Menopause Is a Determinant of Breast Adipose Inflammation

Chronic inflammation is associated with the development of a growing number of epithelial malignancies (1–3). Obesity, defined as body mass index (BMI) ≥ 30, is associated with an increased risk of developing hormone receptor (HR)–positive breast cancer after menopause and a worse prognosis after breast cancer diagnosis (4–10). Obesity results in chronic, subclinical inflammation characterized by elevated circulating proinflammatory mediators (4, 11), which have been linked to the development and progression of breast cancer (12–14). Other mechanisms suggested to contribute to the obesity–cancer link include increased circulating estrogens (4, 15, 16), altered adipokine levels, and insulin resistance (12, 17).

Locally, adipocyte hypertrophy occurs in the obese fat pad leading to immune cell recruitment and inflamed white adipose tissue (WAT; refs. 4, 18–20). This WAT inflammation, which occurs commonly in overweight and obese individuals, is histologically defined by the presence of inflammatory foci known as crown-like structures (CLS; ref. 18). Composed of a dead or dying adipocyte encircled by macrophages, CLS were first observed in the visceral and subcutaneous fat in association with the metabolic syndrome (18, 19, 21). More recently, we demonstrated both in experimental models of obesity and in obese women with breast cancer that CLS occur in the WAT of the mammary gland and breast (CLS-B), respectively (22, 23). The association between elevated BMI and CLS-B has been confirmed in women without breast cancer who underwent reduction mammoplasty (24).

Breast WAT inflammation, manifested as CLS-B, correlates with elevated BMI, activation of NF-κB, and increased levels of proinflammatory mediators and aromatase, the rate-limiting enzyme for estrogen synthesis (22, 23). The severity of breast WAT inflammation is a better correlate to tissue levels of proinflammatory mediators and aromatase activity than is BMI (22, 23). Thus, WAT inflammation, which commonly occurs in obese individuals but has been demonstrated to occur across the spectrum of BMIs, is likely to contribute to the pathogenesis of breast cancer.

Despite substantially lower circulating levels of estrogen after menopause, somewhat paradoxically, estrogen-dependent tumors are the most common subtype of breast cancer in postmenopausal women. Similar to the obese state, postmenopausal status is associated with elevated circulating levels of C-reactive protein (CRP) and proinflammatory mediators (25, 26). Furthermore, ovariectomy sensitizes mice to WAT inflammation (27). Thus, as in obesity-associated malignancies, inflammation may provide a key link between the postmenopausal state and cancer development. However, a relationship between menopausal status and breast WAT inflammation in humans, as quantified by assessing CLS-B, has not been reported.

In the current study, we had three main objectives. First, we determined whether menopausal status, independent of BMI, contributes to breast WAT inflammation. Second, we expanded upon our previous pilot study to confirm the correlations between BMI and breast WAT inflammation, and between breast WAT inflammation and adipocyte diameter (23). Third, we investigated whether WAT inflammation of the breast is a sentinel for adipose tissue inflammation in other fat depots. We show that postmenopausal status, independent of BMI, is associated with breast WAT inflammation. Moreover, our findings confirm that the majority of obese and overweight women, and a minority of women with normal BMI, have CLS-B. Finally, we present evidence that breast WAT inflammation is a sentinel for adipose inflammation in another adipose depot. Collectively, our findings demonstrate that both obesity and the postmenopausal state independently contribute to breast WAT inflammation, thereby providing a plausible explanation for the elevated risk of HR-positive breast cancer in obese postmenopausal women.

This study was approved by the Institutional Review Boards of Memorial Sloan Kettering Cancer Center (MSKCC) and Weill Cornell Medical College (New York, NY). Informed consent was provided by women undergoing mastectomy at MSKCC, and breast WAT specimens were obtained under a standard tissue acquisition protocol. Patients underwent mastectomy for treatment of breast cancer (n = 211) or to reduce risk of breast cancer (n = 26). To ensure adequate tissue for analysis, patients undergoing lumpectomy were excluded. Bilateral breast WAT was obtained from a subgroup of women undergoing bilateral mastectomy. Breast and abdominal WAT were obtained from a subgroup of women undergoing mastectomy with immediate autologous flap reconstruction. Clinicopathologic data (age, race, BRCA mutation status, tumor subtype, diagnoses of hypertension, diabetes, dyslipidemia, medications used, e.g., statins, breast cancer treatments including endocrine therapy and preoperative chemotherapy) were systematically extracted from the electronic medical record by research staff and physicians, and independent data review was carried out for quality assurance. Height and weight were prospectively recorded before surgery and used to calculate BMI. Standard definitions were used to categorize BMI as under- or normal weight (BMI < 25), overweight (BMI, 25.0–29.9), or obese (BMI ≥ 30). Menopausal status was categorized as either premenopausal or postmenopausal based on National Comprehensive Cancer Network (NCCN) criteria (28). In brief, women were classified as postmenopausal if they had bilateral oophorectomy or reported permanent cessation of menses for 12 or more months in the absence of chemotherapy or endocrine therapy. All data were reviewed twice for accuracy independently by 2 physicians.

