Many established risk factors for breast cancer are related to endogenous, lifetime estrogen exposure [1, 2]. The relationship between elevated levels of serum steroid hormone and risk of postmenopausal breast cancer has been established in many well-conducted epidemiological studies [38]. In premenopausal breast cancer, associations between circulating ovarian steroid levels and risk have been more difficult to determine, largely due to the extreme fluctuations in hormone levels related to the ovarian cycle.

Breast tissue contains all of the enzymes required for the synthesis of estradiol from androgen and estrogen sulfate precursors. There is increasing realization that the local hormonal environment in the breast can differ from that in the serum, and that the levels of estradiol, its precursors, and progesterone are higher in ductal fluid than those in serum [9]. The local production of steroid hormones in the breast is a more stable contributor to the hormonal environment [10]. Nipple aspiration fluid (NAF) provides a window into this environment and allows the measurement of hormone and protein content, which may have stronger relations to breast cancer risk than corresponding serum concentrations [11]. Besides hormones in the breasts, many protein markers have also been found to be related to the presence of breast cancer when tumor-bearing breasts were compared with the normal breasts of unaffected women [1224]. In addition, several proteins have been recognized as candidate cancer risk markers in the high-risk contralateral unaffected breasts of breast cancer patients with increased levels. These proteins are known to be growth factors, proteases, and markers of inflammation and oxidative stress in the breast; they include basic fibroblast growth factor (bFGF) [2527], cathepsin D [28], C-reactive protein (CRP) [2931], YKL40 (chitinase-3-like protein 1, also known as human cartilage glycoprotein-39) [32], and superoxide dismutase (SOD-1) [33], among others.

In previous work, we have shown that the NAF concentrations of some estrogen response proteins (cathepsin D and epidermal growth factor EGF) are significantly correlated with those of estradiol and its precursors in NAF, and no significant relation between serum hormone levels and nipple fluid proteins D [34]. These within-NAF correlations suggest that the estrogen content of NAF is biologically relevant, and that NAF composition may provide a coherent picture of the breast physiologic and pathologic environment. The relation of other protein markers to the local steroid hormone content of NAF has not been previously studied. A recent proteomic analysis showed that the proteomic profile of NAF did not change substantially during the menstrual cycle [35], even though some earlier studies identified several hormone-responsive proteins in NAF [36, 37], including apolipoprotein D (gross cystic disease fluid protein 24, GCDFP-24), cathepsin D, epidermal growth factor (EGF), prolactin-induced protein (gross cystic disease fluid protein 15, GCDFP-15), and pS2 (TFF-1). We hypothesized that the within-NAF correlations seen between NAF estrogens and cathepsin D and EGF may also be observed for other cancer-relevant NAF proteins, and have investigated the correlation between selected protein markers in NAF and both systemic (serum) and local (NAF) abundance of estradiol and its precursors to better understand the contributing role of hormones on breast cancer risk.

Materials and Methods

Study Population

We have conducted a case-control study to examine the breast cancer risk associations of NAF hormones, comparing incident cases with unilateral breast cancer to healthy screening controls seen at the Lynn Sage Breast Center of Northwestern University. Study subjects were recruited with the approved informed consent under an IRB-approved protocol. Nipple aspirate fluid (NAF) was collected from both breasts of healthy women (control group) and the contralateral unaffected breast of patients with breast cancer. Blood was collected at the same visit. In the present case-control study, we included samples from breasts of 114 subjects yielding >10 μl of NAF (sufficient for analysis of both steroids and proteins). Of these, 54 were controls (27 premenopausal and 27 postmenopausal), and 60 were cases (28 premenopausal and 32 postmenopausal).

