Study population

We collected urine from 75 women at 3 different sites: (i) at the Center for Mammographic Screening at the University of Naples, Italy (42 women), (ii) at the Breast Diagnostic Clinic and Oncology Breast Clinic of the Mayo Clinic, Rochester, MN (18 women) and (iii) at the Olson Center for Women’s Health, University of Nebraska Medical Center (UNMC), Omaha, NE (15 women). Women were recruited between March 2005 and September 2006 and their ages ranged between 34 and 73 years—healthy women: range, 34–67; mean, 50 ± 8; high-risk women: range, 44–64; mean, 52± 6; women with breast cancer: range, 34–73; mean, 54 ± 10.

All women recruited at the University of Naples were healthy (they did not receive a diagnosis of breast cancer at the time of their mammographic test). Among the women recruited at the Mayo Clinic, 12 were classified as high-risk women (Gail Model score = 1.67%–11.7%) and 6 were breast cancer cases. At UNMC, 4 women were healthy, that is, had no known cancer, and 11 were diagnosed with breast cancer. None of the subjects received estrogen-containing treatment for at least 3 months prior to providing a urine sample. The 3 groups were frequency matched on age, race and menopausal status.

All procedures were approved by the University of Naples, Mayo Clinic and UNMC Institutional Review Boards. Signed consents included authorization to collect and bank urine samples and collect demographic and clinical information.

Sample collection

A standardized method was followed to collect all of the urine samples. A spot urine sample of about 50 ml was collected from each participant and 1 mg/ml ascorbic acid was added to prevent oxidation of the catechol moieties in the various estrogen compounds. The urine samples were aliquoted, frozen and four 10-ml aliquots were transferred to the Eppley Institute, UNMC, on dry ice and were stored at −80°C until analysis. Thus, each analytical sample was thawed only once prior to analysis.

Solid-phase extraction of urine

Two milliliter aliquots of urine were partially purified by SPE. The SPE was performed using a 20-port SPE vacuum manifold with phenyl cartridges (Fig. 2). Urine samples were adjusted to pH 7 with 1 M NaOH or 1 M HCl. For method development and validation, 2-ml aliquots of charcoal-treated human urine samples were spiked with a total of 250, 500 or 1,000 pg of the 40 estrogen-related compounds (final concentration 0.125, 0.25 and 0.50 pg/μl) and loaded onto the phenyl 100-mg cartridges preconditioned with CH₃OH and the loading buffer, 10 mM ammonium formate, pH 7. The cartridges were washed with the loading buffer, and then the compounds of interest were eluted from the cartridge by using an elution buffer, methanol/10 mM ammonium formate, pH 7 (90:10) with 1% acetic acid. This procedure led to enrichment of the 40 estrogen-related compounds after elution. Charcoal-treated urine (2 ml) was used in controls, and the eluates from the SPE cartridges were spiked with 250, 500 or 1,000 pg of the 40 estrogen-related compounds. The eluates from both the experimental and control samples were concentrated using a Speed-Vac and lypholizer, and subjected to ultraperformance liquid chromatography/tandem mass spectrometry (UPLC/MS-MS) analysis. To determine the recovery of the standards by the SPE method, comparison was made between the corresponding concentrations of experimental and control samples (Fig. 3). Study samples were cleaned in duplicate by using the above optimized SPE conditions and analyzed by UPLC/MS-MS.

UPLC/MS-MS analysis of urine samples

The 40 analytes (Table I) included the androgens androstenedione and testosterone; the estrogens E₁ sulfate, E₁ and E₂; the catechol estrogens 2-OHE₁(E₂) and 4-OHE₁(E₂); the 16α-OHE₁(E₂); the methylated 2- and 4-catechol estrogens; the 2- and 4-catechol estrogens conjugated with GSH, cysteine (Cys) or N-acetylcys-teine (NAcCys); and the depurinating DNA adducts of 4-OHE₁(E₂) and 2-OHE₁(E₂). All of the estrogen compounds were analyzed as both E₁ and E₂ derivatives because the interconversion of these 2 estrogens is carried out continuously by 17β-estradiol dehydrogenase.

