A Liquid Chromatography-Mass Spectrometry Method for the Simultaneous Measurement of 15 Urinary Estrogens and Estrogen Metabolites: Assay Reproducibility and Interindividual Variability

The link between elevated endogenous estrogen levels and postmenopausal breast cancer is well established, with substantial evidence accruing from prospective studies of both circulating and urinary estrogens (1-4). However, it has not been possible to determine which of the estrogens commonly measured is most predictive of risk. Indeed, the precise mechanism of estrogen-related breast carcinogenesis, including the contribution of individual patterns of estrogen metabolism, has yet to be fully elucidated. The prevailing hypotheses suggest that specific estrogens not only can stimulate mitogenic activity, inhibit apoptosis, and promote tissue growth but also be metabolized into genotoxic and mutagenic forms that may directly damage DNA (5).

Estrogen metabolism occurs primarily in the liver along an oxidative pathway, although enzymes involved in hormone metabolism are also expressed and are functional in estrogen target tissues, such as the breast (5). Hydroxylation of estrone (E₁) and estradiol (E₂) is catalyzed on either the A-ring or the D-ring by various cytochrome P450 enzyme isoforms and results in the formation of several hydroxy and keto metabolites. Hydroxylation on the A-ring is predominately at the C2 position, which produces 2-hydroxyestrone (2-OHE₁) and 2-hydroxyestradiol (2-OHE₂), and, to a lesser extent, at the C4 position, which produces 4-hydroxyestrone (4-OHE₁) and 4-hydroxyestradiol (6). These four estrogen metabolites (EM) are catechol estrogens, with adjacent hydroxyl groups on the aromatic A-ring. The 2- and 4-hydroxy derivatives are further converted by catechol-O-methyltransferase to 2-, 3-, and 4-methoxy estrogens, including 2-methoxyestrone (2-MeOE₁), 2-methoxyestradiol (2-MeOE₂), 2-hydroxyestrone-3-methyl ether (3-MeOE₁), 4-methoxyestrone (4-MeOE₁), and 4-methoxyestradiol (4-MeOE₂; ref. 7). Hydroxylation at the 16α position of the D-ring produces 16α-hydroxyestrone (16α-OHE₁), which can be further metabolized to estriol (E₃), 17-epiestriol (17-epiE₃), 16-ketoestradiol (16-ketoE₂), and 16-epiestriol (16-epiE₃; ref. 8). Following conjugation with sulfate, glucuronic acid, or glutathione, EM are eliminated by excretion in the urine and/or feces. Evidence is accumulating that substantial differences exist in the genotoxic, mutagenic, and proliferative activities of the individual estrogens and EM and that these differences may play a role in breast cancer etiology (5, 8-13). Thus, the measurement of a wide spectrum of EM is needed to further our understanding of breast carcinogenesis.

Most studies that have assessed the role of urinary estrogens in breast cancer etiology in human populations have used direct or indirect RIA or enzyme immunoassays to measure a limited number of EM (10). Direct assays can be conducted with relative ease and at low cost; however, because each analyte must be assayed separately, at least moderate quantities of biological specimen are required. Moreover, concerns about the specificity, accuracy, reproducibility, and sensitivity of the commercial kits persist (14). Indirect assays, which incorporate extraction and chromatography before RIA, are more specific and accurate but also more costly, time-consuming, and require more biological sample (14). Unlike these assay methods, mass spectrometry (MS) methods have the potential to measure a spectrum of EM at once using a single sample (15). The stable isotope dilution gas chromatography/MS method that has been considered the gold standard for steroid hormone measurement is reasonably accurate, specific, and sensitive; however, sample preparation is complicated and laborious (16).

