Simultaneous determination of a suite of endogenous steroids by LC-APPI-MS: Application to the identification of endocrine disruptors in aquatic toxicology

Introduction

Decades since the US Environmental Protection Agency (USEPA) first established the Endocrine Disruptor Screening Program (EDSP) [1], there remains substantial interest in the potential effects of endocrine-disrupting compounds (EDCs) in both regulatory and research settings. Small fish are among the most commonly used nonmammalian species for testing of EDCs [2, 3], with model laboratory species including fathead minnow (Pimephales promelas), Japanese medaka (Oryzias latipes), and zebrafish (Danio rerio) [4]. Small fish are critical for testing of potential ecological impacts of EDCs [4] and, due to the highly conserved nature of the hypothalamic-pituitary gonadal (HPG) axis, data from tests with fish also can be extrapolated to other vertebrate species, including humans [5]. Endocrine-disrupting compounds can perturb the HPG axis in vertebrates by interacting with various molecular targets, in many cases leading to changes in circulating steroid hormones and potential impacts on development and reproduction [6–8]. Steroid hormones are a critical element of normal HPG axis function, and a variety of chemicals have been shown to cause adverse effects through altering circulating steroid concentrations in small fish species [9–12]. As such, plasma steroid concentrations are frequently used as a biomarker of effect in identifying EDCs [7].

The small size and short life cycle of model small fish species are highly advantageous for toxicity testing; however, the same attributes pose a challenge when measuring steroid hormones, as sample volumes can be quite limited. For example, obtainable plasma volume in the fathead minnow is frequently <10 μL per individual leading to the need for highly sensitive methods for quantification of steroids [13]. Radioimmunoassays (RIAs) or enzyme-linked immunosorbent assays (ELISAs) are commonly used to quantify steroids of interest due to their high level of sensitivity and relative ease of access [14–17]; however, these assays have demonstrated cross-reactivity with structurally similar compounds, which can lead to discrepancies in assay results [18, 19]. Additionally, only one analyte can be quantified in a single RIA, which becomes limiting in relation to the small sample volumes and sample sizes typically employed in small fish testing. For example, it is recommended that 17β-estradiol, testosterone, and 11-ketotestosterone be measured in short-term reproduction assays using fathead minnow, but plasma volumes from individual fathead minnows are commonly insufficient to perform three independent RIAs [14]. In cases where there is inadequate plasma for steroid analyses, exposure water itself may be used as a surrogate for steroid plasma concentrations (see Scott and Ellis for review [20]). Steroids can partition directly from plasma to the surrounding water by passive diffusion through the gills [21], and steroids also can be actively excreted as glucuronide and sulfate conjugates in urine or feces [31]. Some of these conjugated steroids have direct pheromonal activity and play a critical role in reproductive signaling [22, 23]. However, RIAs pose similar limitations in analysis of water where only one analyte can be measured per extracted water sample.

To overcome these limitations, liquid chromatography-mass spectrometry (LC-MS) based methods have become increasingly popular for measuring steroids. LC-MS methods demonstrate increased specificity and can allow for simultaneous quantification of multiple steroids, reducing the need to split samples for multiple RIAs, for instance. However, challenges remain for quantifying multiple steroid classes in a single sample. Previously developed methods have largely focused on quantification of just a few key steroids, namely 17β-estradiol, testosterone, and 11-ketotestosterone. However, teleost fish are somewhat unique among vertebrates in that they utilize several “non-classical” steroids, including 11-deoxycortisol, 17,20β-dihydroxyprogesterone, and 17,20β,21-trihydroxyprogesterone, during final oocyte maturation [15, 24] (see Fig. 1). Further, insights can be gained as to a chemical’s toxic mechanism of action through assessing changes in a suite of steroids. Consequently, including these additional steroids in experimental testing can provide a more complete picture of steroidogenesis in small fish.

Fig. 1.

Steroidogenic pathway in teleost fish. Steroids included in the method are in blue, bold font. Critical enzymes throughout the steroidogenic pathway are noted in italics along pathway arrows.