NAF Collection

Moist heat was applied to the breasts, and 2.5 % lidocaine and 2.5 % prilocaine cream (Emla, AstraZeneca, Wilmington, DE) were applied to the nipple and areola area for anesthesia. Fluid was obtained by applying a partial vacuum with a syringe-like device (Cytyc Corp, Boxborough, MA). The fluid was collected from one or more ducts in a calibrated capillary tube (1.0 mm in length = 1.0-μl volume). The sample was placed on ice and brought to the laboratory where the sample was measured and flushed out with 200 μl of PBS into a small Eppendorf tube and was then frozen at −80 °C.

Steroid Hormone Measurement in NAF and Blood

Steroid hormones were measured according to a previous report [38]. Briefly, steroids were extracted into water-saturated ethyl acetate:hexane (3:2) containing 250 ng of the internal standard dexamethasone acetate. The extract was then applied to a 25 cm × 4.6 mm C18 reversed phase high performance liquid chromatography column and eluted with 58 % 15 mM phosphate buffer, pH 6, and 42 % of a 50:50 mixture of acetonitrile:methanol. A gradient was started at 40 min to a final concentration of 71 % of the acetonitrile:methanol solvent at 50 min. The flow rate was 1.0 ml/min. Retention times of estradiol, estrone, testosterone, androstenedione, and progesterone were 33, 38, 46, 48, and 57 min, respectively. A single fraction containing estrone and androstenedione was collected, and the fraction was divided equally for the assay of these steroids. Each 4-ml fraction was collected in an automated fraction collector, evaporated at 50 °C in a water-bath under nitrogen, and assayed by specific immunoassays. Estradiol was measured in serum by an ultrasensitive radioimmunoassay (DSL-4800) from Beckman Coulter. Estradiol is the limiting steroid concentration in these samples. The immunoassay method used for quantification of the purified product has a limit of detection of 4.0 pg/ml. With a volume of 0.1 ml in the assay vial, the mass detected is thus 0.4 pg. This is the same as that described for the frequently cited LC/MS/MS method of Xu et al. [39]. Other analytes, measured by ELISAs with catalog numbers, were estrone (DSL-8700) and androstenedione (DSL-3800) from Beckman Coulter; DHEA (20-DHEHU-E01) from Alpco, and testosterone (1-2402) and progesterone (1-502-5) from Salimetrics. Aliquots from a pool of female control serum were inserted into each batch for quality control. The intra-assay and inter-assay %CVs for the serum immunoassays were as follows: estradiol 10.4 and 17.2, progesterone 2.6 and 8.1, estrone 6.6 and 14.0, androstenedione 5.3 and 6.6, testosterone 10.5 and 16.4, and DHEA 8.7 and 10.2. All the samples were assayed in duplicate. Intra-assay and inter-assay percent coefficients of variation (CV) in NAF were as follows: estradiol 5.3 and 6.6 %, estrone 5.9 and 17.2 %, progesterone 8.1 and 16.5 %, androstenedione 6.9 and 17.2 %, testosterone 10.2 and 15.0 %, and DHEA 7.1 and 16.8 %. The sensitivity was as follows: estradiol 6.4 pg/ml, estrone 5.3 pg/ml, progesterone 64.9 pg/ml, androstenedione 13.0 pg/ml, testosterone 6.2 pg/ml, and DHEA 9.0 pg/ml. For estradiol assay, 17β-estradiol 3-glucuronide cross-reacts 2.56 %; all other steroids cross-react <1 %. For progesterone assay, among naturally occurring steroids, deoxycorticosterone cross-reacts 1.7 %; all others cross-react <1.0 %.

Protein Marker Measurement in NAF

C-reactive protein (CRP), basic FGF (bFGF), and chitinase 3-like 1 (YKL40) were measured by immunoassays using kits from R&D Systems Inc (Minneapolis, MN). Cu/Zn superoxide dismutase (SOD1) and cathepsin D (CatD) were measured with ELISA kits from Calbiochem (San Diego, CA) according to the manufacturer protocols. All the samples were assayed in duplicate. The intra-assay and inter-assay percent CVs were SOD1 8.5 and 12.5 %, CRP 4.4 and 7.0 %, YKL40 4.3 and 6.9 %, CatD 7.6 and 13.2 %, and bFGF 3.0 and 9.1 %. The sensitivity was SOD1 40.0 pg/ml, CRP 10.0 pg/ml, YKL40 3.6 pg/ml, CatD 4 ng/ml, and bFGF 3.0 pg/ml.

