Conversion of Classical and 11-Oxygenated Androgens by Insulin-Induced AKR1C3 in a Model of Human PCOS Adipocytes

Abstract

Polycystic ovary syndrome (PCOS) is the most prevalent endocrinopathy in women. A common symptom of PCOS is hyperandrogenism (AE); however, the source of these androgens is uncertain. Aldo-keto reductase family 1 member C3 (AKR1C3) catalyzes the formation of testosterone (T) and 5α-dihydrotestosterone (DHT) in peripheral tissues, which activate the androgen receptor (AR). AKR1C3 is induced by insulin in adipocytes and may be central in driving the AE in PCOS. We elucidated the conversion of both classical and 11-oxygenated androgens to potent androgens in a model of PCOS adipocytes. Using high-performance liquid chromatography (HPLC) discontinuous kinetic assays to measure product formation by recombinant AKR1C3, we found that the conversion of 11-keto-Δ4-androstene-3,17-dione (11K-4AD) to 11-ketotestosterone (11K-T) and 11-keto-5α-androstane-3,17-dione (11K-5AD) to 11-keto-5α-dihydrotestosterone (11K-DHT) were superior to the formation of T and DHT. We utilized a stable isotope dilution liquid chromatography high resolution mass spectrometric (SID-LC-HRMS) assay for the quantification of both classical and 11-oxygenated androgens in differentiated Simpson-Golabi-Behmel syndrome adipocytes in which AKR1C3 was induced by insulin. Adipocytes were treated with adrenal derived 11β-hydroxy-Δ4-androstene-3,17-dione (11β-OH-4AD), 11K-4AD, or Δ4-androstene-3,17-dione (4AD). The conversion of 11β-OH-4AD and 11K-4AD to 11K-T required AKR1C3. We also found that once 11K-T is formed, it is inactivated to 11β-hydroxy-testosterone (11β-OH-T) by 11β-hydroxysteroid dehydrogenase type 1 (HSD11B1). Our data reveal a unique role for HSD11B1 in protecting the AR from AE. We conclude that the 11-oxygenated androgens formed in adipocytes may contribute to the hyperandrogenic profile of PCOS women and that AKR1C3 is a potential therapeutic target to mitigate the AE of PCOS.

Polycystic ovary syndrome (PCOS) is the most prevalent endocrine disorder in women, affecting 8% to 13% of reproductive-aged women (1-4). According to the Rotterdam Criteria, to be diagnosed with PCOS, a patient must have 2 of the 3 following symptoms: oligo-anovulation, polycystic ovarian morphology, and hyperandrogenism (1, 4). The most common of the 3 is hyperandrogenism, often referred to as androgen excess (AE). The ovaries were believed to be the dominant source of androgens in women, and at the present time, serum testosterone (T) levels are used as diagnostic biomarkers of AE in PCOS (2, 5). As a result, PCOS women with AE are prescribed oral contraceptives to suppress ovarian androgen biosynthesis and raise steroid hormone binding globulin levels to lower free androgens, yet their hyperandrogenism can still persist (6-9). This suggests that nonovarian androgens produced in peripheral tissues may contribute to the AE of PCOS.

Aldo-keto reductase family 1 member C3 (AKR1C3, also known as type 5 17β-hydroxysteroid dehydrogenase) predominantly catalyzes conversion of androgens such as Δ4-androstene-3,17-dione (4AD) to T and 5α-androstane-3,17-dione (5AD) to 5α-dihydrotestosterone (DHT) (10, 11) in peripheral tissues. Additionally, AKR1C3 can convert 11-keto-Δ4-androstene-3,17-dione (11K-4AD) to 11-keto-testosterone (11K-T) and 11-keto-5α-androstane-3,17-dione (11K-5AD) to 11-keto-5α-dihydrotestosterone (11K-DHT) (12, 13). T and DHT are the classical potent agonists for the androgen receptor (AR), while 11K-T and 11K-DHT are equipotent to T and DHT, as seen in AR reporter gene assays (14). Besides the ovaries, another source of T and 11K-T could be adipocytes, where AKR1C3 is induced by insulin (15, 16) (Fig. 1). Insulin resistance correlates with AE in PCOS women (17), and insulin resistance has been accepted as a common characteristic of PCOS women. Additionally, AKR1C3 expression is greater in PCOS adipocytes vs control adipocytes (16). Therefore, an alternative source of AE in PCOS women could be due to insulin-induced AKR1C3 conversion of classical and 11-oxygenated androgens in white adipose tissues.

Figure 1.

Insulin-induced AKR1C3 converts both classical and 11-oxygenated androgens into the potent analogs T, DHT, 11K-T, and 11K-DHT, which all activate the AR in adipocytes and may contribute to AE in women with PCOS.

The source of precursors for T and DHT in the adipocyte could be the adrenal glands, which primarily produce dehydroepiandrosterone (DHEA) and 4AD. The adrenals also produce 11β-hydroxy-Δ4-androstene-3,17-dione (11β-OH-4AD) (18) by 11β-hydroxylase in the zona reticularis (19). In peripheral tissues, 11β-OH-4AD can be metabolized to 11K-4AD by 11β-hydroxysteroid dehydrogenase type 2 (HSD11B2), which could be converted to 11K-T by AKR1C3 (12, 20, 21). 11β-OH-4AD in the circulation of PCOS women ranges from 1.3 to 32 nM and by contrast, 11K-4AD ranges from 0.65 nM to 13 nM. In every instance, 11β-OH-4AD levels exceed 11K-4AD levels by at least 2.6-fold (22-25). This suggests that 11β-OH-4AD may act as a major precursor for 11K-T and 11K-DHT. We hypothesize that 11β-OH-4AD and 11K-4AD produced from the adrenals are converted into the potent androgens, 11K-T and 11K-DHT, in adipocytes in an AKR1C3-dependent manner, thereby promoting AE in PCOS.