For all cases, up to 5 paraffin blocks were prepared from WAT obtained from each surgical site (breasts ± abdomen). For cases with ipsilateral invasive breast tumors, samples were obtained from quadrants not involved by tumor. Specimens were examined grossly and with hematoxylin and eosin (H&E) staining by a breast histopathologist (D. Giri) to ensure samples were representative of normal breast tissue or abdominal subcutaneous tissue when available.

Methods to detect CLS have been described previously (23). Briefly, 2 sections from each of 5 tissue blocks (5-μm thick and approximately 2 cm in diameter) were stained by H&E and for CD68, a macrophage marker (mouse monoclonal KP1 antibody; Dako; dilution 1:4,000). If fewer than 5 paraffin blocks were available, additional sections were cut from the same block at 50 μm apart. In this manner, a total of 5 sections, stained with CD68, were generated per case for CLS analysis. The CD68-stained sections were examined under light microscopy by the study pathologist (D. Giri) to detect CLS in breast and abdominal WAT when available. The number of CLS in each section was counted and recorded. Following this process, the density of CLS by surface area was determined as follows: each CD68-stained slide was digitally archived by gross photography using a Nikon COOLPIX digital camera (Nikon Co.). Images were stored in the tagged image file format (TIFF), and the WAT tissue area on each slide was measured using the Image J Software (NIH). Epithelial tissue area and areas of fibrosis were excluded. The severity of WAT inflammation was quantified as CLS-B per square centimeter of WAT (CLS-B/cm²).

Measurement of adipocyte diameters has been described previously (23). Briefly, breast H&E sections were photographed at 20× using an Olympus BX50 microscope and MicroFire digital camera (Optronics). Images were stored as TIFF files, and mean diameters were calculated using measurements from 30 or more individual adipocytes for each patient using the linear dimensional tool in the Canvas 11 Software (ACD Systems International, Inc.).

The primary endpoints of this study were (i) presence or absence of WAT inflammation (binary variable) defined by CLS-B or CLS in abdominal WAT when available, and (ii) severity of breast WAT inflammation, defined categorically as follows and using median number of CLS-B/cm² (0.38) as the cutoff: no inflammation (absent CLS-B), mild inflammation (≤0.38 CLS-B/cm²), and severe inflammation (>0.38 CLS-B/cm²). Associations between presence of CLS-B and clinicopathologic features, including BMI (both as a continuous and categorical variable) and menopausal status, were examined using logistic regression and the Fisher exact test where appropriate. Additional clinicopathologic features that were examined include age, race, BRCA mutation status, tumor subtypes (hormone receptor positive, HER2-amplified, or triple negative), invasive or in situ carcinoma, diagnosis of hypertension, diabetes, dyslipidemia, use of NSAIDs, statins, blood pressure–lowering drugs, endocrine therapy, and preoperative chemotherapy. Associations between categorized number of CLS-B/cm² and clinicopathologic features were examined using ANOVA or the nonparametric Kruskal–Wallis test for continuous variables and the Fisher exact test or χ² test for categorical variables. Among subjects with CLS-B, associations between CLS-B/cm² (as a continuous variable) and clinicopathologic features were evaluated using the Spearman correlation for continuous covariates and either the nonparametric Kruskal–Wallis test or Wilcoxon rank-sum test where appropriate for categorical covariates. Associations between mean adipocyte size and categorical variables, including CLS-B presence, categorized CLS-B/cm², and clinicopathologic features, were evaluated using ANOVA and/or the Student t test where appropriate. The strength of the correlation between BMI and mean adipocyte size was quantified using the Spearman rank correlation coefficient.

