Sex-specific Estrogen Levels and Reference Intervals from Infancy to Late Adulthood Determined by LC-MS/MS

Abstract Context The lack of sensitive and robust analytical methods has hindered the reliable quantification of estrogen metabolites in subjects with low concentrations. Objective To establish sex-specific reference ranges for estrone (E1) and estradiol (E2) throughout life and to evaluate sex-differences using the state-of-the-art liquid chromatography tandem mass spectrometry (LC-MS/MS) method for quantification of E1, E2, and estriol (E3). Design LC-MS/MS method development and construction of estrogen reference ranges. Settings Population-based cross-sectional cohorts from the greater Copenhagen and Aarhus areas. Participants Healthy participants aged 3 months to 61 years (n = 1838). Results An isotope diluted LC-MS/MS method was developed and validated for measurements of serum E1, E2, and E3. Limits of detections (LODs) were 3 pmol/L (E1), 4 pmol/L (E2), and 12 pmol/L (E3), respectively. This sensitive method made it possible to differentiate between male and female concentration levels of E1 and E2 in children. In girls, E2 levels ranged from <LOD to 100 pmol/L during mini-puberty, whereas it was ≤20 pmol/L during childhood. E1 and E2 increased with age and pubertal breast stage and varied during the menstrual cycle; E1 was lower than E2 in girls and premenopausal women, and higher than E2 in postmenopausal women. In boys, E1 and E2 increased with age and pubertal stage, whereas little changes with age were observed in men. High E3 concentrations were confirmed in pregnant women. Conclusion Reference ranges of simultaneous quantification of E1 and E2 by this novel specific and highly sensitive LC-MS/MS method provide an invaluable tool in clinical practice and in future research studies.


Preparation of calibration, quality control and validation materials
All native standards from suppliers were sonicated for 15 min. and diluted in 100% MeOH followed by sonication for 10 min. to native stock solutions (E1-stock, E2-stock and E3-stock) containing 10 g/mL of each of the estrogen standards. From these stock solutions, a native standard mixture (E-mix-1) diluted in 40% MeOH containing E1 (0.2 g/mL), E2 (0.2 g/mL), and E3 (0.6 g/mL) was prepared and further diluted to an E-mix-2 containing 2 ng/mL E1, 2 ng/mL E2, and 6 ng/mL E3 and an E-mix-3 containing 0.2 ng/mL E1, 0.2 ng/mL E2, and 0.6 ng/mL E3. For calibration curves, eight solutions of the native E-mix-2 and E-mix-3 diluted in 0.6 mM NH4F/40% MeOH were prepared in the concentration ranges: E1, 7.4-2589 pmol/L; E2, 7.3-2569 pmol/L; and E3, 21-7282 pmol/L. Furthermore, for determination of limits of detection (LOD), limits of quantifications (LOQ), and intra-day variation, the E-mix-2 and E-mix-3 were diluted to nine solutions (concentration ranges: E1, 3.7-355 pmol/L; E2, 3.7-352 pmol/L; E3, 34-3328 pmol/L) by spiking a randomly collected serum pool from prepubertal children (beforehand tested for low or no content of estrogens). Intra-day variation and inter-day variation were determined based on measurements of serum pool without spiking as well as serum pool spiked in three concentrations (Q low, Q middle, and Q high). All calibration and quality control materials were stored at -20C until use. Solvent blank samples (Milli-Q water), control materials and standard solutions for calibration curves were further treated as the real samples.
E3-is (100 g) was dissolved in 1.0 mL MeOH. Internal standards from supplier (E1-is and E2-is) and dissolved E3-is (100 g/mL) were sonicated for 15 min. and diluted with 100% MeOH for preparation of internal stock solutions (E1-is-stock, E2-is-stock and E3-is-stock) containing 10 g/mL of each the isotope labelled estrogen standards, after which the stocks were sonicated for 10 min. An internal standard mixture (E-is-mix-1) diluted in 40% MeOH containing E1-is (0.1 g/mL), E2-is (0.1 g/mL) and E3-is (0.2 g/mL) were prepared from the stock solutions and diluted further to the final internal standard mixture, E-is-mix-2 containing 1 ng/mL E1-is, 1 ng/mL E2-is, and 2 ng/mL E3-is.

Sample preparation
After thawing, estrogens from all serum samples, including calibration and control materials, were purified by liquid-liquid extraction. Each sample was mixed by brief vortexing and left for 10 min. at room temperature. Two hundred µL of each sample was aliquoted to 1.5 mL Eppendorf tubes and added 40 µL of internal standard solution (E-is-mix-2) and 160 µL 0.6 mM NH4F/40% MeOH. The samples were vortexed on a shaking table (IKA®VIBRAX VXR basic) briefly at 2500 rpm and at 1500-2000 rpm for 10 min. at room temperature. Next, 400 µL 50% heptane/ethyl acetate were added to all samples, and the samples were vortexed on the shaking table for 30 min. at room temperature following the same procedure as described above. Samples were then centrifuged (Eppendorf centrifuge 5430k) for 10 min. at 25200 rcf at 4C and thereafter placed in an ethanol bath that was kept below 0C using dry-ice pills. The organic and aqueous phases were separated by pouring the fluid organic phase (upper layer) into glass tubes. The organic extracts were evaporated to dryness (at 45C and 5-15 psi N2), and the residues containing the estrogens were re-suspended in 120 µL 0.6 mM NH4F/40% MeOH (freshly prepared). The solution was briefly mixed and finally transferred to HPLC vials.