It is currently unknown if adipocytes metabolize 11β-OH-4AD and 11K-4AD to 11K-T and 11K-DHT, and whether AKR1C3 plays a dominant role in the conversion. Knowledge of this mechanism would identify AKR1C3 inhibitors as potential therapeutics to reduce the androgen burden in women with PCOS. We investigated the metabolism of 11β-OH-4AD and 11K-4AD by mass spectrometry in human Simpson-Golabi-Behmel syndrome (SGBS) subcutaneous preadipocytes following their differentiation in mature adipocytes, as a model for PCOS adipocytes. This work implicates insulin-induced AKR1C3 in adipose tissue as a source of 11K-T derived from adrenal 11β-OH-4AD, which may contribute to the androgenic profile of PCOS women.

Materials and Methods

Chemicals and Reagents

4AD, T, 5AD, DHT, 11K-5AD, 11K-DHT, and 11β-OH-5AD were purchased from Steraloids (Newport, RI). 11β-OH-DHT was purchased from Isosciences (Ambler, PA). 11K-4AD, 11K-T, 11β-OH-4AD, and 11β-OH-T were gifts provided by the Rainey laboratory (University of Michigan). [¹³C]-4AD was purchased from Cerilliant (Round Rock, TX). [¹³C]-DHEA, [¹³C]-T, and [¹³C]-DHT were purchased from Cambridge Isotope Laboratories (Andover, MA). Steroid identity and purity were assessed by ultraviolet (UV)–high-performance liquid chromatography (HPLC) when possible, followed by high-resolution mass spectrometry (HRMS). All other chemicals were of the highest grade available, and all solvents were HPLC grade. BMT4-159 (2-{[3-(trifluoromethyl)phenyl]amino}-benzoic acid) and ASP9521 (1-(-{1-[(5-methoxy-1H-indol-2-yl)carbonyl] piperidin-4-yl}-2-methylpropan-2-ol) were synthesized in-house as described by us (26, 27). Indomethacin was purchased from ICN Biomedicals Inc (Costa Mesa, CA). BVT14225 (2-[2-[(3-Chloro-2-methylphenyl)sulfonylamino]-1,3-thiazol-4-yl]-N,N-diethylacetamide) was purchased from Med Chem Express (Monmouth Junction, NJ).

Discontinuous Reverse Phase HPLC Kinetic Analysis

Recombinant AKR1C3 was purified from Escherichia coli C41 (DE3) cells following transformation with a pET16b-AKR1C3 vector by sequential chromatography by fast protein liquid chromatography (FPLC) on DEAE-Sepharose and Hi-Trap Blue Sepharose HP resin. Enzyme specific activity was monitored by measuring the NAD⁺dependent oxidation of S-tetralol (28). Specific activity was found to fall within the acceptable range of 2.4 to 2.6 µmoles S-tetralol oxidized/min/mg under standard conditions and purity was accessed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) (29, 30). 4AD, T, 11K-4AD, and 11K-T were titrated at 242 nm (E = 17 200 M^-1cm^-1) to determine their concentrations.

Kinetic reactions with recombinant AKR1C3 were performed in systems containing 50 mM potassium phosphate pH 7.0, 180 µM NADPH (Sigma, St. Louis, MO), varying concentrations of the androgen in 4% acetonitrile, and recombinant enzyme. Timed 500 µL reactions were continuously stirred in a 37 °C water bath. Upon completion, the reaction was quenched and extracted with 1 mL of ethyl acetate. This liquid-liquid extraction was repeated for a second time. The organic phases were pooled and dried in a vacuum centrifuge at 35 °C. The residue was resuspended in 90 µL of acetonitrile and an aliquot was injected onto the HPLC. Substrate and product were separated on a Waters 2695 Separations Module with Waters 996 Photodiode Array Detector and Agilent Zorbax ODS 5 um, 4.6 × 250 mm C18 column (880 952-702) using a mobile phase that went from 85% A (water) and 15% B (acetonitrile) to 100% B over 35 minutes at a flow rate of 0.5 mL/min. The column was washed at 100% B for 5 minutes and then was taken to 95% A and 5% B over 1 minute for 14 minutes, and then returned to 85% A and 15% over 1 minute and re-equilibrated for 4 minutes. Reactions with 4AD, 11K-4AD, and 11β-OH-4AD were monitored at 242 nm. Peak areas were integrated and product formation was quantified using calibration curves constructed against authentic synthetic standards. The rate of product formation was monitored as pg of product formed per minute. Velocity (v) was then plotted against substrate concentration to generate v vs S curves. Global curve fitting was analyzed with PRISM graphpad (version 9.0) to determine the K_m and k_cat of each reaction. Reactions were initially accessed for product formation over 3 hours and optimized in terms of linearity with time and enzyme concentration.