For the multivariable model, covariates of interest including BMI and menopausal status were identified by the univariate analyses results in all study subjects and in subgroups defined by BMI categories and menopausal status, the Akaike information criterion–based backward variable selection procedure, and previous related studies (29). Potential confounding factors, including age, use of NSAIDs, statins, and preoperative chemotherapy, were also included in the final multivariable model for presence or absence of WAT inflammation (binary variable). Using a similar variable selection procedure, the following covariates including, BMI, menopausal status, regular use of NSAIDs, use of endocrine therapy at the time of surgery, and preoperative chemotherapy, were included in the multivariable logistic regression model for severity of breast WAT inflammation in subjects with CLS-B.

When bilateral breast tissue and abdominal subcutaneous WAT were available, the strength of correlation in CLS/cm² between breasts or between breast and abdominal tissue were quantified using the Kendall rank correlation coefficient. Correlation coefficients were tested against the null hypothesis that the correlation coefficients would be 0. Results with P values less than 0.05 were considered statistically significant.

From January 2011 through August 2013, we expanded our prior pilot study (n = 30) leading to a total enrollment of 238 women who underwent mastectomy. Breast WAT was obtained from 237 women (only abdominal WAT was obtained from 1 patient) of median age 48 years (range, 22–90; Table 1). Median BMI was 25.3 (range, 17.3–50.0). Overall, 144 of 237 (61%) women were premenopausal at the time of surgery, whereas 93 of 237 (39%) women were postmenopausal. In addition, 148 of 237 (62%) patients had no evidence of ipsilateral invasive cancer in the breast examined for CLS-B. Thirty-four (14%) patients received preoperative chemotherapy. Fifteen (6%) patients were on tamoxifen or an aromatase inhibitor at the time of surgery. Known BRCA1 or BRCA2 gene mutations were present in 49 of 125 (39%) patients who were tested, reflecting the institutional catchment. Regarding co-morbid conditions, 36 of 237 (15%) women had hypertension and 10 of 237 (4%) had diabetes mellitus.

Among all study patients, 122 of 237 (51%) women had evidence of CLS-B (Fig. 1A and B; Table 1). CLS-B were observed in 19 of 34 (56%) patients who received preoperative chemotherapy, which is slightly higher compared with that among patients who did not receive preoperative chemotherapy (51%, P = 0.71). Statistically significant differences were not found for rates of hypertension, diabetes, or use of NSAIDs or statins in women with versus without CLS-B (Table 1). As in our pilot study of 30 women (23), we examined the correlation between CLS-B (presence/absence) and BMI, both as a continuous and as a categorical variable. In this expanded population, BMI correlated with CLS-B positivity (Table 1). Specifically, CLS-B were present in 43 of 48 (90%) obese patients, 39 of 73 (53%) overweight patients, and 40 of 116 (34%) patients with BMI < 25 (P < 0.001; Fig. 1C). The correlation between CLS-B positivity and BMI remained statistically significant in the multivariate model which adjusted for age, menopausal status, and use of NSAIDs or statins (P < 0.001).

The severity of breast WAT inflammation was quantified as number of CLS-B/cm² in 233 patients who had measurable WAT area. Among these patients who also had breast WAT inflammation, median CLS-B/cm² was 0.38 (range, 0.05–64.44). Using this median as a cutoff, the severity of breast WAT inflammation was categorized as severe (>0.38 CLS-B/cm²) in 59 of 233 (25%) patients and mild (≤0.38 CLS-B/cm²) in 59 of 233 (25%) patients (Table 2). Rising BMI was associated with more severe breast WAT inflammation (P < 0.001; Table 2). Specifically, among women with CLS-B, breast WAT inflammation was severe in 29 of 41 (71%) obese women, 17 of 38 (45%) overweight women, and 13 of 39 (33%) patients with BMI < 25 (P = 0.003). Strikingly, among women with breast WAT inflammation, the severity of inflammation was greater in those with hypertension (P < 0.05) or diabetes (P = 0.05; Table 2). Use of NSAIDs or statins did not vary based on the severity of breast WAT inflammation. Use of preoperative hormonal therapy was highest in those with severe breast WAT inflammation, but this difference was not statistically significant (P = 0.08; Table 2). Use of preoperative chemotherapy was more frequent in patients who did not have severe breast WAT inflammation (P = 0.03; Table 2); however, the usage of preoperative chemotherapy did not vary with the presence or absence of inflammation (Table 1). Among patients tested for BRCA mutations, BRCA mutations were found in a greater number of patients with severe breast WAT inflammation, but this difference was not statistically significant (P = 0.06; Table 2).