Analytical method
The content of E1, E2 and E3 in serum samples, controls and calibration materials were measured using a newly developed method for simultaneous quantitative determination of estrogens in human serum by isotope dilution TurboFlow-LC-MS/MS. Analyses were performed on a Dionex UltiMate 3000 UHPLC system (Thermo Scientific, San Jose, CA, USA) with the integrated Transcend TLX TurboFlow sample preparation system (Thermo Scientific, San Jose, CA, USA) coupled with a triple quadrupole mass spectrometer (TSQ Quantiva, Thermo Scientific, San Jose, CA, USA) controlled by Aria MX 2.2 and Xcalibur 4.0 software (ThermoFinnigan, Bellefonte, PA, USA). Samples were introduced with a HTS PAL autosampler (CTC Analytics AG, Zwingen, Switzerland) and kept at 10°C. For sample extraction and chromatographic separation of the estrogens the TurboFlow-LC system was equipped with a loading Cyclone-P TurboFlow column (0.5 x 50mm) (Thermo Scientific, Franklin, MA, USA) followed by an analytical Kinetex® Phenyl-Hexyl column (100Å, 2.1 x 50 mm, 2.6 µm particle size) equipped with a Drop-in SecurityGuard Ultra cartridge (UHPCL Phenyl, 2.1 mm) in front of the analytical column. Both pre-and analytical column were purchased from Phenomenex. The loading column was operated at 22°C (controlled room temperature) and the analytical column was kept at 30°C with a Multisleeve column heater from Thermo Scientific. The MS/MS-system was equipped with a heated electrospray ionization source (HESI) running in negative mode with short shift to positive mode at the end of the time period duration. Solvents used on the TurboFlow (loading) system were: A, 5 mM NH4AC in 2% MeOH; B, MeOH; and C, acetone/isopropanol/acetonitril (10:45:45 v/v). Solvents used on the eluting system were: A, 0.4 mM NH4F in H2O adjusted to pH 7.5 by adding 200 µl 0.25% NH3/L solvent (freshly prepared), and B, MeOH. A bypass valve was used for directing the flow to waste for the first two minutes and the last minute of the analysis time, respectively. The total duration time was 5.50 min. The injection volume was 80 µL with a flow rate and solvent programming as shown in Supplementary Table 2

Operation procedure and method validation
For calibration curves, the ratio between the area of native standard and internal standard was plotted as a function of concentrations of native standards. Through linear regression based on area ratios (sample area/internal standard area), the concentration of unknown samples and the control material were determined.
For method validation and all other analyses, two calibration curves in Milli-Q water were included at the beginning and the end of all sample batches.
Matrix effect and ion suppression were investigated in duplicate calibration curves in Milli-Q water and serum pool at 9 different concentration levels as described above for each compound. The responses from standards prepared in the serum pool were plotted as functions of the responses from standards prepared in Milli-Q water. Subsequently, 95% confidence intervals (CI) were calculated for slopes and intercepts of the linear regression (Supplementary Table 5) using the regression function in Analysis Toolpak for Microsoft Excel 2007. If the 95% CI included 1 for the slopes and 0 for the intercepts, no matrix effect was present. In cases in which the slope-and/or intercept constants in the equation for estrogen/serum calibration curves differed from the equation for estrogen/water calibration curves, matrix effects occured. Sample results might then be corrected for this matrix effect by dividing with the slope coefficient or by deducting the intercept value estimated for the estrogen/serum calibration curve equation from all sample results. However, in this case in which the 95% CI´s of the slope and intercept of the calibration curve in serum was almost 1 and 0, further adjustment was not performed.
The intra-day variability was estimated based on five repeated calibration curves made in the serum pool: Accuracy (% recovery) and precision (relative standard deviation (RSD)) were calculated for a low, mean, and high concentration levels (Q low, Q middle and Q high) from the five repeated serum calibration curves. These five repeated estrogen/serum calibration curves were further used for determination of linearity and limit of detection (LOD) and quantification (Supplementary Table 6 and Supplementary Figure  2A-C). For this, the approach described by the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) (1) on validation of analytical procedures was used: 3.3 times the standard error of the intercept of the calibration curve with the y-axis divided by the slope of the calibration curve using the five lowest calibration levels for each analyte (1,2). The standard error of the intercept and the slope of the calibration curve were calculated using the regression function in Analysis Toolpak for Microsoft Excel 2007 (Microsoft Corp., Redmond, WA). The inter-day variation (precision) was estimated from analysis of control material (Q low, Q middle, and Q high) examined in duplicates in 18 batches over a period of 2 months (Supplementary Table 7 .0 a Inter-day based on serum pool spiked in low, mean and high level of standards, n=108 analyzed in 18 batches over an period of 2 month. RSD: relative standard deviation Supplementary Table 8. Median concentration and selected percentiles estrogens and androgens in 188 women from 24.7 to 43.9 years of age separated in early follicular phase from day 1-7, late follicular phase from day 8-14, the lutheral phase post ovulation from day 15 (15+ (all)), and from day 15, where ovulation was confirmed by a progesterone level > 10 nmol/L (15+ (ovul)) Menstrual cycle day Supplementary Figure 2. Estrogen standards spiked in serum a pool from boys and in milli-Q water.
Supplementary Figure 3. Logarithmic plots of female serum concentrations of free estradiol (E2) and free estrone (E1) as a function of age, from 0.25-61 years, n=1055 (a,b) and calculated free E2 and free E1 from 0.25-61 years, n=985 (c,d). Solid lines represent medians and dashed lines represent ± 1 and ± 2 standard deviations, respectively.