To facilitate the study of the role of estrogen metabolism in large epidemiologic studies of hormonal carcinogenesis, accurate, precise, sensitive, robust, and relatively simple laboratory methods that require low volumes of biological samples are needed. We have developed a stable isotope dilution high-performance liquid chromatography-tandem MS (LC-MS/MS) method that requires only 0.5 mL urine and a simple sample preparation, consisting of hydrolysis, extraction, and derivatization, yet is capable of measuring simultaneously at least 15 EM (17, 18). In this report, we evaluated laboratory variability for our assay based on a formal reproducibility study in which blinded, randomized urine samples from premenopausal women in the follicular and luteal phase, postmenopausal women, and men were assayed over several weeks. We also assessed the interindividual variability of EM levels for each menstrual/sex group.

A total of 45 subjects provided overnight urines for this study, including 10 men ages 21 to 65 years, 20 premenopausal women ages 21 to 45 years, and 15 naturally postmenopausal women ages 50 to 65 years. Because the purpose of this study was to assess the performance of the assay method, we did not select participants based on characteristics suspected of modifying EM profiles, such as age, race, medication use, or anthropometry. However, subjects reporting current use of hormone replacement therapy, oral contraceptives, or other hormone supplements were ineligible. Subjects were instructed to collect all urine from the overnight period, including the first morning void. For the 20 premenopausal women, urines were obtained as follows: 10 women collected urine during the midluteal phase, between days 19 and 23 of the current menstrual cycle, and another 10 women collected urine during the follicular phase, between days 1 and 10 of the menstrual cycle. Urine was collected in a half-gallon container with a teaspoon of l-ascorbic acid as a preservative and kept in ice or refrigerated until the following day when urines were decanted, aliquoted, and stored at −70°C.

To establish the range of values and optimize the assay technique, the laboratory initially measured EM in urines from 25 of the subjects, including 10 premenopausal (5 luteal and 5 follicular) women, 10 postmenopausal women, and 5 men. Samples were measured in four batches, with each batch containing duplicate urine aliquots from each of the 25 subjects. For this exploratory phase, the laboratory was aware of the sex and menstrual status of the participant that provided the sample. Following this, the reproducibility of the method was evaluated using urines from the remaining 20 subjects (5 follicular, 5 luteal, and 5 postmenopausal women and 5 men). For this component of the study, the laboratory received four batches of urines at one time, each containing two aliquots per subject, and was instructed to assay one batch at the start of each of 4 consecutive weeks. The 40 samples in each batch were placed randomly, and the laboratory was unaware of whether a sample came from a premenopausal or postmenopausal woman or a man. Samples were stored at −80°C until assayed.

A total of eight measurements were obtained for each subject in both the initial optimization effort and the reproducibility study. All steps of the assay procedure, including hydrolysis, extraction, derivatization, and LC-MS/MS, were done separately on each aliquot.

Details of our analytical method for urinary EM, which includes hydrolysis, extraction, derivatization, and LC-MS/MS with stable isotope-labeled internal standards, have been published elsewhere (17, 18). A summary of the laboratory method is presented here. Fifteen estrogens and EM, including E₁, E₂, E₃, 2-OHE₁, 2-MeOE₁, 2-OHE₂, 2-MeOE₂, 3-MeOE₁, 4-OHE₁, 4-MeOE₁, 4-MeOE₂, 16α-OHE₁, 17-epiE₃, 16-ketoE₂, and 16-epiE₃, were obtained from Steraloids. Four deuterium-labeled estrogens and EM (d-EM), including E₂-2,4,16,16-d₄ (d₄-E₂), E₃-2,4,17-d₃ (d₃-E₃), 2-OHE₂-1,4,16,16,17-d₅ (d₅-2-OHE₂), and 2-MeOE₂-1,4,16,16,17-d₅ (d₅-2-MeOE₂), were purchased from C/D/N Isotopes. A fifth d-EM, 16-epiE₃-2,4,16-d₃ (d₃-16-epiE₃), was obtained from Medical Isotopes. Charcoal-stripped human urine (Golden West Biologicals) that contained 0.1% (w/v) l-ascorbic acid and had no detectable level of any EM was used for preparation of calibration standards and quality-control samples. For all the deuterated internal standards, the purity was >98%.

