Estrogen and Progesterone Levels in Nipple Aspirate Fluid of Healthy Premenopausal Women: Relationship to Steroid Precursors and Response Proteins

Epidemiologic studies have consistently shown a significant positive association between serum concentrations of estrogen and breast cancer risk in postmenopausal women (1, 2). A relationship between risk and serum progesterone has not been shown, although a considerable amount of indirect evidence suggests that endogenous progesterone does contribute to breast cancer development (3), and the presence of the progesterone receptor is an indicator of hormone-responsive cancer (4). In premenopausal breast cancer, the link between sex steroid levels and risk has been difficult to evaluate due to menstrual fluctuation of these compounds in serum, particularly for estradiol (5). Although some studies have failed to show a relationship (6), the European Prospective Investigation into Cancer and Nutrition (by far largest prospective cohort study among premenopausal women conducted to date) recently reported significant risk elevations for women with higher serum levels of testosterone and adrenal androgens and no association for estrogens despite considerable efforts to control for temporal fluctuation (7). Hormone concentrations in breast tissue itself rather than serum could be far more closely related to risk and more temporally stable as well. However, measuring tissue concentrations in population studies is problematic due to the heterogeneity of breast tissue extracts and the need for relatively invasive sampling techniques. Nipple aspirate fluid (NAF), which is continuously secreted into breast ducts by apocrine and merocrine processes, can be obtained noninvasively and could provide a better reflection of breast hormone concentrations than serum (8).

We have developed assays that measure estradiol, progesterone, and related analytes in NAF and have previously reported that these assays, which are capable of measuring multiple analytes in a single small-volume sample, are sensitive and reproducible (9). More recently, we reported on a comparison of NAF hormone levels in groups of women who varied by menopausal status and exogenous hormone use (10). In the studies presented here, our goals were to assess the relationship of estrogen and progesterone levels in NAF to the corresponding levels in serum and saliva, to evaluate the association between estradiol and its potential estrogenic and androgenic precursors in NAF, and to determine whether concentrations of estradiol and related hormones in NAF are associated with NAF concentrations of the estrogen response proteins epidermal growth factor (EGF) and cathepsin D.

Apart from addressing the intuitive or content validity of hormone assays in NAF, the present results provide insight into the capability of the breast for synthesizing its own growth-promoting steroids. Breast tissue is known to contain aromatase (CYP19) and other enzymes involved in steroid biosynthesis and metabolism (11). The recent success of aromatase inhibitor drugs for treatment and prevention of breast cancer emphasize the need for better understanding of the role of estrogen synthesis in the breast and other nonovarian tissues (12). This could eventually lead to a more precise explanation of the relationship between ovarian steroid levels and breast cancer risk, or to the use of NAF measurements as intermediate markers in breast cancer prevention trials.

Women were recruited by mass mailing and advertisements within the Chicago area to participate in a dietary intervention trial involving a low-fat diet and soy supplements; this trial has been described in detail elsewhere (13). Eligible women were between the ages of 20 and 40 years, were having regular menstrual cycles of 25 to 35 days in length, were not taking any medications that interfere with ovarian function, and were at least 6 months post-lactation. We obtained written consent for participation following procedures approved by the Institutional Review Board at the Northwestern University. Participants provided biological samples, including serum, saliva, and NAF, during four menstrual cycles: at baseline, then during the 4th, 12th, and 15th cycles after enrollment. In these months, the subjects collected urine specimens on days 10 to 15 for self-assessment of ovulation using a commercial test kit for luteinizing hormone. They collected daily saliva samples, which were stored in their home freezers and came to the clinic to have blood drawn and to have samples of NAF collected 5 to 8 days (mean = 7.2 days) after the urinary luteinizing hormone peak, was detected. For NAF collection, the breasts were warmed and massaged by the subjects themselves. Then a small volume of fluid was aspirated from each breast with a syringe connected to a bell-shaped applicator. When more than one duct in a breast provided fluid, the fluid from those ducts was combined. Overall, NAF was obtained from 48% of the 196 women enrolled in the diet trial. Generally, if NAF was obtained from a subject at the first visit, it was obtained on all subsequent visits. For this study, we included a total of 47 women who had an adequate amount of NAF obtained from both breasts on at least three visits. The fluid was collected in calibrated capillary tubes, the volume was recorded, and the ends were sealed with clay. The median volume of NAF collected was 11 μL (range = 0-151 μL). The samples were stored at −20°C until they were thawed for assay.