Derivatization With Dinitrophenylhydrazine and Quantitation of Steroid Hydrazones

As 5AD and 11K-5AD are UV transparent, reactions containing these substrates required post derivatization with dinitrophenylhydrazine (DNPH) (31, 32), to form hydrazones that are UV-visible. Following extraction, samples were dried down and derivatized with 200 µL of 1% trifluoroacetic acid containing 20 µL of fresh Brady’s reagent (1 mg DNPH/mL in methanol and 100 µL of concentrated sulfuric acid per mL). Reactions were incubated at 60 °C for 20 minutes. Reactions were dried and residues were dissolved in methanol and analyzed by reverse phase (RP)-HPLC using ion-pair chromatography in which solvent A was 10 mM ammonium acetate, 0.05% formic acid in water pH 4.25 and solvent B was 10 mM ammonium acetate, 0.05% formic acid in 60/40 methanol/acetonitrile pH 5.6. To separate 5AD and DHT-DNPH derivatives, the method began at 20% A and 80% B with a flow of 1 mL/min. The flow was increased to 1.5 mL/min, and a gradient to 93.7% B was applied over 30 min. The column was taken to 100% B for 0.1 minute and washed under these conditions for 9 minutes before being returned to 20% A and 80% B over 0.1 minute and the system re-equilibrated for 6 minutes. To separate 11K-5AD and 11K-DHT-DNPH derivatives, the method began at 50% A and 50% B at a flow rate of 1 mL/min. A gradient to 100% B was applied over 50 minutes and the column was washed at 100% B for 8 minutes. The system returned to 50% A and 50% B over 1 minute and was re-equilibrated for 6 minutes. Substrate and product were monitored at 370 nm and product was quantified using calibration curves constructed from known amounts of derivatized authentic synthetic standards.

SGBS Cell Culture and Differentiation

SGBS preadipocytes were a kind gift from the Wabitsch laboratory at the University of Ulm and grown and differentiated into mature adipocytes over 14 days based on their protocol (33). Preadipocytes were passaged in growth media (Dulbecco’s Modified Eagle Medium/Nutrient Mixture F12 [DMEM/F12], 10% fetal bovine serum [FBS], 1% Pen/Strep, 33 µM biotin, 17 µM pantothenic acid) and were maintained up to 50 generations at a growth rate of 2 generations per 3 days. In preparation for differentiation, preadipocytes are seeded into 6 cm dishes (Corning, NY) and grown for 3 days to reach 600 000 cells at 90% confluency. After 3 days of growth, cells were washed with phosphate-buffered saline and quick differentiation was initiated with differentiation media (growth media without FBS) containing 10 µg/mL apo-transferrin, 20 nM insulin, 100 nM cortisol, 200 pM triiodothyronine, 25 nM dexamethasone, 250 µM isobutylmethylxanthine, and 2 µM rosiglitazone. After 4 days, the quick differentiation media was removed and replaced with a similar differentiation media as before without 25 nM dexamethasone, 250 µM isobutylmethylxanthine, and 2 µM rosiglitazone. Cells were phenotyped using Oil Red O staining for lipid droplet formation, which was normalized to cell count by Hoechst nuclei staining and imaged using a Biotek Cytation 5 plate imager.

Real-Time Quantitative Polymerase Chain Reaction

RNA was extracted with 600 µL of lysis buffer with 2-mercaptoethanol by RNeasy mini kit (Qiagen, Hilden, Germany) and treated with DNase (Qiagen) as per protocol. RNA concentrations are determined with a Nanodrop ND-2000 spectrometer. RNA was reverse transcribed to cDNA using High Capacity RNA to cDNA kit (Applied Biosystems/Thermo, Waltham, MA). Expression was measured by QuantiTect SYBR Green PCR Master Mix (Qiagen) in the Chromo4 System.

The conditions for the real-time polymerase chain reaction (PCR) for AKR1C3 were 95 °C for 15 minutes followed by 40 cycles of 94 °C for 15 seconds, 57 °C (annealing temperature) for 30 seconds, and 72 °C (extension temperature) for 30 seconds. The conditions for the real-time PCR for αTubulin were 95 °C for 15 minutes followed by 40 cycles of 94 °C for 15 seconds, 51 °C (annealing temperature) for 30 seconds, and 72 °C (extension temperature) for 30 seconds. Standard curves were generated using full-length standards (2 500 000 to 0.025 fg) as previously described by us (34). Full-length standards were generated from their appropriate cDNA plasmids (pcDNA3-AKR1C3, pCMV-αTubulin [a gift provided by the Dominguez laboratory at the University of Pennsylvania]). PCR product standards were generated for AKR1C3 with the following: forward primer 5′- GAA GTA AAG CTT TGG AGG TC-3′ and reverse primer 5′- GTC AAC ATA GTC CAA TTG AGC-3′. PCR product standards were generated for αTubulin with the following: forward primer 5′- AGA CTT GGA ACC CAC AGT CA-3′ and reverse primer 5′- AAG CCC TGA AGA CCG-3′. All samples were analyzed in triplicate.

The conditions for the endpoint PCR for 11β-hydroxysteroid dehydrogenase type 1 (HSD11B1) and 11β-hydroxysteroid dehydrogenase type 2 (HSD11B2) were 95 °C for 15 minutes followed by 40 cycles of 94 °C for 30 seconds and 60 °C for 30 seconds. PCR products were generated for HSD11B1 with the following: forward primer 5′- CTC TAC AGA AGG TGG TAT CC-3′ and reverse primer 5′- AAT GAG CAT GTC TAG TCC TC-3′. PCR products were generated for HSD11B2 with the following: forward primer 5′- TGC TTC AAG ACA GAG TCA GTG-3′ and reverse primer 5′- GGC ATC TAC AAC TGG GGT GA-3′. PCR amplicons were visualized on a 2% agarose gel stained with RedSafe.