Because endogenous estrogens and their metabolites are mostly present in urine as glucuronide and sulfate conjugates, an initial hydrolysis step that has been optimized for this purpose was included (16, 19). To a 0.5 mL aliquot of urine, 20 μL of the d-EM internal standard solution (1.6 ng of each d-EM) was added followed by 0.5 mL of freshly prepared enzymatic hydrolysis buffer containing 2.5 mg l-ascorbic acid and 5 μL β-glucuronidase/sulfatase from Helix pomatia (type HP-2, ≥500 Sigma units β-glucuronidase and ≤37.5 units sulfatase activity; Sigma) in 0.5 mL of 0.15 mol/L sodium acetate (pH 4.6). The sample was incubated for 20 h at 37°C and then extracted with 8 mL dichloromethane. After extraction, the aqueous layer was discarded and the organic solvent portion was transferred to a clean glass tube and evaporated to dryness at 60°C under a stream of nitrogen gas. Ascorbic acid had been added to the samples to prevent oxidation of individual EM. Levels of the calibration and quality-control standards were relatively stable, which suggested that degradation was minimal during sample preparation.

All EM and d-EM were quantitatively dansylated to improve their ionization efficiency and thus the sensitivity of the assay. The dried sample residue was redissolved in 100 μL of 0.1 mol/L sodium bicarbonate buffer (pH 9.0) and 100 μL dansyl chloride solution (1 mg/mL in acetone) and incubated at 60°C for 5 min. A 20 μL sample (1:10 of final volume) was injected for LC-MS/MS analysis.

The LC-MS/MS analysis was done using a ThermoFinnigan TSQ Quantum-AM triple quadrupole MS equipped with an electrospray ionization source and coupled directly to a Surveyor HPLC System (ThermoFinnigan). Both the chromatography system and MS were controlled using Xcalibur software (ThermoFinnigan). Reverse-phase LC was carried out on a 150 mm long × 2.0 mm i.d. C₁₈ column packed with 4 μm Synergi Hydro-RP particles (Phenomenex) and maintained at 40°C. The mobile phase, operating at a flow rate of 200 μL/min, consisted of methanol as solvent A and 0.1% (v/v) formic acid in water as solvent B. A linear gradient changing solvent A from 72% to 85% in 75 min was employed. After washing with 100% solvent A for 13 min, the column was reequilibrated with a mobile phase of 72% solvent A for 12 min before the next injection. The general MS conditions were as follows: source: electrospray ionization, ion polarity: positive, spray voltage: 4,600 V, sheath and auxiliary gas: nitrogen, sheath gas pressure: 49 arbitrary units, auxiliary gas pressure: 23 arbitrary units, ion transfer capillary temperature: 350°C, scan type: selected reaction monitoring, collision gas: argon, an collision gas pressure: 1.5 mTorr. The following MS variables were used for all experiments: scan width: 0.7 units, scan time: 0.50 s, Q1 peak width: 0.70 units full width at half-maximum, and Q3 peak width: 0.70 units full width at half-maximum.

Quantitation of each EM in urine was carried out using the Xcalibur Quan Browser (ThermoFinnigan). Calibration curves for the 15 EM were constructed by plotting EM-dansyl/d-EM-dansyl peak area ratios obtained from calibration standards versus amounts of EM and fitting these data using linear regression with 1/X weighting. The amount of each EM in a urine sample was then interpolated using this linear function. Quality-control samples at three concentrations (0.12, 0.96, and 6.4 ng of each EM/mL) were included in each batch of assays. The limit of quantitation is 2 pg on the column for each EM. The reported EM concentrations, in pg EM/mg creatinine, are for unconjugated forms of the EM. No corrections were made for molecular weight differences between the conjugated and the unconjugated forms.