The procedures for sample preparation and analysis have been described in detail in previous publications (9, 14). The volume of NAF was diluted 1:9 with PBS (pH 7.4). Cellular components were not separated from the fluid. From the diluted sample, 50 μL were generally taken for extraction, although as little as 30 μL were used in some cases. This volume was further diluted to 700 μL with PBS. The samples were extracted twice with 1.0 mL of ethyl acetate-hexane (3:2). The remaining aqueous fraction was kept for assay of water-soluble analytes, and the organic solvent was evaporated and redissolved in 1.0 mL isooctane. Estradiol was extracted from the isooctane solution with 1.0 ± 0.5 mL of 0.4 N aqueous NaOH, the alkaline solution was neutralized with HCl, and the estradiol was reextracted into ethyl acetate. The ethyl acetate was washed with 0.5 mL of water and evaporated. The residue was dissolved in 1.0 mL PBS (pH 7), containing 0.1% gelatin for assay. Standards of estradiol for the assay were prepared in the same buffer. The isooctane solution containing the neutral steroids was diluted with 0.1 mL of ethyl acetate, washed with water, and evaporated. It was redissolved in 800 μL PBS (pH 7.4) containing 0.1% bovine serum albumin. Standards for dehydroepiandrosterone (DHEA), androstenedione, and progesterone assays were prepared in the same buffer. Water used in the procedures was deionized (Milli-Q water purification system, Millipore Corp., Bedford, MA) and then redistilled in an all-glass still. The sodium phosphate salts used to prepare the buffers were recrystallized from aqueous methanol. Ethyl acetate, hexane, and isooctane were redistilled within 2 weeks of use. Pipette tips were washed in an ultrasonic cleaner with Liquinox (Alconox Corp., White Plains, NY) detergent before use and rinsed with deionized water. The concentration of DHEA sulfate (DHEAS) that was observed in direct assays was extremely high, suggesting that something in the samples was interfering with binding of the steroid to the antiserum. Therefore, the solvolysis technique (15) was used to hydrolyze the sulfate from DHEAS, and the resulting DHEA was assayed in an aliquot of the free steroid.

Assays for estrogens and progesterone in serum and saliva were done as described previously (16). All NAF samples from each subject were assayed in duplicate within one assay run. Each run included two or more quality controls that were prepared from a pool of NAF that had been aliquoted in small volumes and stored at −20°C before use. Buffer blanks were carried through the assay, and the values obtained were subtracted from the final results. When commercial kits were used, they were modified such that the standards for the assays were prepared in the same buffer that was used to redissolve the samples. Estradiol and progesterone were measured as described previously for salivary and serum assays (17). DHEA and androstenedione were assayed using kits supplied by Diagnostic Systems Laboratories (Webster, TX). The intra-assay coefficients of variation for estradiol, progesterone, DHEA, and androstenedione were 13.7, 16.0, 11.0, and 11.1, respectively. Interassay coefficients of variation were obtained from a pool of excess NAF from a previous study. Interassay coefficients of variation for estradiol, DHEA, and androstenedione were 16.8, 43.5, and 20.3, respectively. Values for progesterone in the pool were below the lowest standard and could not be used for this purpose. Using the sample preparation procedure described above, the buffer blanks for estradiol, progesterone, DHEA, and androstenedione were 1.37 ± 0.95 pmol/L, 0.11 ± 0.16, 0.0 ± 0.009, and 0.07 ± 0.18 nmol/L, respectively. The efficiency of the extraction and purification procedure for estradiol was evaluated to determine procedural losses. The recovery of 22.9 and 45.9 pmol estradiol/L in buffer, after fractionation and measurement in the RIA, was 77 ± 4% and 83 ± 3%, respectively.

In the aqueous fraction, DHEAS was assayed directly by an ELISA from Diagnostic Systems Laboratories and, after solvolysis with 0.1% HCl in acetone (15), by an ELISA for DHEA (Diagnostic Systems Laboratories). Estrone sulfate was assayed directly by a RIA from Diagnostic Systems Laboratories. EGF and cathepsin D were assayed by ELISA kits from R&D Corp. (Minneapolis, MN) and Calbiochem (San Diego, CA), respectively. The intra-assay coefficients of variation for DHEAS (after hydrolysis), estrone sulfate, cathepsin D, and EGF were 15.0, 7.8, 17.6, and 7.1, respectively. Inter-assay coefficients of variation were 53.7, 20.0, 23.2, and 12.7, respectively. The variability must be improved for between-subject comparisons, especially for DHEAS for which the hydrolysis procedure varied in efficiency between batches. The buffer blanks for DHEAS (after hydrolysis), estrone sulfate, cathepsin D, and EGF were 5.9 ± 11.8 μmol/L, 0.09 ± 0.13 nmol/L, 0.92 ± 1.0 ng/mL, and −2.0 ± 3.0 ng/mL, respectively. A pool of NAF was initially tested by comparison of direct assay with assay after the extraction procedure. After extraction, the average measured values for cathepsin D and EGF were 96.8% and 94.4%, respectively, of that in unextracted samples. The values presented have not been corrected for procedural losses. The sensitivity of the individual assays was sufficient for measurement of the mean value of all analytes in NAF when the sample size was 5.0 μL.