Though LC-MS is considered by many to be the preferred method for steroid measurements, the diversity of steroid structures and their physicochemical properties pose a challenge for LC-MS analysis. Many steroids do not ionize efficiently by atmospheric pressure ionization techniques due to their relatively low proton affinities [25]. In electrospray ionization (ESI) based methods, steroids are commonly derivatized to increase sensitivity. Previous LC methods have utilized dansyl chloride for derivatization of the phenolic hydroxyl group of estrogens or 2-hydrazinopyradine for derivatization of ketone-groups [26–29]. Atmospheric pressure chemical ionization (APCI), and atmospheric pressure photoionization (APPI) have also frequently been used for analysis of steroids [30–33]. APPI methods have been applied to quantify steroids in a variety of matrices and demonstrate a decrease in matrix effects due to the softer ionization [25, 27, 34, 35]. Unlike other ionization techniques, APPI is also capable of ionizing all steroids of interest, including estrogens, in positive ionization mode [35], removing the need for multiple injections using both negative and positive polarity or for an instrument capable of fast polarity switching.

Developing a derivatization-free method to measure a suite of steroids in small volumes of plasma or water was the primary goal of the present study. To accomplish this, APPI was selected as the ionization source, and a new LC-APPI-MS/MS method was developed to simultaneously quantify 13 endogenous steroids without the need for chemical derivatization. The final steroids selected for the method included key C21 progestogens and corticosteroids (11-deoxycortisol, 17α-hydroxyprogesterone, 17,20β-dihydroxyprogesterone, 17,20β,21-trihydroxyprogesterone, cortisol, progesterone), C19 androgens (11-ketotestosterone, androstenedione, testosterone), and C18 estrogens (17α- and 17β-estradiol, estriol, estrone) (Fig. 1). The method was applied to two separate experiments to demonstrate effectiveness in identifying perturbation of the HPG axis and steroidogenesis in small fish species. Experiment 1 analyzed plasma steroids from female fathead minnows while Experiment 2 analyzed tank water from a toxicity test with Japanese medaka as an alternative matrix for assessing impacts of a model EDC, fadrozole, on steroid synthesis.

2. Experimental

2.1. Chemicals and materials

HPLC grade water, methanol, and ethyl acetate, ACS grade ammonium hydroxide, and GC grade hexane and dichloromethane were purchased from Fisher Scientific (Hampton, NH). Formic acid, acetic acid, sodium acetate, and β-glucuronidase (Type HP-2) were from Sigma-Aldrich (St. Louis, MO). Charcoal stripped fetal bovine serum (FBS) was from GE Healthcare Hyclone (Chicago, IL). Fadrozole was provided as a gift from Novartis, Inc.

2.2. Standards

All chemical standards were ≥98% purity. Androstenedione (CASRN 63-05-8), cortisol (CASRN 50-23-7), 11-deoxycortisol (CASRN 152-58-9), 17β-estradiol (CASRN 50-28-2), estrone (CASRN 53-16-7), 17α-hydroxyprogesterone (CASRN 68-96-2), progesterone (CASRN 57-83-0), and testosterone (CASRN 58-22-0) were purchased as solutions from Sigma-Aldrich. Dehydroepiandrosterone (53-43-0), 17α-estradiol (CASRN 57-91-0), estriol (CASRN 50-27-1), 17α-hydroxypregnenolone (CASRN 387-79-1), 11-ketotestosterone (CASRN 564-35-2), pregnenolone (CASRN 145-13-1) were purchased as neat material from Sigma-Aldrich, while 17,20β-dihydroxyprogesterone (DHP; CASRN 1662-06-2) and 17,20β,21-trihydroxyprogesterone (20β-5; CASRN 128-19-8) were purchased from Steraloids as neat material (Newport, RI). Stable isotope labeled androstenedione (2,3,4-¹³C3), 17β-estradiol (13,14,15,16,17,18-¹³C6), 17α-hydroxyprogesterone (2,3,4-¹³C3), progesterone (2,3,4-¹³C3), and testosterone (2,3,4-¹³C3) were purchased as neat material from Cambridge Isotope Laboratories (Tewksbury, MA). Cortisol (9,11,12,12-d4), 11-deoxycortisol (2,2,4,6,6-d5), estriol (2,3,4-¹³C3), and estrone (2,3,4-¹³C3) were purchased as solutions from Sigma-Aldrich.