Western Blot

Cellular extracts were harvested on ice with 100 µL RIPA buffer (Pierce/Thermo) containing 1:100 Protease Inhibitor (Sigma), homogenized by syringe, and spun for 10 minutes at 15 000 rpm at 4 °C. Protein concentrations were determined with a BCA Assay Kit (Pierce/Thermo). Protein extracts (20 µg) were loaded on 10% SDS polyacrylamide gels. Gels were transferred to nitrocellulose membranes and blocked with 5% blocking buffer (Bio-Rad, Hercules, CA) in Tris-buffered saline containing Tween 20 (TBST). Blots were then incubated with murine anti-human AKR1C3 (35) (AB_476751; A6229; Sigma) and 1:3000 murine anti-human αTubulin (AB_521686; NB100–690; Novus Biologicals, Littleton, CO) as primary antibodies overnight at 4 °C. Blots were washed in TBST and developed with 1:2000 murine-IgG horseradish peroxidase conjugate (AB_2687626; sc516102; Santa Cruz; Dallas, TX) in 5% blocking buffer in TBST for 1 hour at room temperature. Bands were then detected using ECL Western Blotting Substrate (Thermo) and imaged with a ChemiDoc MP Imaging System (Bio-Rad). Band intensity was quantified by Image Lab software (Bio-Rad). AKR1C3 band intensity was normalized to the band intensity of the loading control αTubulin.

Stable Isotope Dilution Liquid Chromatography High Resolution Mass Spectrometry Sample Preparation and Girard P Derivatization

Stable isotope dilution liquid chromatography high-resolution mass spectrometry (SID-LC-HRMS) samples were prepared as follows: fully differentiated adipocytes were treated with 10 nM 4AD, 10 nM 11K-4AD, or 100 nM 11β-OH-4AD in phenol red-free DMEM/F12 media (21041025, Gibco/Thermo). Aliquots of media (1.0 mL) were taken over time at 0, 1, 2, 4, 8, 24, 48, and 72 hours and quenched in 2 mL of ethyl ether containing internal standards [¹³C]-DHEA, [¹³C]-4AD, [¹³C]-T, and [¹³C]-DHT in a 10-mL screw-cap centrifuge tube. Each sample was vortexed for 1 minute and spun down at 4000 rpm at room temperature for 30 minutes. The upper organic layer was extracted to a clean borosilicate test tube. This liquid-liquid extraction was repeated once more by adding another 2 mL of ether and pooling the extracted organic layers. The extracts were dried down in a vacuum centrifuge at 35 °C. The dried extract containing keto-androgens was then derivatized with Girard P to form oximes, which could be detected in the positive-ion mode to increase sensitivity of detection (36-38). Derivatization was performed using 200 µL of 10% acetic acid in methanol containing 20 µL of Girard P (1 mg/mL in water) and reacted at 60 °C for 10 minutes. Samples were then dried under nitrogen and dissolved in 100 µL of 1:1 methanol/water. Samples were filtered with HPLC filter tubes (Corning) at 15 000 rpm for 3 minutes. Analytes are identified by SID-LC-HRMS based on correct retention time, correct high-resolution mass (M + H), and correct fragmentation/daughter ion formation due to loss of Girard P.

SID-LC-HRMS Androgen Quantification

An Ultimate 3000 UPLC system (Thermo Fisher Scientific) was coupled to a Q Exactive-HF mass spectrometer (Thermo Fisher Scientific) for androgen quantitation. Chromatographic separation was achieved on a Cortecs Phenyl 2.1 × 100 mm, 1.6 μm column (186008381) with solvent A as 0.1% formic acid in water (v/v) and solvent B as 0.05% formic acid in methanol/acetonitrile 60/40 (v/v/v) using gradient elution. The method began at 85% A and 15% B with a flow of 0.5 mL/min. A stepwise gradient was applied in which B was increased to 25%, 28%, and 80% in 4-minute increments. The system was then taken to 100% B and washed for 5 minutes. The column was returned to 15% B and re-equilibrated for 5 minutes. Data acquisition was performed in full scan/data dependent tandem mass spectrometry (MS² mode) at 120 000 resolution in the positive-ion mode with a CID of 5.0 eV. Data were analyzed using Xcalibur (Thermo Fischer) to integrate peaks of interest to calculate the ratio of analyte to internal standard. Quantitation was achieved using calibration curves constructed in cell culture medium in which the ratio of analyte to internal standard vs known amount of authentic standard were plotted. T, 11K-T, and 11β-OH-T were quantified using [¹³C]-T as an internal standard. Data are reported as pg androgen formed/ 1 × 10⁵ cells (means ± SEM). Across different days, low, middle, and high calibration points met precision and accuracy standards (< 20% coefficient of variation).

Statistical Analysis for Pharmacological Inhibition With SID-LC-HRMS

All statistical tests on inhibition experiments were computed using PRISM graphpad (version 9.0). Each treatment condition was assessed for normality using the Shapiro-Wilk test and QQ plots when appropriate. Significance was determined using a 1-sample t test against a hypothetical mean of 100% with an alpha of 0.05.