We next investigated the relationship between adipocyte size in the breast and the presence and severity of breast WAT inflammation in the first 101 patients enrolled. In this subset, we found a positive correlation between mean adipocyte size (μ) and BMI (ρ = 0.63, P < 0.001; Fig. 2A). Mean adipocyte size was larger in patients with CLS-B (107 μ; SD, 12 μ) versus those without CLS-B (92 μ, SD, 16 μ, P < 0.001, Fig. 2B). In addition, increasing severity of breast WAT inflammation was associated with larger adipocyte size (P < 0.001, Fig. 2C).

We next evaluated the rate of CLS-B positivity and the severity of breast WAT inflammation in pre- versus postmenopausal women. The postmenopausal state was associated with the presence of breast WAT inflammation (P = 0.008; Table 1; Fig. 1D). Specifically, CLS-B occurred in 58 of 93 (62%) postmenopausal women versus 64 of 144 (44%) premenopausal women. In addition, the postmenopausal state was associated with more severe inflammation as evidenced by greater number of CLS-B/cm² (P < 0.001; Table 2). These associations remained statistically significant in both multivariable models for the presence/absence of CLS-B (P = 0.03) and for severity of WAT inflammation in patients with CLS-B (P = 0.01). When categorized by menopausal status and BMI, overweight/obese postmenopausal women were the most likely to have breast WAT inflammation (P < 0.001; Fig. 1E). Similarly, overweight/obese postmenopausal women had the highest incidence of severe breast WAT inflammation (P = 0.003; Fig. 1F).

In light of these findings, and given the association between larger adipocyte size and CLS-B positivity, we examined breast adipocyte size in pre- versus postmenopausal women. Postmenopausal women had larger mean adipocyte size (105 μ, SD, 14 μ) than premenopausal women (96 μ, SD, 16 μ, P = 0.002; Fig. 2D). When grouped by menopausal status and BMI, overweight/obese postmenopausal women had the largest mean adipocyte size in the breast (P < 0.001; Fig. 2E).

Given that obesity is a risk factor for the development of HR-positive breast cancer in postmenopausal women (4, 5), we next examined the relationship between WAT inflammation and HR-positive breast cancer. Among women with invasive breast cancer inclusive of all BMIs and menopausal status, we did not observe an association between CLS-B and HR-positive tumors (Tables 1 and 2). When stratified by BMI and menopause status, the incidence of HR-positive, HER2-nonamplified breast cancer was highest in obese postmenopausal women (84.2%, P = 0.04). Moreover, the incidence and severity of WAT inflammation were highest in obese postmenopausal women (P < 0.001 for both).

As obesity is associated with low-grade systemic inflammation characterized by elevated levels of circulating proinflammatory mediators (4, 11), we investigated whether WAT inflammation occurs simultaneously at the tissue level among different adipose depots. Bilateral breast WAT was obtained from a subset of 63 women undergoing bilateral mastectomy. In this subgroup, there was a high rate of concordant CLS-B status (positive or negative) between bilateral breasts. Specifically, 49 of 63 (78%) women had concordant CLS-B status between breasts, whereas 14 of 63 (22%) had discordant status (Table 3). Notably, the median number of CLS-B/cm² was greater in women who had WAT inflammation present in both breasts (0.6, range, 0.2–51.9) versus women with WAT inflammation in one breast but not the other (0.2, range, 0.1–0.5; P < 0.001; Table 3).

Abdominal subcutaneous WAT was obtained from a subset of 13 women undergoing mastectomy with immediate autologous flap reconstruction. In this subgroup, 10 of 13 (77%) had evidence of CLS in abdominal WAT. Similar to findings in women with bilateral breast WAT, there was a high rate of concordant CLS status between breast and abdominal subcutaneous WAT. Specifically, 10 of 13 (77%) women had concordance in the presence or absence of CLS in breast and abdominal WAT (Table 3).

In the current study, we demonstrate that the postmenopausal state is independently associated with breast WAT inflammation. Our findings in breast tissue are consistent with recent reports linking the postmenopausal state to systemic inflammation (25, 26). For example, circulating levels of CRP are higher in postmenopausal compared with premenopausal women (25). In a nested case–control study, Gross and colleagues demonstrated that increased circulating levels of soluble TNF receptor 2 (TNF-R2), a receptor that binds TNF and other cytokines, is associated with higher risk of developing breast cancer in postmenopausal women (26). In addition, systemic inflammation characterized by elevated levels of prostaglandin E₂ (PGE₂) metabolite (PGE-M) in the urine is associated with a greater risk of breast cancer after menopause (29, 30). Here, we provide the first evidence that postmenopausal women, compared with their premenopausal counterparts, have larger adipocytes and greater prevalence and severity of WAT inflammation in the breast itself. These findings are consistent with prior evidence that estrogen has a strong inhibitory effect on the expression of key lipogenic genes and that larger adipocytes appear to be more likely to die leading to WAT inflammation (31, 32).