2.3. Standard stock and calibration curve preparations

Stock solutions of neat steroids were prepared in methanol (MeOH) at a concentration of 1.0 mg/mL. A mixed stock solution of all 16 target steroids was prepared at 2.0 μg/mL in MeOH, and serially diluted in 25:75 MeOH:water (v/v) to make standard or QC solutions. Stock solutions of neat internal standards were prepared in MeOH at a concentration of 1.0 mg/mL. A mixed stock solution of all nine internal standards was made at 200 ng/mL in 25:75 MeOH:water (v/v). All solutions were stored at −20°C and brought to room temperature before use.

Calibration standards were prepared by adding 100 μL of internal standards to 900 μL of target standards. Standards were prepared at two different concentration ranges depending on the matrix type. For plasma, standards were prepared at 0.01, 0.05, 0.1, 0.5, 1.0, 5.0, and 10 ng/mL, and for water analysis standards were prepared at 0.1, 0.5, 1.0, 5.0, 10, 50, and 100 ng/mL. Quality control (QC) samples were prepared by spiking a known amount of target standards into charcoal stripped FBS, or Lake Superior water. Samples were then processed as described below for each sample matrix.

2.4. Fish exposure and sample collection

2.4.1. Experiment 1: Fathead minnow plasma analysis

Female fathead minnows were exposed for 48 h to waterborne fadrozole (FAD; 0, 0.5, 5 μg/L; n = 12 individuals/treatment, 6 individuals/replicate), a well characterized cytochrome P450 19A1A (aromatase) inhibitor (see Ankley 2020 for additional exposure details [36]). Aromatase catalyzes the conversion of androgens such as testosterone to estrogens like 17β-estradiol in vertebrates, and our laboratory has used it as a model EDC in a variety of studies [6]. Fadrozole stock solutions were prepared without organic solvent by sonicating neat FAD in ultrapure water, bringing to final volume, and stirring for a minimum of 24 h. Exposures were conducted using a constant flow (approx. 45 mL/min) of water from Lake Superior. After exposure, fish were anaesthetized with buffered tricaine methanesulfonate and blood collected. Plasma was obtained by centrifugation and stored at −80°C until extraction.

If needed, plasma from individual fish within treatment and replicate were combined to reach a total volume between 8 – 10 μL per sample. Because of this, sample numbers vary across treatments (FAD 0: n = 6; FAD 0.5: n = 7; FAD 5: n = 7). Plasma was transferred to a 1.5 mL microcentrifuge tube, spiked with 5 μL (0.125 ng) of mixed internal standards, diluted with 1% formic acid in water to a final volume of 200 μL, and vortexed to mix thoroughly. Plasma was extracted using supported liquid extraction (SLE) cartridges (Phenomenex Novum, 1cc). Diluted plasma samples were loaded onto SLE cartridges and allowed to sorb for 5 min. Cartridges were washed with 0.75 mL hexane (discarded) then eluted with 1.2 mL of dichloromethane. Extracts were evaporated to dryness under a gentle stream of nitrogen at 35°C, reconstituted in 50 μL of 75:25 water:MeOH (v/v), and transferred to a high recovery autosampler vial. Vials were capped and stored at −20°C until analysis.

2.4.2. Experiment 2: Japanese medaka exposure water analysis

Male and female Japanese medaka were exposed to waterborne FAD (0, 2, 10, 30 μg/L in a group spawning design (8 males, 8 females per tank, n = 3 tanks/treatment) for 21d (see Doering et al. for additional exposure details [37]). Fadrozole stock solutions were prepared without organic solvent by sonicating neat FAD in ultrapure water, bringing to final volume, and stirring for a minimum of 24 h. Exposures were conducted under constant flow (approx. 45 mL/min) of water from Lake Superior. During exposure, 0.25 L of tank water was collected periodically (days −1, 1, 2, 4, 21 of exposure) and immediately extracted for steroid analysis.