Results

Steady-State Kinetics for the Reduction of 11-Oxygenated Androgens Catalyzed by Recombinant AKR1C3

Steady-state kinetic parameters were determined for the following reactions catalyzed by recombinant AKR1C3: (i) 11K-4AD to 11-Keto-T; (ii) 11K-5AD to 11K-DHT, (iii) 11β-OH-4AD to 11β-OH-T, (iv) 11β-OH-5AD to 11β-OH-DHT, (v) 4AD to T, and (vi) 5AD to DHT using UV reverse phase HPLC discontinuous enzyme assays to measure direct product formation. Velocity vs substrate curves were generated to measure the K_m and k_cat for each reaction. Androgens of the 5α-series (5AD and DHT analogs) required an additional DNPH derivatization to make UV-visible derivatives. Reduction of 11K-5AD yielded two 11K-DHT-DNPH derivatives (Fig. S1) (39). When the catalytic efficiencies (k_cat/ K_m) for the reduction of 11K-4AD and 11K-5AD were compared with those for the reduction of 4AD and 5AD, the 11-oxygenated androgens gave values that were 4-fold higher compared with the respective classical androgens (Fig. 2 and Table 1). This confirmed that the 11-oxygenated androgens were superior substrates for AKR1C3. By contrast AKR1C3 did not convert 11β-OH-4AD to 11β-OH-T or 11β-OH-5AD to 11β-OH-DHT.

Figure 2.

Steady-state kinetics for the conversion of (a) 11K-4AD to 11K-T, (b) 11K-5AD to 11K-DHT, (c) 4AD to T, or (d) 5AD to DHT catalyzed by recombinant AKR1C3 using discontinuous RP-HPLC assays. Mean +/- SEM is given for each time point (n = 2-5).

Table 1.

Steady-state kinetic parameters for the 17-ketoreduction of adrenal androgens catalyzed by recombinant AKR1C3 measured by RP-HPLC

Substrate	K m (µM)	k cat (min−1)	k cat/ Km (min−1 mM-1)
11K-4AD	38.5 ± 10.9	3.57 ± 0.65	92.8
11K-5AD	14.4 ± 1.7	1.18 ± 0.09	81.9
4AD	10.5 ± 2.9	0.22 ± 0.02	20.9
5AD	7.1 ± 1.6	0.14 ± 0.02	20

Abbreviations: 11K-4AD, 11-keto-Δ4-androstene-3,17-dione; 11K-5AD, 11-keto-5α-androstane-3,17-dione; 4AD, Δ4-androstene-3,17-dione; 5AD, 5α-androstane-3,17-dione.

Induction of AKR1C3 in SGBS Adipocytes

In the SGBS cellular model, the induction of AKR1C3 was recapitulated, as previously reported, during the differentiation of these preadipocytes into mature adipocytes at the RNA and protein levels. SGBS cells were stained with Oil Red O to confirm the production of lipid droplets in the mature adipocytes. AKR1C3 is not expressed in preadipocytes, but there is insulin-dependent AKR1C3 induction during differentiation that is time dependent. The largest increase in AKR1C3 mRNA occurs before day 7 of differentiation with a > 600-fold induction relative to the preadipocytes by RT-qPCR, which is followed by a corresponding increase in protein expression first seen by day 9 and robustly induced by day 14 as measured by immunoblot analysis (Fig. 3).

Figure 3.

AKR1C3 is induced by 20 nM insulin in SGBS cells. Cells were stained with Oil Red O (a) at day 0 as preadipocytes, (b) at day 4 after the quick differentiation step, and (c) at day 14 as mature adipocytes for lipid droplets and viewed under 4× color brightfield. (d) Day 7 adipocytes were viewed under 20× phase contrast to visualize individual lipid droplets. AKR1C3 expression was measured by (e) RT-qPCR and (f) immunoblot during differentiation using 20 ng of recombinant AKR1C3 as a standard. Mean ± SEM is given for each RT-qPCR time point with at least an n = 3 and significant differences with a P value < 0.05 using a 2-sample t test are indicated with an (*).

Stable Isotope Dilution LC-HRMS Assay for 11-Oxygenated Androgens

We utilized a SID-LC-HRMS assay for the separation, detection and quantitation of a 12-analyte steroid panel consisting of classical androgens (4AD, T, 5AD, DHT), keto-androgens (11K-4AD, 11K-T, 11K-5AD, 11K-DHT), and 11β-OH androgens (11β-OH-4AD, 11β-OH-T, 11β-OH-5AD, and 11β-OH-DHT) using Girard P derivatization, in which [¹³C]-T, [¹³C]-DHT, and [¹³C]-4AD were used as internal standards. Each of the analytes were well separated and gave the correct MH⁺ and ion transition for the loss of Girard P group. Calibration curves were linear up to 1 ng and representative calibration curves for T, 11K-T, and 11β-OH-T, where [¹³C]-T was used as an internal standard, are provided in the supplemental material (Fig. S2) (40).