Univariate distributions for each analyte were reviewed and we determined that log transformation gave approximately normal distributions for parametric analyses. For salivary estradiol, average daily levels were computed for the 7-day segment from luteal days 2 to 8 and for the 15-day segment spanning day 6 before ovulation to day 8 after. Salivary progesterone levels were computed as the average of the 5-day segment centered on the midluteal day, which was the day NAF was collected. These segments of consecutive day saliva samples were selected to provide optimal reproducibility (16). To examine relationships between hormone levels, data were fit to repeated measures analysis of covariance models, in SAS, with the following general form:

where Y_ijkl is the log NAF estradiol (or other dependent variable) concentration for the ith woman in the jth treatment group (j = 1, 2, 3, 4) at the kth visit (k = 1, 2, 3, 4) corresponding to the lth breast (l = 1, 2); μ is the mean intercept variable, γ_j is the jth group effect; ω_i(j) is a random effect due to the ith woman within the jth group (overall woman-to-woman variation); ν_k is the kth visit effect; (γν)_jk is a group by visit interaction effect; δ_i(jk) is a random effect due to the ith woman within group j and visit k (i.e., breast-to-breast variation within women); β_l is the effect of lth breast (left versus right); (γβ)_jl, (νβ)_kl, and (γνβ)_jkl are corresponding interaction terms with breast; and ε_i(jkl) is within subject variation (i.e., within woman variation). Inclusion of terms for treatment group means that the associations between NAF constituents (the focus of this analysis) are independent of any dietary effects on NAF hormones in the trial. The basic model was extended by adding covariates for other hormone concentrations. Overall model fit and R² were obtained using a SAS macro that provided both average and conditional R² and concordance correlations. All Ps evaluated were two sided.

Table 1 summarizes important characteristics of the study participants and the samples used in this analysis. As previously reported, salivary hormone concentrations were considerably lower than corresponding concentrations in serum, because saliva excludes protein-bound fractions. Concentrations of estrogen in NAF were higher, on average, than those in serum.

Table 2 shows the associations between estradiol and progesterone levels in NAF and contemporaneous levels of estrogens and progesterone in serum and saliva. Both regression model coefficients and correlation coefficients indicated very small and nonsignificant associations between NAF estradiol and various estrogens and progesterone in serum and saliva. On the other hand, NAF progesterone was significantly associated with progesterone levels in both serum and saliva. Models used for computing these associations accounted for sampling of one or both breasts on repeated visits.

Results concerning the relationship of NAF estradiol levels to possible precursor steroids and progesterone in NAF are shown in Table 3. Estrone sulfate, DHEA, and progesterone were highly associated with NAF estradiol concentrations (each P < 0.02), whereas DHEAS and androstenedione were positively but nonsignificantly associated. In a multihormone model that included all of the NAF hormones in Table 3, none of the individual hormones was a dominant predictor of NAF estradiol. However, although the correlation between single hormones and NAF estradiol were not very large, the full model explained most of the variance in NAF estradiol. The average model R² was 0.83, and the concordance correlation (between observed and predicted NAF estradiol) was 0.85.

Regression model results with the estrogen response protein cathepsin D in NAF as the dependent variable are shown in Table 4. Three estradiol precursors in NAF (DHEA, DHEAS, and estrone sulfate) were significantly associated with cathepsin D in NAF (all P < 0.001). Estradiol itself was positively associated with cathepsin D, but there was little association with androstenedione and essentially none with progesterone. None of the serum or salivary hormone measurements were notably predictive of NAF cathepsin D. Because saliva samples from up to 15 consecutive days (6 days before ovulation to 8 days after) were assayed, and because NAF samples were obtained on a midluteal day, results were obtained both for the entire segment of salivary estradiol and for the shorter luteal segment from days 2 to day 8. Salivary progesterone samples from up to 5 days bracketing the midluteal point were analyzed.