Cell Culture and Protein Detection

Human breast cancer cell lines MDA-MB-453 and MDA-MB-231 were obtained from the American Type Culture Collection (Manassas, VA). The cells were cultured in Dulbecco’s modified eagle’s medium (DMEM, Invitrogen, Carlsbad, CA) supplemented with 10 % fetal bovine serum (FBS) at 37 °C in a humidified CO2 incubator. Prior to the experiments, the cells were cultured in phenol-red-free DMEM (Invitrogen) for 24 h and then in serum-free DMEM overnight. The cells were treated with testosterone (Sigma) at 1 or 10 nM for 24 h and then collected for protein extraction. The modulation of YKL40 and bFGF by DHT was detected in MDA-MB-453 and MDA-MB-231 cell lines using Western blot. Anti-YKL40 and anti-bFGF antibodies were purchased from Sigma.

Statistical Analysis

The comparison between groups was performed using Wilcoxon rank sum test with Bonferroni adjustment for multiple comparisons. The correlation of protein markers with hormones and other protein markers was examined using Spearman’s rank correlation test with Bonferroni adjusted P < 0.05 as statistically significant and considered as a strong correlation.


Steroid Hormone in Serum and NAF from Breast Cancer Patients

From a total of 355 subjects that had been recruited to the study, we included 114 subjects that yielded >10 ul NAF sufficient for both steroid hormone and protein measurements. The average age was 52.1 years with a range of 30–71 years. Fifty-five (48.2 %) were premenopausal and 59 (51.8 %) postmenopausal. There were no significant differences between cases and control groups by age or menopausal status (Table 1). The measurements of steroid hormones in serum and NAF (Table 2) showed that most of the measured hormones were significantly higher in NAF, including estrone (P = 0.037), progesterone (P = 0.0010), androstenedione (P = 0.0055), and DHEA (P = 0.0001). Although the estradiol level in NAF was also higher than that in serum, it was not significantly different (P = 0.093). Only NAF testosterone level was marginally lower than serum level (P = 0.069).

Table 1 Summary of age and menopausal status of the subjects in control and case groups
Table 2 Hormone levels in serum and NAF

NAF Protein Biomarkers in Serum and NAF and the Correlation with Steroid Hormones

Protein marker concentrations were measured in NAF as shown in Table 3. There was no significant difference between the controls and cases in these protein levels, although cathepsin D tended to be higher in cases than controls (3.3 vs 2.3 ng/μl, P = 0.13). We then examined the correlation among the protein markers using Spearman’s rank correlation test with Bonferroni adjusted P < 0.05 as significant. SOD1 was significantly correlated with CRP (R = 0.276, P = 0.033) and CatD (R = 0.340, P = 0.0036). bFGF was also significantly correlated with CRP (R = 0.343, P = 0.0021, Table 4). Correlations between NAF protein markers and serum hormones (Table 5) indicated only a negative correlation between bFGF and testosterone (R = −0.338, P = 0.017). In contrast, NAF protein markers showed stronger correlation with hormones in NAF. bFGF was negatively correlated with NAF testosterone (R = −0.339, P = 0.015), to a similar degree as serum testosterone. In addition, SOD1 was correlated with NAF DHEA (R = 0.333, P = 0.019). YKL40 was correlated with NAF testosterone level (R = 0.389, P = 0.0012).