Metabolism of Adrenal Androgens in Mature SGBS Adipocytes by SID-LC-HRMS

We used fully differentiated SGBS adipocytes obtained at day 15 to determine whether they would metabolize adrenal androgens when treated with concentrations that may be reached in the serum of PCOS patients, such as 10 nM 4AD, 10 nM 11K-4AD, or 100 nM 11β-OH-4AD. It is noteworthy that 11β-OH-4AD can be as high as 32 nM and may be even higher following ACTH stimulation of the adrenals (18). The media was sampled and extracted over 3 days to quantify androgen conversion by the SGBS cells. When treated with 10 nM 4AD, SGBS adipocytes produced T that reached a concentration of 2 nM after 72 hours (Fig. 4A-4C). When treated with 10 nM 11K-4AD, SGBS adipocytes produced 11K-T at a concentration of 2 nM within 24 hours, showing more rapid conversion than 4AD to T. After 1 day, the maximum concentration of 11K-T decreased over the remaining time course. Concurrent with the decrease of 11K-T, there was an increase in 11β-OH-T, which reached 4 nM after 72 hours. These data support the finding that 11K-4AD is a superior substrate compared with 4AD for AKR1C3; however, the presence of the 11-ketone group in 11K-T makes it available for further downstream metabolism to 11β-OH-T. Ratios of 11K-T to T, 11β-OH-T to T, and (11K-T + 11β-OH-T) to T confirm that 11K-T and 11β-OH-T are produced in amounts that far exceed T early into the time course (Table 2).

Figure 4.

Metabolism of adrenal androgens in differentiated SGBS adipocytes measured by SID-LC-HRMS. Cells were treated with: (a) 10 nM 4AD and T formation was quantified, (b) 10 nM 11K-4AD and 11K-T formation was quantified, or (c) 10 nM 11K-4AD and 11β-OH-T formation was quantified. Alternatively, cells were treated with (d) 100 nM 11β-OH-4AD and 11K-T formation was quantified or (e) 100 nM 11β-OH-4AD and 11β-OH-T formation was quantified. Mean ± SEM is given for each time point with at least an n = 3.

Table 2.

Ratios of androgen metabolites formed in SGBS adipocytes over time

Time (Hours)	11K-T/ T	11β-OH-T/ T	(11K-T + 11β-OH-T)/ T
4	42.0	10.2	52.3
8	9.5	3.5	13.0
24	2.6	3.2	5.8
48	1.1	3.2	4.3
72	0.5	2.2	2.7

Abbreviations: 11β-OH-T, 11β-hydroxy-testosterone; 11K-T, 11-ketotestosterone; T, testosterone.

When adipocytes were treated with 100 nM 11β-OH-4AD, the adrenal precursor of 11K-4AD, linear formation of 11K-T was observed, reaching 1.5 nM after 72 hours. However, the amount of 11K-T formed is comparable to that observed with 10 nM 11K-4AD (10-fold less precursor) suggesting that the conversion of 11β-OH-4AD to 11K-4AD is limiting 11K-T formation (Fig. 4D and 4E). Overlays of the progress curves emphasize that over time the dominant metabolite is 11β-OH-T, but this is not formed by the reduction of 11β-OH-4AD to 11β-OH-T by AKR1C3 (Fig. S3) (41).

Formation of T, 11K-T, and 11β-OH-T Are Blocked by AKR1C3 Inhibitors

We repeated these experiments with 3 different classes of AKR1C3 competitive inhibitors (BMT4-159, ASP9521, and indomethacin) (26, 42-44) and found that the conversion of 4AD to T and 11K-4AD to 11K-T were robustly inhibited by 30 µM of each inhibitor, suggesting that these reactions were AKR1C3 dependent (Fig. 5A-5C). The formation of 11β-OH-T was blocked by AKR1C3 inhibitors when using 11K-4AD as a substrate and reflects the requirement that 11K-4AD is first converted to 11K-T by AKR1C3, which is then converted to 11β-OH-T. Similar trends were observed when cells were treated with 11β-OH-4AD followed by the AKR1C3 inhibitors. Thus, both 11K-T and 11β-OH-T are produced in an AKR1C3-dependent manner (Fig. 5D and 5E).

Figure 5.

Metabolism of adrenal androgens in differentiated SGBS adipocytes in the presence of AKR1C3 competitive inhibitors (BMT4-159, ASP9521, and indomethacin) measured by SID-LC-HRMS. Cells were treated with: (a) 10 nM 4AD and T formation was quantified, (b) 10 nM 11K-4AD and 11K-T formation was quantified, (c) 10 nM 11K-4AD and 11β-OH-T formation was quantified, (d) 100 nM 11β-OH-4AD and 11K-T formation was quantified, or (e) 100 nM 11β-OH-4AD and 11β-OH-T formation was quantified in the presence or absence of 30 µM AKR1C3 inhibitors after 24 hours. Mean ± SEM is given for each time point with at least an n = 3. All product inhibitions were significant with a P value < 0.05.

Formation of 11β-OH-T Is Blocked by HSD11B1 Inhibition

The delayed but robust formation of 11β-OH-T when using either 11β-OH-4AD or 11K-4AD as a precursor, compared to the more rapid formation of 11K-T, implied that HSD11B1, which acts as a reductase in human omental fat (45), might be responsible for the formation of 11β-OH-T. We repeated our assays in SGBS adipocytes with 10 µM BVT14225, a HSD11B1-specific inhibitor (46) (Fig. 6). We find that inhibitor treatment increased 11K-T formation from both 11β-OH-4AD and 11K-4AD with corresponding inhibition of 11β-OH-T formation. To confirm expression of HSD11B1 and HSD11B2, we performed endpoint PCR and found that transcripts for both enzymes are expressed in differentiating SGBS adipocytes (Fig. S4) (47).

Figure 6.