All experiments were performed on a Waters (Milford, MA) Quattro Micro triple quadrupole mass spectrometer by using electrospray ionization (ESI) in positive ion (PI) and negative ion (NI) mode, with an ESI-MS capillary voltage of 3.0 kV, an extractor cone voltage of 2 V, and a detector voltage of 650 V. Desolvation gas flow was maintained at 600 l/h. Cone gas flow was set at 60 l/h. Desolvation temperature and source temperature were set to 200 and 100°C, respectively. For all the studies, a methanol:-water (1:1) mixture with 0.1% formic acid was used as the carrier solution. ESI interface tuning and mass calibration were accomplished in the PI mode by using a standard sodium iodide-rubidium iodide solution. The test sample (compounds 1 through 40) was introduced to the source at a flow rate of 10 μl/min by using an inbuilt pump. PI or NI detection was used in cases where the sample was readily ionized to cation or anion, respectively. The masses of parent ion and daughter ions were obtained in the MS and MS-MS operations. The parent and daughter ion data obtained for each compound were used to generate the multiple reaction monitoring (MRM) method for UPLC/MS-MS operation (Table I).

Measurements of estrogen-related compounds in urine extracts were conducted by using UPLC/MS-MS. UPLC/MS-MS analyses were carried out with a Waters Acquity UPLC system connected with the high-performance Quattro Micro triple quadrupole mass spectrometer. Analytical separations on the UPLC system were conducted using an Acquity UPLC BEH C18 1.7 μm column (1 × 100 mm) at a flow rate of 0.15 ml/min. The gradient started with 80% A (0.1% formic acid in H₂O) and 20% B (0.1% formic acid in CH₃CN), changed to 79% A over 4 min, followed by a 6-min linear gradient to 45% A, resulting in a total separation time of 10 min. The elutions from the UPLC column were introduced to the Quattro Micro mass spectrometer.

The ionization method used for MS analysis was ESI in both the PI and NI mode. MS-MS was performed in the MRM mode (see above), and resulting data were processed by using QuanLynx software (Waters) to quantify the estrogen metabolites. To calculate limits of detection, various concentrations, 0.1, 0.25, 0.5, 1.0, 2.5, 5.0, 10, 25, 50 and 100 pg/μl, of the analyte were injected to UPLC/MS-MS. The injected amount that resulted in a peak with a height at least 2 or 3 times as high as the baseline noise level was used as the limit of detection (Table I). Pure standards were used to optimize the UPLC/MS conditions prior to analysis. After UPLC analysis, the mean value was calculated for all the compounds obtained from each sample.

Median values of the urinary estrogen-related compounds in the 3 groups of women

Using the newly developed SPE/UPLC/MS-MS methodology, we have analyzed urine samples of various women’s groups for estrogen-related compounds. The data obtained were used to calculate median values for each of the 40 compounds (Table III).

The median andostenedione, testosterone, E₂/E₁, 16α-OHE₂/ 16α-OHE₁, 4-OCH₃E₂/4-OCH₃E₁, 2-OHE₁(E₂) GSH conjugate and derivative values were higher for controls compared to high risk participants, and the median 4-OHE₂-1-N3Ade/4-OHE₁-1-N3Ade values were lower for controls compared to high risk participants. Compared to breast cancer participants, the median 2-OHE₂/2-OHE₁, 4-OHE₂/4-OHE₁, 16α-OHE₂/16α-OHE₁, 4-OCH₃E₂/4-OCH₃E₁, 2-OHE₁(E₂) GSH conjugate and derivative values were higher for controls, while the median 4-OHE₂-1-N7Gua/4-OHE₁-1-N7Gua values were lower for controls. Of particular interest are the significantly lower levels of the methoxycatechol estrogens in the women with breast cancer or at high risk compared to the control women, because this represents a major protective pathway in estrogen metabolism. In addition, the levels of the 4-OHE₁(E₂)-1-N3Ade and 4-OHE₁(E₂)-1-N7Gua adducts are higher in the women with breast cancer or at high risk than in the control women, although only 2 of the differences are statistically significant.

Depurinating estrogen-DNA adducts in the 3 groups of women

In the second analysis, the ratios of depurinating N3Ade and N7Gua adducts to the sum of estrogen metabolites and conjugates in urine samples from healthy control women are generally low (Fig. 4). In contrast, high ratios of these adducts to estrogen metabolites and conjugates were observed in urine from high-risk women (Gail Model score >1.66%) and women with breast carcinoma. In general, the value obtained from the high-risk women and women with breast carcinoma derives from the ratio between a high level of adducts and low levels of metabolites and conjugates. In some women, however, the level of adducts was not particularly high, but the levels of metabolites and conjugates were very low, suggesting that a substantial proportion of the metabolites was converted to adducts.

