluster was examined and scored by the Masood cytology index (Table 1). Morphologic assessment, Masood cytology index scores, and cell count were assigned by a single dedicated pathologist who was blinded to BMI.

RPFNA samples were immediately placed and fixed in modified CytoLyt (Hologic Inc.) containing 1% formalin for 24 hours at room temperature. Samples were washed with CytoLyt until the majority of red blood cells were removed, followed by storage at 4°C. Cytospin slides were made for each of the samples using a Shandon Cytospin 4 Cytocentrifuge (Thermo Scientific). RPFNA samples were spun onto slides, using Shandon Single Cytofunnels at 420 rpm for 1 minute, which were then immediately placed in 75% ethanol for 30 seconds. The slides were then stained with hematoxylin (Sigma-Aldrich) for 20 seconds, dehydrated in 75%, 95%, and 100% ethanol for 30 seconds each, followed by xylene for 5 minutes, and allowed to air dry. The 75% ethanol and hematoxylin staining solutions were supplemented with Complete protease inhibitor tablets (Roche Applied Science). Approximately 5,000 epithelial cells were microdissected with a PixCell II Laser Capture Microdissection system (MDS Analytical Technologies) and stored onto microdissection caps which were maintained at −80°C until lysed.

As previously described (15), cellular lysates were printed in triplicate onto nitrocellulose-coated slides, using an Aushon 2470 arrayer equipped with 350-μm pins. The following cellular lysates served as positive controls for antibody staining: A431 ± epidermal growth factor, HeLa ± pervanadate, or Jurkat ± calyculin. Immunostaining was performed as previously described (15). Polyclonal and monoclonal antibodies were purchased from Cell Signaling Technologies, Biosource/Invitrogen, BD Biosciences, and Upstate/Millipore. Each antibody-stained array was scanned and normalized to total protein, as determined by SYPRO Ruby staining. Image analysis was performed as previously described (15).

Cytology specimens were mounted on Superfrost Excell slides (Fisher) and preserved in 95% ethanol. These slides were dehydrated in 95% and 100% ethanol for 30 seconds each and air-dried for 10 minutes. Specimens were fixed in ice-cold acetone for 20 minutes and then air-dried. Rehydration was achieved in TBS for 5 minutes, followed by immersion in 3% hydrogen peroxide for 5 minutes, deionized water, and TBS.

All immunostaining was carried out on Dako Autostainer (Dako). Specimens were placed in Background Buster (Innovex Biosciences) for 30 minutes, followed by incubation with mouse monoclonal anti-human vimentin antibody (Dako; M7020) at 1:300 dilution for 45 minutes and single wash in TBS. Detection and visualization were performed with Envision+ and DAB+ (Dako) as per manufacturer's recommendations. Specimens were counterstained in hematoxylin for 30 seconds, dipped in ammonium water 4 times, dehydrated in graded series of ethanol and xylene, and coverslipped. Frozen human tonsil tissue and inflammatory cells served as positive external and internal controls, respectively.

Immunostained specimens were reviewed and scored by a pathologist (J.G.) who was blinded to BMI and patient identity. Samples were evaluated for average staining intensity (0–3) and percentage of positively stained cells (0%–100%). As previously described (16), a HistoScore was calculated for each stained specimen (product of the 2 aforementioned parameters).

The Mann–Whitney test was used to compare the relative intensity of each proteomic marker between nonobese and obese samples, given the lack of normal distribution of these variables. To compare difference in the median age between nonobese and obese women, the Mann–Whitney U test was performed (Table 2). Fisher's exact test was used to determine the relationships between the two groups and the following variables: race, menopausal status, and Masood cytology. Unsupervised hierarchical clustering analysis of log2-transformed values for each protein was performed with R/Bioconductor version 2.10.1 software. The values of P < 0.05 were considered statistically significant.

This study was performed on 31 RPFNA samples obtained from 26 high-risk women. Of these samples, 10 were bilateral and 21 were unilateral. Table 2 provides a list of the characteristics of study subjects according to 2 BMI categories, nonobese (BMI < 30 kg/m²) and obese (BMI ≥ 30 kg/m²). We have chosen the BMI cutoff of 30 because of well-established associations between poor prognosis and survival of women in the highest BMI categories (4). In the present study, the median BMI of obese women is 31.3 kg/m², whereas the median BMI of nonobese women is 22.3 kg/m². These 2 groups of women are not significantly different with respect to age, race, and menopausal status.

RPFNA samples were stratified according to Masood cytology index scores (refs. 8, 14; Table 1) with the following distribution: 3% were nonproliferative, 16% were hyperplastic, 74% were atypical, and 7% were suspicious of cancer. These scores were further classified into high and low Masood groups (Table 2), with a Masood score cutoff of 15 for cytologic atypia. The presence of atypia has been used as a surrogate marker of short-term breast cancer risk in high-risk women (8). Here, the short-term breast cancer risk in obese and nonobese women is not significantly different (P = 0.67).