Breast cancers that express the estrogen receptor (ER) and progesterone receptor represent the most common subtype of the disease including in postmenopausal women. After the ovaries cease to produce estrogen, circulating levels of estrogens drop significantly compared with the premenopausal state. Despite this relative estrogen depletion, the incidence of estrogen-dependent tumors rises with age, accounting for nearly 85% of breast cancers diagnosed in women 80 years and older (33). Given that WAT inflammation is associated with increased amounts of aromatase in the breast (23, 34), the increased presence and severity of breast WAT inflammation after menopause provides a plausible explanation for the paradoxical observation that hormone-dependent tumors occur with increasing frequency in the setting of falling circulating estrogen levels. Specifically, locally produced estrogens as a result of WAT inflammation may explain the increased incidence of estrogen-dependent tumors in postmenopausal women. The mechanisms of breast carcinogenesis could include ER-mediated activation of multiple signaling pathways that affect cell proliferation and apoptosis (35). Genotoxic estrogen metabolites could also be important. Another possibility is that ligand-independent activation of ERα could occur in the context of inflamed WAT. In support of the potential link between WAT inflammation and estrogen-dependent breast cancers, we found that the incidence and severity of breast WAT inflammation were highest in obese postmenopausal women, the group with the highest incidence of HR-positive breast cancer.

We also confirmed that WAT inflammation occurs commonly in obese and overweight women, and that inflammation is present in a minority of women who are under- or normal weight. Strikingly, 90% of obese women had CLS-B, and the presence of WAT inflammation was associated with larger adipocyte size. These results are consistent with the findings in our original pilot study of 30 women (23). Although the postmenopausal state is associated with WAT inflammation independent of BMI, the addition of obesity represents a “two-hit” phenotype as obese, postmenopausal women had the highest rate and greatest severity of breast WAT inflammation. Several recent studies have suggested a link between metabolic syndrome and an increased risk of breast cancer (36–39). WAT inflammation is now recognized to be an important component of obesity-related disorders that occur as part of the metabolic syndrome including diabetes mellitus and cardiovascular diseases (20). Notably, in the current study, women with diabetes or hypertension were more likely to have severe breast WAT inflammation. Possibly, WAT inflammation in the breast itself will help to explain the link between metabolic syndrome and increased risk of breast cancer.

We previously quantified the severity of breast WAT inflammation using a CLS-B index and demonstrated that this index rises with increasing BMI and correlates more strongly with aromatase activity than does BMI (23). Here, we refined our method to assay the severity of breast WAT inflammation by determining the density of CLS-B quantified as number of CLS-B/cm². In addition to the increased prevalence, the severity of breast WAT inflammation was worse in obese and overweight women compared with lean women. Importantly, however, although the majority of obese women had breast WAT inflammation, 5 of 48 (10%) obese women in our study did not. Furthermore, 40 of 116 (34%) under- and normal weight women (BMI < 25) were found to have breast WAT inflammation. Thus, the presence and severity of CLS-B are not merely surrogates of BMI. The absence of breast WAT inflammation in a minority (10%) of obese women and its presence in 34% of under- or normal weight women illustrate this and raise additional issues. First, these findings are consistent with the observation that a small minority of obese individuals, as defined by BMI, are metabolically healthy (40–42). Second, a subset of lean individuals are known to be metabolically unhealthy (43). Based on our findings, it will be worthwhile to determine in normal sized individuals if WAT inflammation is associated with metabolic abnormalities. This should enable us to better understand the contribution of WAT inflammation to the pathogenesis of breast cancer in lean as well as in obese women.