Water extraction was adapted from Ma et al.[38] using solid phase extraction (SPE, Phenomenex Strata-×, 200mg) to capture both free and conjugated steroids. Water collected during the Japanese medaka exposure was spiked with 25 μL (0.625 ng) of mixed internal standards and allowed to equilibrate for 15 min. Extraction cartridges were conditioned sequentially with 6 mL each of ethyl acetate, methanol, and water and then samples loaded onto cartridges. After loading, cartridges were washed with 10 mL each of 93:5:2 water:MeOH:acetic acid (v/v/v) and 93:5:2 water:MeOH:ammonium hydroxide (v/v/v) then aspirated to dryness. Steroids and steroid conjugates were eluted sequentially with 4 mL 90:10 ethyl acetate:MeOH (v/v) and 4 mL 2% NH40H in MeOH. Extracts were evaporated to dryness under a gentle stream of nitrogen at 35°C.

Enzymatic deconjugation of excreted steroids allows for measurement of total steroids in exposure water, increasing the concentration of free compounds and reducing the need to include individual conjugates in an analytical suite. Consequently, extracts were enzymatically digested using β-glucuronidase (Millipore Sigma, Type HP-2) to enable detection of both free and conjugated steroids in the free form. Dried extracts were reconstituted in 200 μL of 100 mM sodium acetate (pH 5.2) containing 6250 units/mL of β-glucuronidase. Samples were vortexed gently, covered, and incubated at 50°C for 2h. After incubation, samples were cooled to room temperature and cleaned up by SLE (Phenomenex Novum, 3cc). Samples were loaded onto SLE cartridges and allowed to sorb for 5 min. Cartridges were washed with 0.75 mL hexane (discarded) then eluted with 1.8 mL of dichloromethane. Extracts were evaporated to dryness under a gentle stream of nitrogen at 35°C, reconstituted in 250 μL of 75:25 water:MeOH (v/v), and transferred to a high recovery autosampler vial. Vials were capped and stored at −20°C until analysis.

2.5. Liquid chromatography atmospheric pressure photoionization tandem mass spectrometry (LC-APPI-MS/MS)

The LC-MS system consisted of an Agilent 1290 Infinity LC (Agilent Technologies Inc., CA, USA) coupled to an Agilent 6490 triple quadrupole mass analyzer with an APPI ion source. Separation was performed using a Zorbax RRHD SB-C18 column (2.1 × 50mm, 1.8μm; Agilent Technologies Inc) held at 40°C. Gradient flow consisted of 95:5 water:MeOH (mobile phase A; v/v) and 95:5 MeOH:water (mobile phase B; v/v) at a flow rate of 0.4 mL/min. The gradient program began at 22% B, held for 0.25 min, ramped to 83% B at 10 min, held at 100% B for 2 min, and re-equilibrated at 22% B for 2 min for a total 14 min run. Under these conditions, target steroids eluted in under 10 min (Fig. 2). The mobile phase was directed to waste from 0-4 min, to the mass spectrometer from 4-10 min, and again to waste for the remainder of the run. A 20 μL injection volume was used for both sample types.

Fig. 2.

Chromatogram of mixed steroid standard at 5 ng/mL. Analytes are numbered as follows: 1) estriol; 2) cortisol; 3) 11-ketotestosterone; 4) 17,20β,21-trihydroxyprogesterone; 5) 11-deoxycortisol; 6) 17β-estradiol; 7) estrone; 8) 17α-estradiol; 9) androstenedione; 10) testosterone; 11) 17α-hydroxyprogesterone; 12) 17,20β-dihydroxyprogesterone; 13) progesterone.

The mass spectrometer was operated in positive APPI mode using toluene as a dopant at a flow rate of 0.04 mL/min. Ion source gas flow was 11 L/min, gas temp 290 °C, desolvation heater 450°C, and capillary voltage 4.5 kV. Mass transitions for individual compounds were determined and optimized using flow injection analysis (Table 1). Analysis was performed using the dynamic multiple reaction monitoring (MRM) setting, with dwell time ranging 23 – 400 ms depending on the number of concurrent MRMs. All chromatographic data were analyzed using Agilent MassHunter Quantitative Analysis software (v.B.06.00). Calibration curves were generated as the relative response ratio of analyte area to internal standard area and fit with a linear curve with 1/× weighting to account for heteroscedasticity.

Table 1.

Optimized ion transitions, collision energy, retention time, and corresponding internal standard for each analyte. Refer to Table 2 for internal standard details.