Table 3 Protein concentrations in NAF from cases and controls
Table 4 Correlation among five protein biomarkers in NAF
Table 5 Correlation between NAF protein biomarkers and hormones in serum or NAF

Regulation of YKL40 and bFGF by Testosterone in Breast Cancer Cells

In order to investigate the causal correlation between NAF testosterone and NAF protein YKL40 and bFGF, we treated breast cancer cell lines MDA-MB-453 (AR+) and MDA-MB-231 (AR−) using testosterone at 1 nM (0.29 ng/ml) or 10 nM (2.9 ng/ml) for 24 or 48 h. As shown in Fig. 1, in MDA-MB-453 cells, YKL40 was up-regulated by testosterone at 1 and 10 nM after 24-h treatment (P < 0.05). bFGF was down-regulated by testosterone at 1 and 10 nM after 48-h treatment (P < 0.05 for 1 nM, P < 0.01 for 10 nM). In the AR-negative cell line MDA-MB-231, YKL40 and bFGF were not regulated by testosterone treatment, suggesting that the effects of testosterone were mediated through androgen receptor. Testosterone up-regulating YKL40 and down-regulating bFGF confirmed the positive correlation between YKL40 and NAF testosterone level and negative correlation between bFGF and NAF testosterone level in vivo.

Fig. 1
figure 1

Effects of testosterone on protein expression YKL40 and bFGF in breast cancer cell lines MDA-MB-453 and MDA-MB-231 by Western blot (a) and average quantification (b) of three experiments. The error bars indicate standard error. *P < 0.05, ** P < 0.01 vs untreated cells


The first studies of NAF constituents as risk biomarkers for breast cancer risk focused on the epithelial cell component, with encouraging results [40]. Since then, the hormonal and protein components of NAF have been the subject of numerous studies. However, the interrelationships of these constituents are also of interest since they may paint a cohesive picture of the intra-mammary environment, and lead to new insights regarding breast cancer etiology and therefore, to avenues of prevention. We have previously demonstrated correlations between NAF estrogens and cathepsin D as well as EGF [34]. We have also shown that NAF hormones do not relate to serum hormones, with the exception progesterone [38]. In the present study, we measured five proteins that are candidate breast cancer risk biomarkers, based on positive results in previous studies by other investigators [2532]. We did not observe significant differences between the high-risk unaffected contralateral breast of breast cancer subjects (at high risk for future cancer) and standard risk healthy controls, but we did find NAF protein markers showed higher correlation with NAF hormone levels than with serum hormone levels, suggesting that NAF protein composition may be determined by local hormone levels in the breast, rather than systemic hormone levels in the blood. Different protein markers were specifically correlated with different steroids, primarily androgen, suggesting that these steroids make distinct contributions to the breast environment. The stronger correlation between NAF protein with NAF hormones (rather than serum hormones) may represent a direct or indirect effect of these hormones on protein synthesis by ductal cells; the proximity of the steroid concentrations to the parenchymal cells that produce the protein factors strengthens the evident direct involvement of the steroids with the appearance of the proteins in NAF. We acknowledge that the measurements of protein in NAF only reflect the secreted protein from tissues. Further studies to examine the abundance of target proteins in breast tissues using immunohistochemistry will be necessary to evaluate the overall expression of the target proteins and their correlation with NAF hormone.

We chose five protein markers in NAF based on their function and potential candidacy as breast cancer risk markers, including angiogenesis-related growth factor (bFGF), cell invasion and migration-associated protease (CatD), inflammation protein markers (YKL40 and CRP), and an antioxidant enzyme (SOD1). Previous studies from different groups found that bFGF [26], CatD [28] and CRP [31] were significantly increased, and SOD1 was lower [34] in NAF from tumor-bearing breasts than that from normal breasts, suggesting their roles in tumor development. Moreover, further studies also found bFGF [26, 27], CatD [28], and YKL40 [32] were elevated in NAF from high-risk breasts, including the contralateral unaffected breast, breasts with high-risk epithelial lesions (ADH or LCIS), as well as high-risk individuals due to family history. CRP in NAF was found to be significantly related to multiple breast cancer risk factors such as age at first pregnancy and recent lactation [29], and also positively correlated to breast cancer risk as predicted by the Gail model [30]. We compared these five protein markers in NAF samples from 60 prospectively recruited breast cancer cases, using their contralateral unaffected breasts as a high-risk model, and 54 frequency-matched screening mammography controls with clinically normal breasts. We did not find significant differences between the two groups in bFGF, YKL40, CRP, and SOD1 levels. Only CatD showed a non-significantly higher level in cases than controls (3.3 vs 2.3 ng/μl, P = 0.13), consistent with prior studies [28]. Our results do not necessarily negate the findings of previous investigators regarding these biomarkers since our sample size is small; however, several other studies have been of similar size. Therefore, these results do point out that future studies should be prospective, well-controlled, and large.