Metabolism of adrenal androgens in differentiated SGBS adipocytes in the presence of a HSD11B1 inhibitor (BVT14225) measured by SID-LC-HRMS. Cells were treated with 10 nM 11K-4AD and either (a) 11K-T formation or (b) 11β-OH-T formation was quantified in the presence or absence of 10 µM BVT14225 after 24 hours. Alternatively, cells were treated with 100 nM 11β-OH-4AD and either (c) 11K-T formation or (d) 11β-OH-T formation was quantified in the presence and absence of 10 µM BVT14225 after 24 hours. Mean ± SEM is given for each time point with at least an n = 3. 11β-OH-T inhibition in b and d were significant with a P value < 0.05.

Discussion

Ovarian androgen production can be a dominant source of hyperandrogenism in women with PCOS and can be suppressed with oral contraceptive use. However, adrenal 11β-OH-4AD and 11K-4AD may contribute to AE in PCOS. Once 4AD is converted by 11β-hydroxylase to produce 11β-OH-4AD, it may act as a precursor of 11K-T and 11K-DHT, which are potent ligands for the AR. In the periphery, AKR1C3 has 17-ketosteroid reductase activity that may contribute to the formation of 11K-T and 11K-DHT. In PCOS adipocytes, AKR1C3 is induced by insulin, and insulin resistance is a characteristic of PCOS. Therefore, insulin resistance may contribute to potent androgen production in adipocytes by inducing AKR1C3. However, no metabolic studies of 11β-OH-4AD and 11K-4AD in adipocyte models have been performed until now.

Using recombinant AKR1C3, we find that 11K-4AD and 11K-5AD are superior substrates for the enzyme compared to 4AD and 5AD, which is consistent with previous findings (12). However, we found that 11β-OH-4AD, the major circulating 11-oxygenated androgen, was not a substrate for AKR1C3, indicating a distinct preference for 11-keto-androgens by the enzyme.

We utilized SGBS adipocytes as a model for PCOS patient adipocytes to study the metabolism of adrenal androgens. This is a convenient model, representative of human white adipose tissue, that can be maintained in passage and differentiated for up to 50 generations independent of FBS. This advancement of a human cellular adipocyte model is superior to previous murine adipocyte models that require FBS during differentiation. Use of patient samples to conduct these intricate metabolism studies would require a large sample size to reduce the variability among the various phenotypes of PCOS, which SGBS cells help to avoid. SGBS cells also recapitulate the characteristics of PCOS adipocytes, including induction of AKR1C3 and lipogenesis by insulin, making SGBS cells ideal for this study (16). We find that there was no baseline production of androgens in the preadipocytes or mature adipocytes contrary to a recent publication using SGBS cells (48).

We used a SID-LC-HRMS assay to monitor a panel of 12 androgens to conduct metabolic studies in SGBS differentiated adipocytes in which AKR1C3 was robustly induced by insulin. Using concentrations of 4AD and 11K-4AD that may be reached in PCOS patients (22-25), we found that these could be converted to T and 11K-T, respectively. 11K-T was formed in much larger amounts than T, as determined by ratios of 11K-T to T. This finding is supported by the higher turnover number (k_cat) with recombinant enzyme for the conversion of 11K-4AD to 11K-T compared to the conversion of 4AD to T. Importantly, the K_m of these reactions are in the micromolar range, which is much higher than the physiologic concentrations of steroids. This indicates that AKR1C3 is operating on the linear portion of the velocity vs substrate curves for both 4-AD and 11K-4AD, supporting the stronger preference for 11K-4AD. We also found that in SGBS adipocytes 11β-OH-4AD and 11K-4AD both produced the same amount of 11K-T, even though the concentration of 11β-OH-4AD was 10-fold higher than that of 11K-4AD. This suggests that the conversion of 11β-OH-4AD to 11K-4AD in the SGBS cells was the rate-determining step, most likely mediated by HSD11B2 (Fig. 7). Despite the robust formation of 11K-T from either 11K-4AD or 11β-OH-4AD, 11K-T levels were eventually dwarfed by the appearance of 11β-OH-T, which has no affinity for AR. During the metabolism of either 11K-4AD or 11β-OH-4AD, we did not detect any 11K-5AD or 11K-DHT, suggesting that the level of SRD5A1/2 expression is not sufficient for their formation. Alternatively, 11K-DHT could be formed but may be further rapidly metabolized by AKR1C1/AKR1C2 to the corresponding 3α- and 3β-diols.

Figure 7.

Metabolic scheme of the metabolism of 11β-OH-4AD in a model of PCOS adipocytes. Once formed, 11K-T is preferentially converted to 11β-OH-T, while the formation of 11K-DHT is not observed. Mean ± SEM is given for each inhibitor (n = 3).

We find that the formation of T and 11K-T are almost entirely dependent on AKR1C3, since their formation is robustly blocked by a panel of AKR1C3 inhibitors. We also find that the formation of 11β-OH-T is blocked by a potent HSD11B1 inhibitor. HSD11B1 is classically known for its ability to convert cortisone to cortisol, where cortisol is a potent agonist for the glucocorticoid receptor (49). Our results identify an unappreciated role for HSD11B1, in which by diverting 11K-T to 11β-OH-T, it can protect the AR from androgen excess. Thus, HSD11B1 and HSD11B2 could act as molecular switches involved in the pre-receptor regulation of the AR in adipocytes whereby HSD11B1 inactivates 11K-T and HSD11B2 forms 11K-T.