Because deuteriums at the α-position to the carbonyl group of ketolic estrogens are especially susceptible to exchange loss, only deuterium-labeled standards that exhibited no significant exchange loss under the assay conditions were employed. Based on structural similarity and retention times, d₄-E₂ was used as the internal standard for E₂ and E₁; d₃-E₃ for E₃, 16-ketoE₂, and 16α-OHE₁; d₃-16-epiE₃ for 16-epiE₃ and 17-epiE₃; d₅-2-MeOE₂ for 2-MeOE₂, 4-MeOE₂, 2-MeOE₁, 4-MeOE₁, and 3-MeOE₁; and d₅-2-OHE₂ for 2-OHE₂, 2-OHE₁, and 4-OHE₁.

Urinary creatinine was measured with the Beckman LX20 analyzer using a technique based on the Jaffe reaction (20). Laboratory coefficients of variation (CV) were 2% for creatinine concentrations of 70 and 155 mg/dL.

Urinary EM levels from all 45 participants were used to examine interindividual differences in metabolite concentrations. In the initial phase of this study, the laboratory was provided samples from the first 25 subjects to optimize the method and was not blinded as to the participant's sex and menstrual group status. Urines from these 25 subjects were assayed eight times, with duplicate aliquots measured in each of four batches. These results were then combined with those from the 20 subjects evaluated in the reproducibility study described below, where blinded urine samples were also measured eight times, twice in each of four batches. Data were analyzed on the logarithmic scale. For each EM, the geometric mean (expressed as pg EM/mg creatinine) and the arithmetic mean for percent of total EM were calculated for each menstrual/sex group. Rank-order statistics and box plots were used to compare the distributions of EM concentrations among the menstrual/sex groups.

The reproducibility study included only the 20 subjects whose samples were randomized and identifiers were removed before shipment to the laboratory. A nested, within-person ANOVA model was used to assess assay reproducibility over the 4-week period (21). Data were analyzed on the logarithmic scale. Estimates of the variability among subjects (σ²_a), assay variability among batches for a given subject (σ²_b), and assay variability associated with different aliquots in the same batch for the same subject (σ²) were obtained from the SAS procedure Proc VARCOMP (22). With y_ijk denoting the mean of the natural logarithm of assay measurements of duplicate aliquots for subject i = 1, 2, 3 … 20 at week j(i) = 1, 2, 3, 4 on aliquot k(ij) = 1, 2, the model is y_ijk = μ + a_i + b_j(i) + ε_k(ij), where a_i, b_j(i), and ε_k(ij) are independent variables, each with a mean of 0 and respective variances σ²_a, σ²_b, and σ². From the variance estimates, we computed estimates of the total, intrabatch, and interbatch CVs of the assay. We also computed the intraclass correlation coefficients [ICC = (σ²_a / σ²_a + σ²_b + σ²)] for each menstrual/sex group.

For each urinary EM, geometric mean concentrations (pg EM/mg creatinine), mean percent of total urinary EM, and rank order by concentration are presented in Table 1 for follicular phase women, luteal phase women, postmenopausal women, and men. Results from all 45 subjects are included in this table. In all groups, the ranking of EM concentrations was similar. Levels of E₃, 2-OHE₁, E₁, E₂, and 16-ketoE₂ were the highest, accounting for 60% to 75% of the total urinary EM. E₃ was the predominant urinary EM in all groups, except follicular phase women where 2-OHE₁ was highest. The catechol estrogens comprised ∼20% of the total, with concentrations of 4-OHE₁ and 2-OHE₂ being much lower than that of 2-OHE₁. With the exception of 2-MeOE₁, concentrations of the methoxy estrogens tended to be very low for all groups, comprising <10% of total urinary EM. Levels of 4-MeOE₂ were below the detection limit for 3 males in the study and measured in some, but not all, batches for 1 postmenopausal woman. The total concentration of urinary EM was similar for follicular and luteal phase women and, of note, comparable for men and postmenopausal women.