Figure 1A depicts the expression pattern of log₂-transformed intensity values of 52 of the original 60 total, phosphorylated, and cleaved proteins obtained from all 31 RPFNA samples. Eight proteins did not yield sufficient intensity signal; these were excluded from statistical analysis.

In unadjusted analysis, the expression of vimentin (P = 0.028) and phosphorylated Bad (p-Bad) S136 (P = 0.049) are statistically different between samples of nonobese and obese women (Fig. 1B and C). However, after adjusting for false discovery rate via the Benjamini–Hockberg method, vimentin and p-Bad S136 expression are no longer statistically significant.

We performed retrospective analysis of vimentin expression by immunohistochemistry (IHC) on a limited number of archived RPFNA cytologic samples. A majority of RPFNA samples had been expended in other experiments, and only 20 of 31 samples were evaluable for vimentin expression by IHC. Representative photomicrographs of epithelial cell clusters from obese and nonobese women are shown in Figure 2A and B. Box plot representation of the distribution of HistoScores of nonobese and obese samples is shown in Figure 2C. Because of the limited samples available for retrospective IHC analysis, we were unable to test for correlation between vimentin expression with RPPM and by IHC.

In this exploratory study, RPPM analysis allowed us to investigate a wide array of proteins that play a role in cell survival, growth, differentiation, and death. The preliminary results presented herein are a first report suggesting that mammary epithelial cells obtained from high-risk obese women may differ in vimentin expression compared with nonobese women. Vimentin is often used as a general marker of mesenchymal cells and has been shown to play a critical role in epithelial-to-mesenchymal transition (EMT) necessary for cancer cell invasion and metastasis (17). In addition, vimentin is an integral structural component of fat droplet organization in adipocytes, lipid metabolism, and adipogenesis (18–20). Treatment of MCF10A cells with arachidonic acid, a metabolite obtained from a high-fat diet, induced EMT and increased vimentin expression with concomitant decrease in E-cadherin expression (21). Considering the established associations between adiposity, adipokines, and vimentin, further investigation into how adipocyte microenvironment influences mammary epithelial cell protein expression in high-risk obese women is warranted.

Few members of the apoptotic pathway have been implicated in adipocyte apoptosis induced by leptin, a hormone that regulates food intake and energy metabolism (22). Animal model studies have shown no difference in protein expression of proapoptotic Bax in mammary tumors of obese and nonobese mice that were fed with high-fat diet and low-fat diet, respectively (23). In the present study, Bax expression is not significantly different (P = 0.055) in obese and nonobese women (Fig. 1D) whereas pro-survival p-Bad S136 expression is higher in nonobese women (Fig. 1C). In the absence of dietary information at the time of recruitment and RPFNA collection, we cannot attribute the differential expression of these proteins to a recent change in caloric intake. But it is interesting to speculate that diet or exercise may impact apoptotic signaling proteins independently of BMI.

An important limitation of this study is small sample size that does not allow for statistical significance of associations between protein expression and risk factors for breast cancer such as age, menopausal status, and cytologic atypia. Thus, it is not possible to draw definitive conclusion on causal relationship between differential protein expression and breast cancer risk in obese versus nonobese women.

In this study, we did not archive adequate RPFNA cytologic samples for a statistically meaningful comparison of RPPM and IHC data. However, our limited data demonstrate that it is possible to prospectively compare RPPM and IHC analysis of specific proteins. The direct comparison of RPPM data with Western blot analysis and/or IHC is clearly important for the future use of RPPM as a tool for multiplexed cell signaling analysis of human samples. Recent studies by Accordi and colleagues show concordance between RPPM results and Western blot analysis of proteins such as Bcl-2 S70 that are activated in primary leukemia samples with mixed lineage leukemia translocations (24). In the present study, we lack statistical power to draw definitive conclusion about the correlation between RPPM and IHC data, given the insufficient or small number of samples. Moreover, the IHC that we employed has obvious limitations such as variable cellularity, heterogeneity of staining, and subjective nature of scoring the cytologic samples. In the future, we will address these limitations by using comparable number of cells for IHC and employing image analysis of stained cells to quantify protein expression.

Our future proteomic studies will include a larger cohort with equal number of obese and nonobese women of Caucasian and African American descent and analysis of adipokines to better understand the underlying pathology of breast cancer initiation and progression in obese women. African American women have higher rates of obesity and more aggressive forms of breast cancer, with greater likelihood of dying from breast cancer in pre- and postmenopausal women (25). Future proteomic studies will better delineate whether these 2 factors are causally related and identify potential biomarkers of response to preventative interventions such as dietary changes and exercise.

No potential conflicts of interest were disclosed.

Primary support was provided by Susan G. Komen Promise grant KG091020 (D. Yu and V.L. Seewaldt), NCI/AVON Partners in Progress grant 3 P30 CA014236-32S1 (V.L. Seewaldt and D. Yu), NIH/NCI grants R01CA88799 and R01CA114068 (V.L. Seewaldt), and Susan G. Komen Career Catalyst in Disparities Research KG090730 (C. Ibarra-Drendall).

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.