Compound	Precursor Ion (m/z)	Product Ions (m/z)a	Collision Energy (V)	Retention Time (min)	Internal Standard
estriol	271.1 [M+H-H20]+	133.1, 159.1	29, 25	4.21	1
cortisol	363.2 [M+H]+	121.1, 97.1	33, 37	5.52	2
11-ketotestosterone	303.3 [M+H]+	121.1, 259.1	33, 25	5.80	3
17,20β,21-trihydroxyprogesterone	313.3 [M+H-2(H20)]+	97.1, 109.1	37, 33	6.75	3
11-deoxycortisol	347.2 [M+H]+	109.1, 97.1	33, 41	6.83	3
17β-estradiol	255.1 [M+H-H20]+	159.1, 133.1	17, 25	7.07	4
estrone	271.1 [M+H]+	133.1, 159.1	21, 25	7.16	5
17α-estradiol	255.1 [M+H-H20]+	159.1, 133.1	17, 25	7.28	4
androstenedione	287.2 [M+H]+	97.1, 109.1	25, 29	7.44	6
testosterone	289.2 [M+H]+	97.1, 109.1	29, 33	7.96	7
17α-hydroxyprogesterone	331.3 [M+H]+	109.1, 97.1	29, 45	8.05	8
17,20β-dihydroxyprogesterone	333.3 [M+H]+	97.1, 109.1	37, 45	8.24	8
progesterone	315.3 [M+H]+	97.1, 109.1	25, 29	9.33	9

^aIons listed as quantifier ion, qualifier ion.

2.6. Method validation

2.6.1. Accuracy and precision

Due to the small volume of plasma obtained from fathead minnows, method validations for plasma were performed using charcoal stripped FBS. Precision and accuracy of the developed method were tested by extracting and analyzing replicate QC samples (n = 9) of spiked, charcoal stripped FBS. Steroids were spiked at a concentration of 0.5 ng/mL except for 20β-S, cortisol, and estriol, which were spiked at 2.5 ng/mL. Precision and accuracy of the method using water samples were tested by extracting and analyzing replicate QC samples (n = 7 total) of Lake Superior water. Steroids were spiked at a concentration of 2.0 ng/L (n = 4) or 4.0 ng/L (n = 3). Due to a lack of detection of 20β-S and E3 in any plasma samples assessed, these compounds were omitted in water samples and should be considered semi-quantitative in tank water samples. Percent precision and percent accuracy were defined as: % precision = (sample standard deviation/sample mean) × 100; % accuracy = (sample mean/sample nominal concentration) × 100.

2.6.2. Method Detection Limit and Quantitation Limit

The method detection limit (MDL), defined as the minimum concentration that can be measured and reported with 99% confidence that the analyte concentration is greater than method blanks [39], was determined in plasma using the same set of spiked charcoal stripped FBS (n = 9). The MDL is statistically determined by calculating the standard deviation, multiplied by Studen’s t-test value (n-1 degrees of freedom, 99% confidence level, 1-tailed) for each analyte. The lower limit of quantification (LLOQ) was also determined at the lowest calibrator concentration that achieved acceptable precision ≤20% and accuracy ≥80% and a signal to noise ratio ≥10.

2.7. Statistical Analysis

All statistical analyses were performed using Microsoft Excel for Office 365 or GraphPad Prism (v.7.04). Differences in plasma steroid concentrations between treatments were assessed using a one-way analysis of variance (ANOVA) followed by a Dunnet’s multiple comparison test. Differences in waterborne steroid concentrations between treatments were assessed for individual time points using a one-way ANOVA followed by a Dunnet’s multiple comparison test. Differences were considered significant at a 95% confidence level (p ≤ 0.05).