The ranges in absolute urinary EM concentrations (pg EM/mg creatinine), according to menstrual/sex group, are shown by box plots in Fig. 1, with E₁ and E₂ in Fig. 1A, catechol estrogens in Fig. 1B, estrogens in the 16α pathway in Fig. 1C, and methoxy estrogens in Fig. 1D. As these plots show, interindividual differences in each menstrual/sex group were very large, with the highest concentration for each EM at least five times the lowest. For nearly all the EM, 10- to 100-fold differences between individuals were observed. The exceptions were 4-MeOE₁ in follicular women; 16-ketoE₂ and 4-MeOE₁ in luteal women; 2-OHE₁, 2-OHE₂, and 2-MeOE₁ in postmenopausal women; and E₁, E₂, 2-OHE₂, 4-OHE₁, 16-epiE₃, and 2-MeOE₁ in men, for which the ranges were <10-fold, and 16α-OHE₁ in follicular women, 4-OHE₁ in luteal women, 16α-OHE₁ and 2-MeOE₂ in postmenopausal women, and 17-epiE₃ in men, where the differences between participants were >100-fold.

In contrast, interindividual differences in relative urinary EM concentrations, expressed as percent of total urinary EM, were not as large. For the majority of the EM, the difference between the highest and the lowest values in each menstrual/sex group was <10-fold; for E₁ and E₂, in particular, the range in percent EM was <6-fold (data not shown).

The design of the reproducibility component of this study is presented in Fig. 2. The duplicate readings for 16α-OHE₁ in each of four batches are shown for 5 follicular phase women in (a), 5 luteal phase women in (b), 5 postmenopausal women in (c), and 5 males in (d). Data are plotted on a logarithmic scale. Each distinct symbol in each plot corresponds to a different subject in that menstrual/sex group. Where only one symbol per subject is shown, the values for the duplicate readings in that batch overlapped. These plots depict several sources of variability in urinary hormones assessed in this study, including differences among subjects, measurement variability over batches for a given subject, and variability among duplicate aliquots in a given batch. Note that this does not measure the variability over time for a given subject; for this, we would need different urine samples collected at different times from each subject. For this EM and all the EM studied (see additional plots in Supplementary Figs. S1-S14), there was little difficulty in distinguishing one subject from another. In men, for example, where levels of 16α-OHE₁ were particularly low, the assay successfully separated the readings for each subject. Furthermore, even at these low concentrations, measurements varied little from one batch to the next or between duplicate aliquots in the same batch.

As these plots suggest, laboratory reproducibility for each EM across the menstrual/sex groups was very similar; thus, we present reproducibility results for the entire study group of 20 subjects combined (Table 2). The overall laboratory CVs were ≤10% for all urinary EM and ≤5% for 11 of the 15. The within- and between-batch CVs were consistently <3% and <10%, respectively.

ICCs are also presented in Table 2. In light of the large differences in EM concentrations among the menstrual/sex groups, the ICCs were calculated separately for each of the four groups. With the exception of 17-epiE₃ in postmenopausal women, the ICCs were uniformly very high (>95%).

To study the role of endogenous estrogens in carcinogenesis, including specific EM and individual patterns of estrogen metabolism, evaluating EM levels in urine may be advantageous because of the ease of sample collection, potentially higher participation rates, and the integration of exposure over time for hormones with pulsatile, circadian, or menstrual cycle variability. In our study, laboratory results were highly reproducible, with overall CVs for each of the 15 EM ≤10%, and the majority being <5%. These findings compare favorably with the gas chromatography/MS method for measuring urinary EM and phytoestrogens developed by Taylor et al. (19), where intrabatch and interbatch CVs for two known standards were 2% to 13% and 4% to 16%, respectively. Further, ICCs for all the 15 EM were very high, generally ≥98% in each menstrual/sex group. Although it may be argued that results from the small number of participants in this study may not adequately represent the underlying population values and provide imprecise estimates of the variance components used to calculate the relevant statistics, the consistency of the reproducibility and ICC findings across the menstrual/sex groups is remarkable.

