Adipokines in Plasma and Breast Tissues: Associations with Breast Cancer Risk Factors

Higher body weight is consistently associated with poorer prognosis and increased risk of postmenopausal breast cancer (1–4), although the molecular mechanisms for these associations remain unclear. The relationship between obesity and breast cancer may be partially explained by the biologic effects of the adipokines leptin and adiponectin, secreted by adipocytes (5–10). Plasma leptin is positively correlated with body mass index (BMI; refs. 11, 12) and has been shown to be positively associated with breast cancer in some studies (13–16), inversely associated in one study (17), and not associated in others (18–22). Plasma adiponectin is inversely correlated with BMI (23–25) and has consistently been shown to be inversely associated with breast cancer (13, 26–28), except for in one study (29).

Leptin and its receptor are found in normal mammary tissue and in breast tumors (30, 31). Leptin is a key growth factor and may play a significant role in breast carcinogenesis, promoting cell growth and proliferation (32–34). Once bound to its receptor, leptin induces the activation of several signaling pathways, including the JAK/STAT, mitogen-activated protein kinase (MAPK), IRS1, and SOCS3 pathways, which modulate cell proliferation and survival in breast epithelial cells (32, 34). Leptin also enhances activation of ERα through the MAPK pathway in MCF-7 cells and increases the estradiol-induced activation of ERα (35). In addition, leptin is known to promote angiogenesis (36, 37). In contrast, the function of adiponectin is less understood, but its biologic effects generally oppose those of leptin. For example, adiponectin inhibits cell proliferation and promotes apoptosis (7, 10, 38). The inverse association between BMI and adiponectin makes it plausible that adiponectin too may play a direct role in breast carcinogenesis.

In previous epidemiologic studies investigating associations between adipokines and breast cancer, investigators have assumed that blood adipokine concentrations are good surrogates of tissue exposure. To our knowledge, there are no studies that address this assumption. In this cross-sectional study of women with no history of breast cancer, undergoing reduction mammoplasty, we examined the association between blood and breast adipokine concentrations and determined what factors influenced blood and breast tissue concentrations of these hormones.

Detailed methods are given in Dumitrescu and colleagues (39). Briefly, between 1997 and 2009, 155 women who underwent elective reduction mammoplasty surgery at Georgetown University Medical Center (Washington, DC), the University of Maryland (College Park, MD), the Washington Hospital Center (Washington, DC), and the Center for Plastic Surgery (Buffalo, NY), were included in this study. Any reduction mammoplasty patient greater than age 18 (or 16–18 years where a parent/guardian was available to provide consent) with no history of any cancer was eligible for participation. Women with a psychiatric disorder or any other reason that precluded understanding the informed consent was considered ineligible. Within 24 hours before surgery, women answered a detailed baseline interview that queried on demographics, lifestyle, medication use, and other recent exposures. A more extensive interview was administered to 132 subjects, either at that time or shortly after surgery, to assess personal and family medical history, occupation, diet and alcohol, smoking history, and reproductive history. If a participant was eligible and willing to participate, but a personal interview was not possible for logistical reasons, relevant data were obtained from the medical record (n = 23). Thus, the present study included 155 participants who underwent reduction mammoplasty. All study participants signed an informed consent before study enrollment and the study received Institutional Review Board approval from all participating institutions.

Race was self-reported. Postmenopausal status was defined as not having a menstrual period in the last 12 months. BMI was calculated as self-reported weight divided by self-reported height squared (kg/m²). Family history of cancer was defined as the occurrence of cancer in any first- or second-degree blood relative. Oral contraceptive (OC) users were defined as those who had used OCs for a period of at least 1 year. Smokers were defined as women who had ever smoked more than 100 cigarettes. Drinkers were defined as those who had ever consumed 12 or more alcoholic beverages.