Our data indicate the importance of the relative expression of AKR1C3, HSD11B1, and HSD11B2 in PCOS adipocytes. Heterogeneous enzyme expression in different stages of adipocyte differentiation could dictate different rates of androgen biosynthesis and metabolism in the adipose tissue. HSD11B1 is not expressed in the preadipocyte but is highly expressed at the beginning of differentiation and declines as differentiation is completed (48), whereas AKR1C3 expression is most robust in mature adipocytes. We confirmed the expression of HSD11B1 and HSD11B2 in our SGBS adipocytes at both days 7 and 14 of differentiation. Whether a patient’s adipose tissue expresses a higher AKR1C3 to HSD11B1 ratio may determine to which extent potent 11-oxygenated androgens contribute to the hyperandrogenism of a patient.

There are a number of weaknesses in our study. First, even though we see inactivation of 11K-T in our cellular model, this may not occur in PCOS patients if 11β-OH-4AD and 11K-4AD are continuously resupplied by the adrenals. Second, our cellular model is dependent on a fixed amount of insulin. However, as PCOS patients experience exacerbated insulin resistance this would increase insulin secretion and dysregulate insulin signaling further. In our model, we only modeled physiological insulin stimulation, but this does not model how adipocytes that are insulin resistant alter adipocyte androgen metabolism. The concentrations of steroids used in our assays were 10 nM 11K-4AD and 100 nM 11β-OH-4AD. We find that values for 11β-OH-4AD range from 1.3 nM to 32 nM in PCOS women (22-25), while values for 11K-4AD range from 0.65 nM to 13.4 nM in the same studies. In each of these studies, the concentration of 11β-OH-4AD was at least 2.6-fold higher than 11K-4AD, supporting the view that 11β-OH-4AD will be the major precursor steroid for 11K-T. Studies by the Rainey group show that levels of 11β-OH-4AD and 11K-4AD increase 3- to 5-fold upon ACTH stimulation, suggesting that their levels will oscillate due to circadian rhythm (18). Based on these assessments, we believe that our use of 10 nM 11K-4AD and 100 nM 11β-OH4-AD are not as supraphysiologic as they appear and recapitulate the differences in the ratios of these 2 steroids seen in the circulation. By using 11β-OH-4AD at this concentration also allowed us to detect 11K-T since this was robustly converted to 11β-OH-T. It is also noted that we utilized [¹³C]-T as the internal standard, as ion suppression may vary by analyte, our quantitation of these analytes may not be as reliable. However, our mass balances indicate that we can account for the bulk of the steroid added as substrate.

11K-T from adipocytes could significantly contribute to the hyperandrogenic profile in serum of PCOS patients due to its robust formation by AKR1C3 even though it may potentially be inactivated by HSD11B1. This would suggest that by only measuring T as a diagnostic indicator for AE in PCOS women, the androgen burden in PCOS women may be underestimated. If serum 11K-T was utilized as a biomarker of hyperandrogenism in PCOS in addition to T, this could result in a reclassification of PCOS patients who do not classify as having hyperandrogenism under the current guidelines.

Our study also highlights the importance of sites other than the ovary, such as adipose tissue, that may be responsible for the hyperandrogenism in PCOS women. This work highlights how serum concentrations only tell a piece of the story and quantifying androgen production directly from the adipocytes emphasizes how both AKR1C3 and the 11-oxygenated androgens may be key players in contributing to the AE of PCOS. Our work identifies the possibility of using AKR1C3-specific inhibitors to block androgen production in adipocytes to reduce the hyperandrogenic burden in PCOS women. Oral contraceptives could be supplemented with AKR1C3 inhibitors to robustly reduce the androgen burden in these patients if the adrenal/adipocyte pathway promotes hyperandrogenism in PCOS women.

Glossary

Abbreviations

11β-OH-4AD

11β-hydroxy-Δ4-androstene-3,17-dione

11β-OH-T

11β-hydroxy-testosterone

11K-4AD

11-keto-Δ4-androstene-3,17-dione

11K-5AD

11-keto-5α-androstene-3,17-dione

11K-DHT

11-keto-5α-dihydrotestosterone

4AD

Δ4-androstene-3,17-dione

5AD

5α-androstane-3,17-dione

AE

hyperandrogenism

AKR1C3

aldo-keto reductase family 1 member C3

AR

androgen receptor

DHEA

dehydroepiandrosterone

DHT

dihydrotestosterone

DMEM/F12

Dulbecco’s Modified Eagle Medium/Nutrient Mixture F12

DNPH

dinitrophenylhydrazine

FBS

fetal bovine serum

HPLC

high-performance liquid chromatography

HRMS

high-resolution mass spectrometry

HSD11B1

11β-hydroxysteroid dehydrogenase type 1

PCOS

polycystic ovary syndrome

PCR

polymerase chain reaction

RT-qPCR

real-time quantitative polymerase chain reaction

SGBS

Simpson-Golabi-Behmel syndrome

SID-LC-HRMS

stable isotope dilution liquid chromatography high resolution mass spectrometry

T

testosterone

TBST

Tris-buffered saline with Tween-20

UV

ultraviolet

Financial Support

This work was supported by grants from the National Institutes of Health from the National Institute of Environmental Health Sciences as follows: P30-ES013508 (awarded to T.M.P.) and R.D.P. was supported by T32-ES019851.

Disclosures

T.M.P. is founder of Penzymes, LLC; he is a consultant for the Research Institute of Fragrance Materials.

Data Accessibility Statement

Original data generated and analyzed during this study are included in this published article or in the data repositories listed in References.