Concerns have been raised regarding the high-throughput capability of this method due to difficulties with sample preparation and the interpretation of MS peaks. Sample preparation for this method, which entails hydrolysis of the conjugated EM with glucuronidase and sulfatase enzymes followed by extraction and then dansylation of the unconjugated EM with established technology, is relatively easy and could be automated. LC-MS/MS is accomplished with commercially available LC and MS systems, and peaks corresponding to the EM of interest are interpreted with automated software. Although the results are reviewed to ensure accuracy, most steps in the LC-MS/MS process are unattended. A total of 40 to 45 study samples, along with the necessary quality-control and calibration curve standards, can be assayed per week. Although relatively time-consuming in comparison with measurement methods for single EM, our method does provide urinary hormone measurements for 15 analytes in a single run.

Several caveats must be borne in mind when evaluating interindividual differences in this study. First, of the 45 subjects studied, samples from the initial 25 were not blinded for the laboratory. Although this has the potential of introducing bias, this is not likely in light of the very strong laboratory performance observed in the formal blinded reproducibility study. Second, the relatively small number of participants in each menstrual/sex group limits our ability to explore hormonal profiles with regards to demographic, clinical, and anthropometric characteristics and awaits hormone measurements in large population-based studies. Further, we did not attempt to evaluate the consistency of hormone measurements within subjects over time, which may be a source of considerable variability, particularly among premenopausal women. Thus, although not conclusive, we did find interindividual differences in absolute urinary levels were quite high in all four menstrual/sex groups for each of the 15 EM, with values differing from 10- to 100-fold within each group. In both women and men, E₃, 2-OHE₁ and E₁ were the most abundant of the urinary estrogens followed by E₂ and 16-ketoE₃. Except for 2-MeOE₁, levels of the methoxy estrogens were relatively low. The finding of relatively high levels of 2-MeOE₁ in urine is notable in light of the suggestion that 2-MeOE₁ and 2-MeOE₂ interconvert under physiologic conditions (23, 24). 2-MeOE₂ shows promise as a potential anticancer agent, with in vitro and in vivo studies suggesting that it may inhibit cell proliferation, angiogenesis, and tumor growth (23, 25, 26).

The correlation of EM levels in various tissues is not known. Urinary EM levels reflect extensive hydroxylation, methylation, ketone formation, and conjugation in the liver and kidneys and may not be predictive of circulating or breast tissue levels. In addition, several estrogen metabolism enzymes are expressed and functional in estrogen target tissues (5). Considerable evidence shows large interindividual variation in the hepatic and extrahepatic expression of some cytochrome P450 enzymes in the estrogen metabolic pathway, which is largely due to genetic and/or environmental factors. Gender differences for some liver cytochrome P450 isoforms involved in estrogen metabolism have been documented in animal models (5), but studies of human liver microsomes have not found gender differences with regard to either the enzymatic activity of cytochrome P450 isoforms with high estrogen-metabolizing activity (27) or the EM formed (28, 29). Although not conclusive, our observation of similar patterns of EM for men and women supports these in vitro findings.

Both normal and neoplastic breast tissues produce the enzymes required for localized estrogen synthesis, oxidation, conjugation, and deconjugation (12, 30, 31), but how estrogen metabolism proceeds in normal and neoplastic tissue is unresolved. More detailed and accurate knowledge about EM patterns in breast tissue may suggest new approaches to the early detection and treatment of breast tumors. Further, although many accepted hormone-related breast cancer risk factors, such as ages at menarche, first birth, and menopause, are not easily modifiable, estrogen metabolism can be altered by diet, smoking, exercise, coffee consumption, alcohol intake, and chemopreventive and other drugs (32-35). Clarifying the role of individual compounds as well as patterns of estrogen metabolism in breast carcinogenesis may, therefore, be relevant to prevention strategies. To these ends, efforts to optimize MS methods such as ours for the measurement of individual EM in blood and tumor tissue are needed.

X. Xu, T. Veenstra, L. Keefer, R. Ziegler; coinventors on relevant government patent.

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.

We thank Glen L. Hortin, M.D., Ph.D. (Clinical Chemistry Branch, NIH) for measuring urinary creatinines.