Besides the five protein markers in this study, several other proteins in NAF have been reported to be related to breast cancer. Prostate-specific antigen (PSA) in NAF appears to be inversely associated with breast cancer stage, with levels decreasing as the stage of breast cancer increases [41, 42]. A study of testosterone- and prostate-specific antigen (PSA) levels in women with and without breast cancer suggested that low NAF PSA and high serum testosterone were significantly associated with the presence of breast cancer [43]. Another study found that uPA and PAI-1 in NAF were increased in women with atypia and cancer compared to women with benign disease [44], suggesting that these markers may be useful in early breast cancer detection.

In our previous study on catD in NAF samples from 47 premenopausal healthy women [34], cathepsin D showed significant correlations with NAF DHEA (R = 0.26, P < 0.0001) and estradiol (R = 0.15, P = 0.018). In that study, each individual contributed three or four NAF samples at mid-luteal phases during four menstrual cycles; the correlation represented a combination of between-person variation and within-person variation. In current study, we confirmed the moderate correlation between cathepsin D and NAF DHEA (R = 0.273, adjusted P = 0.13), but did not find a correlation with NAF estradiol. In the current study, each woman contributed only a single NAF sample, not timed to the menstrual phase, possibly accounting for the lack of correlation with estradiol because of menstrual phase variation.

We observed significant protein-hormone correlations between bFGF and testosterone (negative in both serum and NAF), and a positive correlation of YKL40 with NAF testosterone. We confirmed the up-regulation of YKL40 and down-regulation of bFGF by testosterone in AR+ breast cancer cell lines (MDA-MB-453), but not in AR-negative cell line MDA-MB-231. MDA-MB-453 cells were known to show high expression of AR and stimulated to proliferate by testosterone [45, 46]. An early study found that both bFGF and testosterone stimulated the growth of mouse mammary carcinoma cells (SC-3) through up-regulating FGF receptor [47], suggesting bFGF and testosterone share a similar signaling pathway to promote tumor cell proliferation. The negative correlation between bFGF and testosterone in our study may indicate a competitive inhibition or negative feedback loop in normal breast tissue, a possible mechanism which needs further investigation. YKL40 (also known as human cartilage glycoprotein-39) is a secretory molecule related to the chitinase protein family, which plays roles in post-lactational involution of the breast [48]. YKL40 normally acts as a protective signaling factor that determines which cells survive during breast involution, but cancer cells may use this protein to extend cell survival, since higher serum levels of YKL40 are related to poor survival in breast cancer patients [49, 50]. The possible regulation of YKL40 by testosterone has not been previously reported, and further mechanistic investigation of this is warranted.

In conclusion, NAF proteins that have been implicated as breast cancer risk markers by others showed stronger correlations with NAF hormone levels, rather than with serum hormone levels. Our results indicate that local breast steroid hormone levels may determine the level of expression of these proteins in the breast, and that these hormone-responsive proteins are responsive to the abundant androgenic steroids that are found in NAF. Although we did not see case-control differences in these NAF proteins, our results provide insights into the regulation of these proteins by androgens signaling pathway that may affect breast cancer risk.