In the sum of the ratios of depurinating adducts to estrogen metabolites and conjugates, the preponderant role is played by the N3Ade and N7Gua adducts of 4-OHE₁(E₂), whereas the adducts of 2-OHE₁(E₂) play a very minor role. For example, for the high-risk subject presented in Table II, the overall adduct ratio is 936, but the contribution of 2-OHE₁(E₂)-6-N3Ade is 1, whereas the contribution of 4-OHE₁(E₂)-1-N3Ade plus 4-OHE₁(E₂)-1-N7Gua is 935. In general, the average contribution of the 2-OHE₁(E₂)-6-N3Ade adducts is ~2.5% of the total, whereas the predominant contribution of ~97.5% derives from the 4-OHE₁(E₂)-1-N3Ade and 4-OHE₁(E₂)-1-N7Gua adducts. The observation of high levels of depurinating estrogen-DNA adducts in urine from high-risk women, as well as subjects with breast carcinoma (Fig. 4), is consistent with the hypothesis that these adducts are a causative factor in the etiology of breast cancer.

Analysis by subject characteristics

We first analyzed the data using the ratio of depurinating N3Ade and N7Gua adducts to the sum of their respective estrogen metabolites and conjugates in urine samples as a continuous variable. Analysis using one-way ANOVA revealed a significant difference among the groups (p < 0.001). Additional post hoc analysis using a Bonferroni correction for multiple comparisons revealed significantly higher means for high risk subjects [mean 336.45, standard deviation (SD) 331.92] compared to controls (mean 20.51, SD 37.01, p < 0.001) and for breast cancer patients (mean 176.28, SD 205.68, p < 0.001). The mean for patients known to be at high risk was not significantly different from that of the breast cancer group (p = 0.62).

A limitation of the study is that most of the group of healthy women (42 of 46) were Italian, whereas the remaining healthy women, high-risk women and women with breast cancer were American. All of the subjects in our study, however, were Caucasian. The 3 groups (healthy, high-risk and breast cancer) had similar mean age at recruitment, mean age at menarche and menopausal status (Table IV). These similarities in subject characteristics support the validity of comparing the ratios of adducts to their respective metabolites and conjugates in these 3 groups of women.

Subject characteristics of age at recruitment, age at menarche, menopausal status, and parity were available for 56 of the 75 subjects (Table IV). The mean age of our entirely Caucasian sample was 50 years (SD 8.5). The average age at menarche was 12.0 years (SD 1.4). Only 11% of the women were nulliparous and 43% had at least 2 children. Twenty-six (46%) women were pre-menopausal at recruitment, 30 (54%) were postmenopausal (they did not have menstrual cycles in the last 12 months before recruitment). Analysis using one way ANOVA revealed that health status, that is breast cancer cases versus high risk and healthy individuals, was significantly associated with age at recruitment (p = 0.048). Specifically, the mean age (years) at recruitment for healthy women was 49 (SD 7.8), 52 (SD 6.1) for women at high risk and 57 (SD 12.2) for breast cancer cases. Age at menarche was not statistically different across the disease status groups (p = 0.534). Analysis using a χ² test did not reveal an association between health status and menopausal status (p = 0.95) or parity (parous vs. nulliparous) (p = 0.15).

The correlation coefficient was used to examine the association between the ratio and subject characteristics. We observed evidence of significant correlation between parity and ratio (r = 0.36, p = 0.007) and marginally significant correlation between the ratio and menopausal status (r = 0.26, p = 0.06). Age at recruitment and age at menarche were not significantly associated with the ratio.

Linear regression was used to assess the association between disease status and ratio adjusted for age at recruitment, age at menarche, menopausal status (categorical) and parity for the 56 subjects with patient characteristics available (Table V). After accounting for these characteristics, the ratio was significantly associated with health status. Specifically, the multivariate coefficient for disease status (108.6) was statistically significant (p = 0.007) in a model that explained 10% (p = 0.040) of variance in the ratio after accounting for covariates. All other covariates did not reach the usual level of significance of 0.05 (Table V).

Grant sponsor: National Cancer Institute (U.S. Public Health Service); Grant numbers: P01 CA49210, P30 CA36727, P30 CA15083; Grant sponsor: U.S. Army (Breast Cancer Research Program); Grant number: DAMD 17-03-1-0229; Grant sponsor: National Institutes of Health; Grant number: P30 CA036727-22; Grant sponsor: AIRC (Associazione Italiana Ricerca sul Cancro), Milan, Italy.

The authors thank Dr. M. Saeed and Dr. M. Zahid for the synthesis of estrogen metabolites, conjugates and DNA adducts.