3. Results and Discussion

3.1. Steroid tuning and mass transitions

Each individual compound was optimized by injecting 1 μL of a 1-10 μg/mL solution with no analytical column. A Q1 full scan spectrum was generated to identify the highest intensity ion for precursor ion selection. One compound from each class of steroid (17β-estradiol, cortisol, testosterone, progesterone, pregnenolone, DHP) was initially tuned to determine the optimal APPI source parameters for analysis. It quickly became apparent that APPI probe temperature settings are of critical importance for APPI ionization of steroids. In positive ionization mode, 17β-estradiol produces few ions below a source temperature of 400°C, after which the dominant ion produced is the dehydrated molecular ion [M+H-H20]+ (Fig. S1). Conversely, polyhydroxylated compounds primarily produce the protonated molecular ion [M+H]+ at much lower temperatures from 150-250°C. Increasing source temperature leads to an increase in dehydration of the molecules and a reduction of precursor ion signal (Fig. S2). These results align with previous studies of APPI ionization of steroids, demonstrating a strong influence of APPI vaporizer temperature on in-source fragmentation [25, 34]. Due to the importance of measuring estrogens in studies with EDCs, the final ion source settings were selected to maximize signal of estrogenic compounds. Thus, compounds with the greatest degree of hydroxylation (cortisol, 20β-S) demonstrated the worst performance and highest detection limits (Fig. 2). For studies where improved sensitivity of progestogens and corticosteroids is needed, a reduction in source temperature could be used to enhance performance.

Compound optimization also identified poor performance for the three pregn-5-ene compounds (pregnenolone, dehydroepiandrosterone, 17αOH-pregnenolone) using APPI. The dehydrated [M+H-H20]+ ion was the dominant precursor ion, but poor product ion efficiency led to a lack of stable ion transitions. Quantifiable concentrations of the pregn-5-ene compounds were an order of magnitude above other compounds, so the class was ultimately removed from the final analytical method.

After optimizing source parameters for structures of greatest interest and identifying the dominant precursor ions, each compound was individually tuned to identify optimal collision energy (CE) and subsequent product ions for multiple reaction monitoring (MRM) analysis (Table 1). Two ion transitions were monitored for each compound, and the most abundant transitions were selected as the quantifier ion. Nonspecific transitions (e.g. −H2O) were not selected as quantifier ions to avoid potential matrix-based interferences. Most of the selected steroids have product ions in common. Estrogens form two primary product ions at 159 and 133 m/z, likely representing the hydroxydihydronaphthalene cation [C11H11O]+ and vinyl phenol radical [C9H9O]+, respectively [34, 40]. Most other steroids produce two primary ions 97 and 109 m/z, which represents the protonated cyclic ketone [C6H9O]+ and [C7H9O]+, respectively [34, 41].

3.2. Method validation

A series of calibration standards were prepared including internal standards at a constant concentration of 2.5 ng/mL. Due to the varying concentration ranges expected for the two sample matrices being analyzed, standards were analyzed across two different concentration ranges. For plasma, standards ranged from 0.01 – 10 ng/mL while for water standards ranged from 0.05 ng/mL – 100 ng/mL. All compounds demonstrated linearity over the calibration ranges (R² values ≥0.99).

3.2.1. Accuracy and precision

Precision and accuracy of the developed method (Table 3) were tested by extracting and analyzing replicate QC samples (n = 9) of spiked charcoal stripped FBS. Steroids were spiked at a concentration of 0.5 ng/mL except for 20β-S, cortisol, and estriol, which were spiked at 2.5 ng/mL. Accuracy ranged from 93 – 121% and precision ranged from 3.2 – 13.5% across analytes. Accuracy is similar to reported accuracy in RIAs, which reported under 10% coefficient of variation for estradiol, testosterone, and 11-ketotestosterone [13]. Precision and accuracy of the method using water samples (Table 3) was tested by extracting and analyzing replicate QC samples (n = 7 total) of Lake Superior water. Steroids were spiked at a concentration of 2.0 ng/L (n = 4) or 4.0 ng/L (n = 3). Accuracy ranged from 93 – 121% and precision ranged from 3.2 – 13.5% across analytes. The analytes 20β-S and estriol were not assessed in water samples and should be considered semi-quantitative in tank water samples. Cortisol demonstrated poor precision (31%) and should also be viewed as semi-quantitative in water analyses.

Table 3.

Method detection limit (MDL) and lower limit of quantification (LLOQ) for analytes in both fish plasma and exposure water.