Among the 155 study participants, surgically removed breast tissues were collected from 94 women. The tissues were inspected and determined to be free from gross abnormalities by board-certified pathologists. Women found to have benign breast disease (n = 9) on microscopic examination were excluded; thus, a subset of the study sample was included in the breast tissue-specific analysis (n = 85). Epithelial/stromal tissues were dissected (to remove adipose tissues) and snap frozen in liquid nitrogen within one hour of surgery. Aliquots of tissues were stored at −80°C until analysis. Heparinized plasma samples were collected, processed, aliquoted, and stored at −80°C until analysis.

Plasma leptin and adiponectin concentrations were determined by Human Leptin Quantikine and Human Adiponectin/Acrp30 Quantikine ELISA kits (R&D Systems) according to manufacturer's instructions. Samples were assayed blindly, in duplicate, and in random order. Each batch included replicates, commercial controls, and blinded plasma controls to assess laboratory variation. The coefficients of variation (CV) for plasma assays were 5.92% and 4.48% for leptin and adiponectin, respectively. Assay sensitivity was less than 7.8 pg/mL for leptin and 0.08 ng/mL for adiponectin. There were no samples below the limits of detection.

Breast leptin and adiponectin concentrations were also assayed using the ELISA kits from R&D Systems. Frozen tissues (0.80 g starting weight) were homogenized in 5 mL of ice-cold extraction solution containing T-PER buffer (contains a proprietary detergent in 25 mmol/L bicine, 150 mmol/L sodium chloride; pH 7.6) and protease inhibitors (Pierce). Each sample was homogenized in Lysing Matrix D tubes using the FastPrep-24 Instrument (MP Biomedicals) at 6.0 m/s for 6 cycles of 30 seconds. Homogenates were centrifuged at 2,800 rpm for 10 minutes and supernatants were collected, aliquoted, and stored at −80°C. Total protein concentration was quantified using the Coomassie Plus-The Better Bradford Assay Kit (Pierce). Breast hormone levels were normalized by total protein concentrations and reported in nanograms of hormone per milligram of total protein (ng/mg). Samples were assayed blindly, in duplicate, and in random order. Each batch included replicates, commercial controls, and breast homogenate controls to assess laboratory variation. CVs for breast assays were 9.24% and 9.33% for leptin and adiponectin, respectively.

Although adipose tissue is dissected from epithelia and stroma at the time of specimen collection, it is not a perfect process. Thus, breast tissue heterogeneity was assessed to determine whether differing amounts of adipose versus epithelial/stromal tissues (on hematoxylin and eosin slides) affected adipokine concentrations. The heterogeneity index was scored as follows: 1 (no epithelia/stroma, all adipose); 2 (some epithelia/stroma, mostly adipose); 3 (equal amounts of epithelia/stroma and adipose); 4 (mostly epithelia/stroma, some adipose); and 5 (all epithelia/stroma, no adipose). On average, 5 slides per participant were scored and mean score was used in analysis; no subject had a mean score less than 2. A subset of subjects (n = 15) with available breast tissues from 3 sections of the same breast (upper, center, and lower) were also studied for heterogeneity of concentrations (n = 45). Each frozen breast tissue specimen was homogenized according to the tissue protein homogenization protocol described above; however, the starting weight for this subset study was 0.40 g and samples were homogenized in 2.5 mL extraction solution.

Mean and frequencies were used to assess the distribution of sample characteristics. Student t tests and analysis of variance (ANOVA) were used to evaluate differences in adipokine concentrations by categorical variables and Pearson correlation coefficients were used to describe associations between plasma and breast adipokine concentrations. Plasma and breast adipokine concentrations were right-skewed and therefore were cube root transformed for normality. Cube root-transformed data were used in all comparisons; however, we present back-transformed findings for ease interpretation.