NAF concentrations of EGF were significantly associated with estradiol, estrogen precursors, and progesterone in NAF (Table 5). The association between cathepsin D and EGF in NAF was very weak. As was true for cathepsin D, serum or salivary hormone concentrations in NAF did not predict EGF levels in NAF. None of the individual hormones were strongly correlated with either NAF cathepsin D or EGF (highest r = 0.32). However, in a multihormone model, the concordance correlation between observed and predicted cathepsin D was 0.70. Combined NAF hormone levels were even stronger predictors of EGF (concordance r = 0.93).

In this study, we observed that levels of estradiol in NAF of premenopausal women were significantly associated with estrogen precursor steroids within the same breast. However, the associations between NAF estradiol and estradiol precursors in NAF or estradiol itself in serum (measured on the same midluteal day) were negligible. Estradiol in NAF was also not measurably associated with salivary estradiol, using either a broad 15 consecutive day segment of daily saliva samples or a 7-day segment focused on the luteal phase. Progesterone levels in NAF, by contrast, were fairly strongly correlated with progesterone levels in both serum and saliva. In further analyses, we observed that NAF concentrations of cathepsin D and EGF (both peptides that are up-regulated in response to estrogen signaling) were positively associated with estradiol and its precursors in NAF but not with steroid levels in serum or saliva.

These results lead to at least two important inferences. First, by showing consistent relationships between estradiol and both its precursor and response molecules, these data support the conclusion that measured levels of estrogen in NAF are biologically valid and meaningful. Measurement of steroid concentrations in breast tissue itself is problematic due to the need for relatively invasive sampling and the heterogeneity of tissue extracts (18, 19). Second, the results indicate that serum and salivary levels of estrogen (but not progesterone) are relatively poor predictors of NAF levels. A similar lack of correlation was observed for serum estrogens and estrogens extracted from breast tissue (18). Consequently, we can expect that epidemiologic estimates of breast cancer risk in relation to serum hormone levels will be different, and perhaps severely attenuated, compared with the risk associations that would be seen with NAF hormone levels. It has been observed that the rate of increase in breast cancer incidence with aging is lower after menopause than before (20). However, the reduction in the incidence slope is far less than proportional to the decrease in serum estradiol levels that occur during menopause. Synthesis and conversion of steroids within the postmenopausal breast could explain why relatively high incidence rates can be maintained (21).

In previous research, we showed that the current NAF assays are sensitive and reproducible, with strong correlations between paired breasts and within a single breast over time. More recently, we reported that NAF levels of estradiol, EGF, and cathepsin D were not significantly reduced in postmenopausal relative to premenopausal women and that potential estradiol precursors remain available in the breast after menopause (10). We also reported that drastic changes in ovarian hormone profile, due to oral contraceptive use or hormone replacement therapy (HRT) indeed seem to cause largely predictable changes in NAF estrogen and progesterone. For example, women using hormone replacement therapy had a >2-fold higher mean NAF estradiol than postmenopausal women not using hormone replacement therapy.

Additional studies on hormones in NAF or within breast tissue itself are still limited in number. In 1987, Ernster et al. reported the surprising observation that premenopausal and postmenopausal women had approximately similar levels of estrogen in NAF, and that serum and NAF estrogen levels were poorly correlated (22). These investigators also reported that pregnancy and lactation were associated with a persistent decline in NAF estrogen levels, which would be consistent with their protective effect on breast cancer risk. Subsequent research has shown that breast tissue contains enzymes capable of synthesizing estradiol from its precursors, which are abundant in the breast (23-25). Many of these studies were conducted on tumor tissue. Our group recently reported evidence that estrone sulfate is converted to estradiol in homogenates of normal breast tissue (14). In this respect, the breast may resemble the prostate, where intracrine processes are responsible for regulating concentrations of potent androgens from large pools of precursor substances (21).

The finding that a high percentage of the variation in estradiol concentration in NAF can be predicted from the variation in estradiol precursors in NAF is fortuitous. Thus, the pool size of these precursors, including estrone sulfate, androstenedione, DHEA, DHEAS, and progesterone, are positively and independently related to the accumulation of estradiol in NAF. This may have great use in future studies when effects of environmental or pharmacologic treatments are evaluated for effects on individual steps in the biosynthetic pathways.