We also found evidence that WAT inflammation is systemic, occurring simultaneously in multiple fat depots including the contralateral breast and abdominal subcutaneous tissue. This observation is consistent with growing evidence that obesity is associated with subclinical, systemic low-grade inflammation (4, 11, 18). The concordance of inflammatory status between tumor-bearing and tumor-free breasts demonstrates that WAT inflammation occurs independent of pre-existing neoplasia. Importantly, the concordant inflammatory status (present or absent) between the breast and abdominal subcutaneous WAT strengthens the rationale for analyzing abdominal WAT, a more clinically accessible depot, in future studies of interventions that aim to attenuate WAT inflammation. Moreover, the fact that the breast is a sentinel for adipose inflammation in other fat depots raises the intriguing possibility that this diffuse low-grade inflammatory state will be associated with blood biomarkers. Future studies will be needed to determine if blood biomarkers of CLS-B can be identified.

Our study is strengthened by a prospective tissue acquisition design and tissue assessment by a histopathologist experienced in the detection and grading of WAT inflammation. To our knowledge, this is the largest study of human breast WAT to date. However, a potential limitation of this study is the inclusion of patients with cancer. Although we have attempted to control for potential confounding factors, similar studies should be carried out in individuals without cancer. Obtaining sufficient amounts of breast tissue from cancer-free individuals to assess adipose inflammation poses practical limitations especially if questions concerning menopause are to be addressed. Importantly, there is no evidence that WAT inflammation occurs as a consequence of neoplasia, and, on the contrary, WAT inflammation has been well described in obese, cancer-free individuals including within nontumorous breasts (18, 19, 24). If a noninvasive blood biomarker of WAT inflammation can be developed, future studies in larger cohorts of cancer-free individuals may be conducted. This will be particularly important if such a biomarker is to be evaluated for predicting risk and/or outcomes.

In conclusion, this study demonstrates that the postmenopausal state is independently associated with the presence and severity of breast WAT inflammation. We also confirm that WAT inflammation occurs in most obese women and some who are lean. Given that WAT inflammation is associated with elevated aromatase levels, our study provides a potential mechanistic explanation for the increased incidence of estrogen-dependent breast cancers in obese, postmenopausal women. WAT inflammation, manifested as CLS-B, may represent a more robust biomarker of risk than BMI and could provide a more specific pathological target that can be used for the development of effective prevention and treatment strategies. Interventions that target WAT inflammation, including diet, exercise, medications, or a combination thereof, may prove useful for breast cancer prevention and treatment.

No potential conflicts of interest were disclosed.

Conception and design: N.M. Iyengar, X.K. Zhou, A. Gucalp, D. Giri, L.T. Vahdat, K. Subbaramaiah, C.A. Hudis, A.J. Dannenberg

Development of methodology: N.M. Iyengar, P.G. Morris, A. Gucalp, D. Giri, C.A. Hudis, A.J. Dannenberg

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.M. Iyengar, P.G. Morris, A. Gucalp, D. Giri, D.J. Falcone, M. Morrow, C.A. Hudis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.M. Iyengar, P.G. Morris, X.K. Zhou, C.A. Hudis, A.J. Dannenberg

Writing, review, and/or revision of the manuscript: N.M. Iyengar, P.G. Morris, X.K. Zhou, A. Gucalp, D. Giri, L.T. Vahdat, K. Subbaramaiah, M. Morrow, C.A. Hudis, A.J. Dannenberg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.M. Iyengar, A. Gucalp, D. Giri, M.D. Harbus, M.D. Krasne, C.A. Hudis, A.J. Dannenberg

Study supervision: N.M. Iyengar, A. Gucalp, C.A. Hudis, A.J. Dannenberg

The authors thank all the patients who consented to provide tissue for this study. They also thank the staff of the Clinical Trials Office and Pathology Core at the Memorial Sloan Kettering Cancer Center.

This work was supported by grants NIH/NCI R01CA154481 (to A.J. Dannenberg), UL1TR000457 of the Clinical and Translational Science Center at Weill Cornell Medical College (to N.M. Iyengar and X.K. Zhou), 2013 Conquer Cancer Foundation of the American Society of Clinical Oncology (ASCO) Young Investigator Award (to N.M. Iyengar), 2013 Royal Academy of Medicine in Ireland, St. Luke's Young Investigator Award (to P.G. Morris), 2012 Expedition Inspiration (Brenda M. Williams) Young Investigator Award (to P.G. Morris), the Botwinick-Wolfensohn Foundation (in memory of Mr. and Mrs. Benjamin Botwinick; to A.J. Dannenberg), and the Breast Cancer Research Foundation (to A.J. Dannenberg and C.A. Hudis). This work was presented in part at the 2014 ASCO Annual Meeting, resulting in a 2014 Conquer Cancer Foundation of ASCO Merit Award (to N.M. Iyengar).

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.