Multiple linear regression models were used to investigate associations between adipokine concentrations and demographic, reproductive, and lifestyle variables. The following main effects were considered for inclusion in models: age (continuous), menopausal status (premenopausal/postmenopausal), BMI (kg/m²), family history of any cancer (no/yes), family history of breast cancer (no/yes), parity (nulliparous, 1–2, >2 children), years of OC use (never user, ≤5, >5 years), duration of alcohol use (never drinker, ≤17, >17 years), smoking status (never, former, current smoker), and quartiles of plasma adipokines [plasma leptin (breast leptin model) and plasma adiponectin (breast adiponectin model)].

We hypothesized a priori that there would be interaction effects between BMI and race, as well as between BMI and menopausal status on adipokine concentrations. Interactions were tested by including cross-product terms in multivariable regression models. We used backward elimination (40) to identify significant variables for each adipokine model; α of 0.20 or more was required for inclusion in final models. First, we fit the full model and computed partial F tests for each variable. We then removed the variable with the largest P value (if >0.20) and fit the reduced model and computed partial F tests for each variable in the reduced model. This was repeated until all variables not reaching significance at P ≤ 0.20 were eliminated.

For all analyses, P values were 2-sided and P < 0.05 was considered statistically significant. Analyses were conducted using SAS version 9.2 (SAS Institute) and STATA version 12 (StataCorp LP).

Characteristics of the 155 study participants are given overall and stratified by race in Table 1. Ninety four (61%) of the women in this study were white and 61 (39%) were black. Compared with white women, black women tended to be younger, were more likely to be obese, and to be current smokers. Black women were less likely to have a family history of cancer and drink alcohol. Black women also tended to have higher plasma and breast leptin and lower adiponectin concentrations. Separately, differences in participant characteristics and adipokine concentrations by site of recruitment were assessed and none were found (data not shown).

BMI-adjusted plasma-breast adipokine correlations are shown in Table 2. Overall, there was a positive correlation between plasma and breast concentrations of leptin (r = 0.41, P = 0.0002) and a borderline positive correlation for adiponectin (r = 0.23, P = 0.05). We conducted exploratory subset analyses by BMI, race, and menopausal status based on a priori hypotheses of subgroup differences. In analyses stratified by BMI, we observed a positive correlation between plasma and breast leptin concentrations, which was strongest among normal weight women (r = 0.62, P = 0.009) and attenuated among overweight and obese women. For adiponectin, the positive plasma-breast correlation was strongest among obese women (r = 0.31, P = 0.04) and attenuated among normal and overweight women. Among white women, we observed positive plasma–breast correlation for both leptin and adiponectin (r = 0.28, P = 0.06 and r = 0.30, P = 0.04, respectively), whereas among black women there was no association for either leptin or adiponectin. Among premenopausal women, we observed no significant plasma-breast adipokine concentrations. Conversely, we observed positive plasma–breast adiponectin correlation among postmenopausal women (r = 0.49, P = 0.03). These data indicate that plasma–breast adipokine correlations may vary by subgroup, although due to small sample size, these analyses may have been underpowered to detect significant differences.

BMI-adjusted associations between participant characteristics and plasma adipokines, including the plasma ratio of adiponectin to leptin (A/L), are provided in Table 3. Mean plasma adiponectin differed by age, race, and menopausal status. Plasma adiponectin was 54% higher in whites than blacks (P < 0.0001) and 31% higher among postmenopausal than premenopausal women (P = 0.02). Mean plasma adiponectin concentration increased with increasing years of alcohol consumption, with 34% higher plasma levels among women who consumed alcohol for more than 17 years compared with never consumers (P = 0.04). Similarly, mean plasma A/L levels differed by race and alcohol consumption (P < 0.0001 and P = 0.01, respectively), although these findings were driven almost exclusively by differences in plasma adiponectin. No statistically significant differences were observed between participant characteristics and plasma concentrations of leptin.

Breast leptin and adiponectin concentrations did not differ by breast location (P values = 0.70 and 0.88, respectively; data not shown) and there were no statistically significant correlations between relative amounts of epithelial/stromal to adipose tissues and breast leptin or adiponectin (r = −0.003, P = 0.98 and r = 0.005, P = 0.97, respectively; data not shown).