Compound	Plasma % Accuracy	Plasma % Precision	Plasma MDL (ng/mL)a	Plasma LLOQ (ng/mL)a	Water % Accuracy	Water % Precision	Water LLOQ (ng/L)b
estriol	97	7.3	0.5	1.0	NA	NA	0.5
cortisol	110	7.3	0.6	1.0	99	31.2	0.5
11-ketotestosterone	104	5.5	0.08	0.3	100	20.4	0.1
17,20β,21-trihydroxy-progesterone	110	13.4	1.0	2.0	NA	NA	1.0
11-deoxycortisol	99	6.6	0.09	0.3	80	5.3	0.1
17β-estradiol	101	6.4	0.09	0.3	95	12.8	0.1
estrone	93	12.3	0.16	0.3	109	13.3	0.1
17α-estradiol	95	11.0	0.15	0.3	76	7.1	0.1
androstenedione	121	13.5	0.24	0.3	93.5	7.1	0.1
testosterone	109	11.4	0.18	0.3	101.5	11.3	0.1
17α-OH-progesterone	95	5.2	0.07	0.3	93	3.5	0.1
17,20β-dihydroxy-progesterone	97	4.4	0.06	0.3	96	3.2	0.1
progesterone	102	3.2	0.05	0.1	92	6.0	0.05

NA = not available.

^aValues calculated using 10 μL of plasma.

^bValues calculated using 0.25 L of exposure water.

3.2.4. Method detection limit and quantification limit

The MDL and LLOQ varied across analytes, with those forming the molecular ion in-source demonstrating the lowest detection limits and those with a greater degree of hydroxylation and in-source fragmentation having higher detection limits (Table 3). Using 10 μL of plasma, MDLs ranged from 0.05 – 1.0 ng/mL and LLOQs ranged from 0.1 – 2.0 ng/mL. Detection limits are similar to comparable RIAs in fathead minnow plasma, with reported detection limits for estradiol, testosterone, and 11-ketotestosterone ranging 0.2 – 1.6 ng/mL when using 6 μL of plasma[13]. An MDL was not determined for waterborne steroids, and only a LLOQ was defined (Table 3). Using 0.25 L of tank water, LLOQs ranged from 0.05 – 1.0 ng/L.

3.3. Experiment 1 – Application to fathead minnow plasma

After validation the method was applied to quantify concentrations of steroids in female fathead minnow plasma following exposure to FAD or control Lake Superior water. Fadrozole has been previously shown to inhibit 17β-estradiol production in female fathead minnows as measured by RIA [10, 42]. After 48 h, fathead minnows exposed to 0.5 μg/L FAD showed no change in 17β-estradiol relative to controls while the 5 μg/L treatment showed a significant reduction in 17β-estradiol (Fig. 3). Plasma androstenedione and testosterone had seemingly increased concentrations in both treatments, although due to variability these changes were not statistically significant (Fig. 3). An increasing trend in both androstenedione and testosterone is consistent with the action of FAD as an aromatase inhibitor. Reduced conversion of androstenedione and testosterone to estrone and 17β-estradiol, respectively, could plausibly lead to their accumulation in gonad tissue and/or plasma. While this has long been hypothesized, the limited plasma volumes available for RIA analyses have historically prevented us from measuring plasma androstenedione and testosterone concentrations in previous experiments with aromatase inhibitors (9, 10). The current method will allow this hypothesis to be explored in a more robust fashion.

Fig. 3.

Plasma concentrations of 17β-estradiol, androstenedione, and testosterone (panels A, B, C, respectively) from exposure of female fathead minnow to varying concentrations of fadrozole (Table S1). Dots represent individual measurements (n = 6 or 7) with mean and SEM represented as the midline and error bars. Lowercase letters (estradiol only) indicate significant differences between treatments (p<0.05).

Looking broadly at plasma steroids, four steroids (androstenedione, 17β-estradiol, cortisol, and testosterone) were detected in almost all samples (SI Table S1), regardless of treatment. Other steroids were detected less frequently, and estriol, 17α-hydroxyprogesterone, and 20β-S were not detected in any plasma samples (SI Table S1). Many of these steroids have not been commonly measured in fathead minnow plasma in the past, so this dataset represents a first profiling of multiple androgens, corticosteroids, estrogens, and progestogens simultaneously in plasma.