The human cathepsin D gene, which encodes a lysosomal aspartyl protease, contains multiple estrogen-responsive elements in its 5′ promoter region (26), and estradiol strongly induces cathepsin D expression in cultured breast cancer cells. Harding et al., who measured estrogen-stimulated proteins in NAF following hormonal treatment, reported decreases in cathepsin D and pS2 following treatment with luteinizing hormone–releasing hormone agonist or tamoxifen and increases following estrogen replacement therapy (27). A 2-week trial of dietary soy supplement increased pS2 and decreased apoliproprotein D (an estrogen-inhibited protein) in NAF of premenopausal women (28). Our results indicate that cathepsin D expression in the normal breast may be regulated by endogenous estrogen as well. Estradiol up-regulates EGF receptor at the mRNA and protein levels (29), and EGF has been shown to be elevated in breast cancer (30). We showed in previous work that EGF levels in NAF were correlated with serum estradiol within the same woman (31). Our present data show consistently strong positive relationships, in NAF, between EGF and estrogens, estrogen precursors, and progesterone. The mechanism responsible for this association, or even its causal direction, is not yet clear. When information on several NAF hormones was combined in a multivariate model, we were able to explain a large amount of the total variance in both cathepsin D and EGF. Surprisingly, we found only a weak association between cathepsin D and EGF themselves, despite studies showing that EGF induces cathepsin D expression in breast cell lines (32).

Although the full picture is complex and incompletely resolved, several lines of evidence indicate that progesterone and its metabolites within the breast can contribute to early breast carcinogenesis (33). Progesterone increases the proliferative activity of breast epithelium during the luteal phase of the menstrual cycle, and both observational studies and clinical trials suggest a greater breast cancer risk among women using hormone replacement therapy with combined estrogen plus progestin compared with estrogen alone (34, 35). Progesterone levels in NAF are relatively stable over time within a breast and vary considerably between women. Our finding that progesterone is positively associated in NAF (P < 0.0001) with EGF, a breast mitogen, raises concern that local concentrations of progesterone could be a marker of elevated cancer risk. Serum and saliva levels of progesterone do seem to reflect concentrations in NAF, unlike estradiol; however, the correlations between either serum or saliva and NAF were relatively modest. A recent analysis from the Nurse's Health Study found no association between serum progesterone and postmenopausal breast cancer nor any interaction with estradiol levels despite evidence that adding progestins to hormonal replacement therapy increases breast cancer risk compared with estrogen alone (1). The European Prospective Investigation into Cancer and Nutrition study results suggested that breast cancer risk might be higher among premenopausal women with lower serum progesterone concentrations as well as elevated androgens (7). Therefore, a protective role for progesterone during the premenopausal years, when breast epithelium is exposed to high levels of estrogen, is not implausible (36). A study of NAF progesterone levels might provide an alternative approach to explaining the apparently complex relation of this hormone to breast cancer risk.

The present study is the first to comprehensively examine the inter-relationships between hormones and response proteins in NAF. Some of its strengths include a relatively large number of samples with repeated measures of NAF, serum, and saliva and the synchronization of sampling during the midluteal period. On the other hand, the cross-sectional nature of this study precludes any firm conclusion about the temporal relationships among NAF constituents. The application of NAF biomarkers is also limited by the difficulty in obtaining adequate amounts of NAF from all volunteers. In our experience, adequate samples can be obtained from about 40% to 60% of premenopausal women and a lower percentage from postmenopausal women. Other investigators have reported adequate sampling from a higher proportion of subjects (37). Success on initial sampling is an excellent predictor of subsequent success in obtaining fluid from the same individual. Of course, the relationships reported here are not likely to be explained by nonrandom sampling of the entire eligible population. We are not able to comment on whether the unconjugated steroids measured in this study are in the bound or unbound state; this could affect biological activity in tissues. We did not include testosterone in our panel of NAF analytes; however, we do not anticipate any technical barriers to doing so in the future.

In conclusion, we have determined that estradiol in fluid aspirated from the breasts of healthy premenopausal women is far more strongly associated with the concentration of estrogen precursors and response proteins in the same fluid, compared with the same compounds in either serum or saliva. Progesterone levels in this fluid, which are associated with corresponding levels in serum and saliva, predict levels of EGF, a prominent breast cell mitogen. These findings warrant further investigation of NAF hormone levels as biomarkers of breast cancer risk and perhaps eventual use as intermediate end-point biomarkers in early phase prevention trials. Much more work remains to be done to explain the genetic and environmental basis for interwoman variation in these hormone levels (38), in parallel with work that will lead to insights regarding control of local hormone synthesis and metabolism within the normal breast.

We thank Sue Giovanazzi, Linda Van Horn, Erin Anderson, Ryan Deaton, Yu-Cai Lu, and the participants in the Diet and Hormone Study for their valuable contributions to this project.