There was large interindividual variation for breast adipokines, with levels ranging from 0.05 to 23.13 ng/mg for leptin and 35.50 to 3,023.30 ng/mg for adiponectin. BMI-adjusted associations between participant characteristics and breast tissue concentrations of adipokines and A/L are provided in Table 4. Breast concentration of adiponectin was inversely associated with any use of OCs (P = 0.02), mean breast adiponectin was 71% and 49% lower among women using OCs for 5 or more years and more than 5 years, respectively, than never users (P = 0.02). Mean breast leptin concentration was 114% lower among former than never smokers (P = 0.01). No other participant characteristics were associated with breast adiponectin or leptin concentrations and no characteristics were associated with A/L in the breast.

Multivariable-adjusted models of plasma adipokine concentrations are shown in Table 5. BMI, race, BMI × race, parity, years of OC and alcohol use, and smoking status were shown to explain 50% of the variation in plasma leptin. Adjusting for all covariables, among whites, plasma leptin increased 2.18 ng/mL with each one kg/m² increase in BMI (P = 0.004), whereas among blacks, plasma leptin increased 1.43 ng/mL with each single-unit (kg/m²) increase in BMI (P < 0.0001). Current smokers had plasma leptin concentrations that on average were 7.07 ng/mL lower than never smokers (P = 0.04), controlling for all covariables. BMI, race, menopausal status, BMI × menopausal status, parity, years of OC use, and smoking status explained 29% of plasma adiponectin variation. After adjusting for covariables, plasma adiponectin, on average, was 2.04 μg/mL lower in blacks than whites (P = 0.02) and decreased 0.24 μg/mL for each kg/m² increase in BMI (P = 0.007) among premenopausal women. Plasma adiponectin was also higher among OC users than never users (P = 0.02).

Multivariable-adjusted models of breast adipokine concentrations are also shown in Table 5. BMI, race, BMI × race, years of OC use, and smoking status explained 30% of breast leptin variation. After adjusting for all covariables, breast leptin increased 0.29 ng/mg with each single-unit (kg/m²) increase in BMI (P = 0.001) and former and current smokers had concentrations 1.91 and 1.20 ng/mg lower, respectively, than never smokers (P = 0.007). Of note, plasma leptin was not found to be a significant predictor of breast leptin concentrations through partial F test. In addition, when quartiles of plasma leptin were included in the final breast leptin model (reported above), the adjusted R² decreased from 0.50 to 0.33. BMI, years of OC use, smoking status, and plasma adiponectin explained 20% of breast adiponectin variation. Breast adiponectin increased 19.91 ng/mg with each single-unit increase in BMI (P = 0.05). Compared with never users, those who used OCs for 5 years or more had breast adiponectin concentrations that were 446.74 ng/mg lower and those using them for more than 5 years had breast adiponectin concentrations that were 363.99 ng/mg lower (P = 0.02). Controlling for all other covariables in the breast adiponectin model, breast adiponectin increases monotonically as plasma adiponectin increases (P trend = 0.03). Nonetheless, only the highest quartile of plasma adiponectin was significantly associated with breast adiponectin concentration (P = 0.02).

In this first report of adipokine concentrations in histologically normal breast tissues from healthy women, it was found that there was wide interindividual variation in breast leptin and adiponectin, inferring the potential for these adipokines to be associated with breast cancer in some women but not others. Importantly, we observed modest positive correlations between plasma and breast adipokine concentrations, which for leptin was the strongest in women with BMI less than 25, whereas for adiponectin, the correlation was strongest in those with BMI ≥30 or more. Therefore, it is uncertain whether or not plasma adipokines are good surrogates for breast adipokines among all women. Through multivariable-adjusted regression analysis, we identified factors that explain 20% to 30% of the variation in breast leptin and adiponectin. Among all of the breast cancer risk factors investigated in this study, we found that BMI and race explained most of the observed variation in breast leptin, whereas OC use and smoking status explained most of the observed variation in breast adiponectin.