3.4. Experiment 2 – Application to exposure water

The method was also applied to quantify excreted steroid hormones as a surrogate for plasma measurements in spawning Japanese medaka exposed to FAD or control Lake Superior water. All steroids except 20β-S and estriol were detected in one or more treatments (SI Table S2, Fig. S3). As expected, 17β-estradiol was the most sensitive endpoint showing a significant decrease relative to controls observable after 24 h of exposure to all FAD treatments (Fig. 4). Concentrations of 17β-estradiol decreased in a dose-dependent manner and remained significantly reduced throughout the experiment (Fig. 4). [37]Excreted estrone showed a similar trend to 17β-estradiol and was significantly reduced after 24 h of FAD exposure (Fig. S3). Excreted testosterone was increased after 24 h of exposure to 30 μg/L FAD and continued to increase throughout the 21 d exposure (Fig. 4). Excreted androstenedione was significantly increased after 48 h of exposure to 30 μg/L FAD and remained significantly elevated after 21 d (Fig 4). Two other steroids, 11-deoxyortisol and progesterone were significantly increased in the 30 μg/L FAD treatment after 21 d (Fig. S3). Plasma 17β-estradiol concentrations were measured after 21 d by RIA, and a significant reduction was observed in both the 10 and 30 μg/L treatments (see [37]). The comparable results indicate tank water can serve as an appropriate surrogate for plasma estradiol measurements. While direct comparison of other steroids was not possible in this study, one previous study has demonstrated positive correlation in cortisol concentration in plasma and holding water [21], supporting the hypothesis that holding water can serve as a surrogate for plasma across multiple steroid classes. More direct comparisons of both plasma and water are needed to further validate this hypothesis.

Fig. 4.

Waterborne concentrations of 17β-estradiol, androstenedione, and testosterone (panels A, B, C, respectively) from exposure water holding spawning Japanese medaka exposed to varying concentrations of fadrozole for 21 d (n = 3) at −1, 1, 2, 4, and 21 d of exposure (SI Table S2). Dots and error bars represent the mean ± SEM. Asterisks indicate significant differences of treatments relative to controls (p<0.05) within a single time point.

To look at excretion profiles of free and conjugated steroids, a single control sample was analyzed without glucuronidase enzymatic digestion. Androstenedione, cortisol, and 17α-hydroxyprogesterone were present in the free form, about 40% of estrone was conjugated, and the remaining steroids were primarily present as conjugates (Fig. S4). Measurement of only free steroid would be the optimal surrogate for plasma measurements, but the low concentrations of free steroids make application to flow-through experiments difficult. Overall, however, these data demonstrate that steroid measurements in exposure water from a flow-through experiment were able to identify impacts of aromatase inhibition within 24 h of exposure, showing promise for the use of waterborne steroids as a noninvasive surrogate for plasma steroids.

4. Conclusions

An LC-APPI-MS/MS method for the simultaneous measurement of 13 androgens, corticosteroids, estrogens, and progestogens was developed and validated for use in small fish toxicity testing. The method shows satisfactory performance and increased the number of analytes that can be measured relative to comparable immunoassay-based methods. The method was applied to both small fish plasma and exposure water, demonstrating use for a standard toxicity testing endpoint (plasma) as well as a potential noninvasive surrogate (exposure water). Method development also identified optimal ion source parameters for different steroid classes, allowing for future flexibility and utility if additional steroid analytes or lower detection limits for specific steroid classes (e.g. corticosteroids) are needed.

Table 2.

Optimized internal standard ion transitions and retention time. Internal standard number refers to Table 1.

Internal Standard Compound	Number	Precursor Ion (m/z)	Product Ion (m/z)	Collision Energy (V)	Retention Time (min)
2,3,4-13C3-estriol	1	274.1	136.1	29	4.21
cortisol-d4	2	367.2	121.1	33	5.51
11-deoxycortisol-d5	3	352.2	113.1	33	6.79
13,14,15,16,17,18-13C6-estradiol	4	261.1	159.1	17	7.07
2,3,4-13C3-estrone	5	274.1	162.1	25	7.16
2,3,4-13C3-androstenedione	6	290.2	100.1	25	7.44
2,3,4-13C3-testosterone	7	292.2	100.1	29	7.96
2,3,4-13C3-17α-OH-progesterone	8	334.3	112.1	29	8.05
2,3,4-13C3-progesterone	9	318.3	100.1	25	9.33