In this study, plasma leptin levels were similar to levels reported in other (17, 18, 22) studies. Prior studies investigating the association between plasma leptin and breast cancer have yielded inconsistent data (13–22). Some have reported no association (18–22), others a positive association (13–16), and still others an inverse association (17). Positive studies, interestingly, were in Asian populations (13–16), where women on average had lower BMI (23–24 kg/m²) than those observed in other studies, such as in the United States. Given the findings reported herein that plasma and breast leptin levels are correlated only in women with a BMI less than 25.0 kg/m² (r = 0.62, P = 0.009), this may indicate that plasma leptin level is a breast cancer risk factor for women who are not overweight or obese.

In this study, plasma adiponectin concentrations were similar to prior reports (23, 27, 29, 41) and are shown to be linearly associated with breast adiponectin concentrations (P trend = 0.03). In contrast to leptin, epidemiologic data are relatively consistent for an inverse association between circulating adiponectin and breast cancer (13, 26–28). Given the suggestion of a positive correlation between plasma and breast adiponectin observed in this study, the inverse association for adiponectin and breast cancer risk may be due to local effects in the breast tissue rather than an inverse correlation for plasma and breast levels. In the Nurse's Health Study, a large prospective cohort of women, Tworoger and colleagues (28) reported a 27% reduction in breast cancer risk when comparing the highest to lowest quartile of plasma adiponectin among postmenopausal [RR, 0.73; 95% confidence interval (CI), 0.55–0.98], but not premenopausal women (RR, 1.30; 95% CI, 0.80–2.10). Another study noted this association being stronger among postmenopausal women as well (26). Also, the association between BMI and breast cancer risk is limited mostly to postmenopausal women (1–4). Thus, there may be a complicated association for BMI and menopausal status contributing to breast carcinogenesis. Consistent with the epidemiologic studies, in subset analyses, we observed differences in the plasma-breast adiponectin correlation by menopausal status (r = 0.14, P = 0.30 and r = 0.49, P = 0.03, for premenopausal and postmenopausal women, respectively).

To our knowledge, adipokines have not been previously measured in histologically normal breast tissues; but they were previously investigated in benign and malignant neoplasms as well as adjacent tissues (42, 43). Karaduman and colleagues (43) found that adiponectin concentrations in malignant breast tumors were higher than in benign breast fibroadenomas (mean ± SD, 0.75 ± 0.06 vs. 0.68 ± 0.10 μg/mL, P = 0.02). In another study, the same authors reported that leptin concentrations in malignant tumors were higher than in adjacent normal appearing tissues (5.02 ± 1.06 vs. 2.03 ± 0.83 ng/mL, P = 0.01; ref. 42). The finding that adiponectin is higher in breast tumors, rather than lower as might be expected on the basis of the apparent inverse association between plasma adiponectin and breast cancer (13, 26–29), may support the hypothesized positive association between breast adiponectin and breast cancer, opposite of the inverse plasma association shown by others (13, 26–28).

Multivariate analysis was conducted that included plasma adipokines as predictor variables to explore the factors that might explain why the levels of breast adipokines differ from plasma adipokines. The results were that among the variables investigated in this study, BMI and race explained most of the variation observed in breast leptin [partial R² = 0.26 (all covariables explained 30% of the variation)], whereas OC use and smoking status explained most of the variation in breast adiponectin [partial R² = 0.13 (all covariables explained 20% of the variation)]. Separate from these breast cancer risk factors, however, there remains little information about other predictors and the association of these and others within the mechanism for different partitioning of adipokine levels between the plasma and breast. It may be that because this study dissected adipose tissue from epithelia and stroma, the partitioning ratio for plasma to breast might have been different had this not been done; there might have been a stronger correlation and lower partitioning ratio given that leptin and adiponectin are produced by adipocytes. Future studies are necessary to understand the underlying mechanisms, as well as to identify additional factors that are associated with adipokine concentrations in target tissues.

There were several strengths of this study. The cross-sectional investigation of plasma and breast adipokine concentrations in healthy women without breast cancer as well as without benign breast disease provides information about women before cancer develops. Our use of highly reproducible assays (i.e., with low CVs) and assessment of breast tissue heterogeneity also strengthened this study. Given the confirmation for the reliability of sampling across the breast, this study provides novel data about variation in adipokines in the target organ from subjects with no history of cancer.

As with any study, there were limitations that should be considered in the interpretation of our findings. We recognize that the sample size was somewhat small; nonetheless, there was significant adipokine variability that can be further investigated in larger studies. Separately, women who undergo reduction mammoplasty may not represent all women. Particularly, they have larger breasts and higher BMI and socioeconomic status, and may have increased breast cancer risk due to large breast size (44, 45). However, the analysis done here includes women with a wide range of BMI and within group comparisons strengthens the validity of the results. This study may suffer from selection bias, which cannot be assessed given that there are no data about women who did not participate in this study. Another limitation relates to the epidemiologic data that were collected. Lack of information for additional confounders, as well as a lack of information pertaining to recency of certain exposures (e.g., alcohol and OC use) could have affected the results. Finally, regarding the tissue-specific analyses, there may have been heterogeneity in the proportions of cell types within and among individual breast tissue samples that could impact adipokine concentrations. However, the validation study presented here for different parts of the breast suggests that where the breast was sampled did not affect our results. Furthermore, the ratio of epithelia/stroma to adipose was shown to have no bearing on breast adipokine concentrations.

In conclusion, we found homogeneity of breast adipokine concentrations across the breast and considerable interindividual variation in breast adipokine concentrations. Although only 20% to 30% of the variation in breast leptin and adiponectin could be explained by factors investigated in this study, these data suggest that BMI and race may be important determinants of breast leptin, whereas years of OC use and smoking status may be important determinants of breast adiponectin. Importantly, the data presented herein indicate that although overall plasma leptin concentrations are correlated with leptin concentrations within the breast, plasma leptin may not be a good surrogate for breast leptin among all women. Nonetheless, these findings are suggestive of positive plasma–breast correlation among normal weight women. Similarly, plasma adiponectin may not be a good surrogate for adiponectin concentrations in the breast, although the data presented herein indicate the possibility of a positive plasma–breast correlation for this hormone as well. It is evident that a clearer understanding of the associations between the easily measured plasma adipokines and the likely more pertinent levels within the target organ is necessary to better understand the impact of these hormones on breast cancer risk and prognosis and their implications for breast cancer prevention.

Amy E. Millen has a commercial research grant from Mushroom Council. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.A. Llanos, J.L. Freudenheim, P.G. Shields

Development of methodology: A.A. Llanos, R.G. Dumitrescu, B.V.S. Kallakury, J.L. Freudenheim, P.G. Shields

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.G. Dumitrescu, B.V.S. Kallakury, D.J. Perry, R.J. Convit, M.E. Platek

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.A. Llanos, R.G. Dumitrescu, C. Marian, K.H. Makambi, B.V.S. Kallakury, J.L. Freudenheim, P.G. Shields

Writing, review, and/or revision of the manuscript: A.A. Llanos, R.G. Dumitrescu, C. Marian, K.H. Makambi, B.V.S. Kallakury, D.J. Perry, R.J. Convit, M.E. Platek, A.E. Millen, L.L. Adams-Campbell, J.L. Freudenheim, P.G. Shields

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.G. Dumitrescu

Study supervision: C. Marian, D.J. Perry, P.G. Shields

Provided material for the research: S.L. Spear

The authors thank all of the participating plastic surgeons at the Georgetown University Medical Center, the University of Maryland, the Washington Hospital Center, and the Center for Plastic Surgery in Buffalo, New York as well as their staff for assistance in identifying potential participants and informing them of the Reduction Mammoplasty Study.

This work was supported by the Department of Defense (grant DOD BC022346).

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.