The standards of E1 (1,3,5 (10)-estratrien-3-ol-17-one) and E2 (1,3,5-estratriene-3,17β-diol) were from Steraloids (Newport, RI, United States), the standards of E1-S (1,3,5 (10)-estratrien-17-one 3-sulfate), E2-S (1,3,5 (10)-estratriene-3,17β-diol sulfate), E2-d₂ (1,3,5 (10)-estratriene-2,4-d2-3,17β-diol) were from Sigma Aldrich (St. Louis, MO, United States). Methanol and ammonium fluoride (NH₄F) were from Honeywell International Inc. (Charlotte, NC, United States), and ethanol was from Merck (Darmstadt, Germany). Ultrapure water with a resistivity of 18.2 MΩ × cm was obtained from a Milli-Q water purification system (Merck Millipore, Darmstadt, Germany). The STS inhibitor STX64 was a kind gift from Dr. Barry Potter (Oxford University, Oxford, United Kingdom).

The RL95-2 cell line (RRID: CVCL_0505) was originally established from a grade 2 moderately differentiated adenosquamous carcinoma of the endometrium from a 65-year-old patient (Way et al., 1983), and it was purchased from American Type Culture Collection (ATCC; CRL-1671; lot 62130010) on October 18, 2017, as passage 125. The RL95-2 cells were cultured in Dulbecco’s modified Eagles’s medium/F12 (D6421), with 10% fetal bovine serum (FBS; F9665), 2.5 mM L-glutamine (G7153), and 5 μg/ml insulin (I9278) (all from Sigma–Aldrich GmbH). The RL95-2 cells were used as passages +8 to +14. Authentication by short tandem repeat (STR) profiling was performed by ATCC.

The KLE cell line (RRID: CVCL_1329) was originally established from a poorly differentiated endometrial carcinoma from a 68-year-old patient (Richardson et al., 1984), and it was purchased from ATCC (CRL-162; lot 70001143) on October 18, 2017, as passage +12. The KLE cells were cultured in Dulbecco’s modified Eagles’s medium/F12 (D6421) supplemented with 10% FBS (F9665) and 2.5 mM L-glutamine (G7153) (all from Sigma–Aldrich GmbH). The KLE cells were used as passages +21 to +27. Authentication by STR profiling was performed by ATCC.

The HIEEC cell line was obtained from Michael A. Fortier (Laval University, Quebec, Canada) on April 4, 2014, as passage 14. The HIEEC cells were grown in RPMI-1640 Medium supplemented with 2 mM L-glutamine and 10% fetal bovine serum (FBS) (all from Sigma-Aldrich, St. Louis, MI, United States). The HIEEC cells were used as passage +7. Cells in passage +8 were authenticated by STR profiling performed by ATCC on March 8, 2018.

These cell lines were all negative for mycoplasma infection, according to MycoAlert mycoplasma detection kits (Lonza, Basel, Switzerland).

The RL95-2 and KLE cells were cultured four different times independently (n = 4) as two replicates. The total RNA from RL95-2 and KLE cells was isolated and purified using RNA isolation kits (Nucleospin; Macherey-Nagel GmbH & Co. KG, Düren, Germany), according to the manufacturer instructions. The RNA quality was determined using a bioanalyzer (2100: Agilent) and RNA nanokits (RNA 600; Agilent Technologies Inc, Santa Clara, CA, United States). The mean RNA integrity number was 9.00 ± 0.53, which demonstrated that the RNA was of good quality. Samples of the total RNA (4 µg) were reverse transcribed into cDNA (in 40 µg) using cDNA synthesis kits (SuperScript VILO; Invitrogen, Thermo Fisher Scientific, Carlsbad, CA, United States), according to manufacturer instructions. The cDNA samples were stored at −20°C.

The expression of the genes that encode the enzymes involved in estradiol biosynthesis and oxidative metabolism was examined using quantitative PCR (qPCR) with exon-spanning hydrolysis probes (dye labeled, FAM), as commercially available as ‘Assay on Demand’ (Applied Biosystems, Foster City, CA, United States) (Table 1), or with primers and probes for AKR1C3 that were designed by our group previously (Rižner et al., 2006) (Table 2), using TaqMan Fast Advanced Master Mix and the universal thermocycling parameters recommended by Applied Biosystems (1 cycle of 20 s at 50°C; 1 cycle of 20 s at 95°C; 40 cycles of 1 s at 95°C; 20 s at 60°C). The expression of the genes that encode transporters was examined using SYBR green I master (Roche, Basel, Switzerland) and primers that were designed in our laboratory (Table 3), using the following program: 1 cycle of 5 min at 95°C; 40 cycles of 10 s at 95°C; 10 s at 60°C; and 21 s at 72°C. Quantification was accomplished using a real-time PCR system (ViiA 7; Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, United States). All of the cDNA samples were run on the PCR machine in triplicates, using 0.25 μl cDNA, and the reactions were performed in 384-well plates (MicroAmp Optical; Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, United States), in a reaction volume of 5.0 μl. The PCR amplification efficiency was determined from the slope of the log-linear portion of the calibration curve for each gene investigated, and this was allowed for in the further calculations.

For gene expression analysis in RL95-2 and KLE cells, the normalization factor for each sample was calculated based on the geometric mean of the three most stably expressed reference genes (POLR2A, HPRT1, RPLP0). The gene expression for each sample was calculated from the crossing-point value (Cq) as E^−Cq, divided by the normalization factor, and multiplied by 10¹³. The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines were considered in the performance and interpretation of the qPCR reactions (Bustin et al., 2009).

For comparison of RL95-2 and KLE cell lines with HIEEC, expression of HPRT1 was used as a normalization control. Inter-plate variability in comparison of these two cell lines with HIEEC cells was minimized by the use of relative quatification method and by considering as important only the genes with 10-fold or higher significant differences in expression.

Cell lysates were prepared using RIPA Lysis buffer (EMD Millipore Corporation, Temecula, CA, United States) according to the manufacturer instructions. Total protein concentrations were determined using Bradford reagent (Carl Roth GmbH + Co. KG, Karlsruhe, Germany), bovine serum albumin (BSA) as standard, and BioTek (Winooski, VT, United States) PowerWave XS Microplate reader.

Samples of 50 μg protein were separated using SDS PAGE in 10% Tris-glycine gels, and then transferred to poly (vinylidene fluoride) membranes (Millipore, Billerica, MA, United States). For STS detection, the membranes were blocked in 5% non-fat milk overnight and 5% bovine serum albumin in TTBS buffer for 2 h. The membranes were incubated with primary anti-STS antibodies (1:5,000; in TTBS, 5% bovine serum albumin; 2 h at 4°C), which were kindly provided by Dr. Gerhard Schuler (Faculty of Veterinary Medicine, Justus-Liebig-University, Giessen, Germany). The membranes were then incubated for 2 h at 4°C with horseradish-peroxidase-conjugated secondary goat anti-rabbit antibodies (111-035-045, 1:5,000; in TTBS with 2.5% bovine serum albumin; Jackson ImmunoResearch Laboratories Inc, West Grove, PA, United States). For GAPDH detection, the membranes were blocked in 5% non-fat milk at room temperature, and incubated for 1 h in the primary anti-GAPDH antibody (G8795; lot number: 045M4799V; 1:5,000; in TTBS with 1% non-fat milk at room temperature; Sigma-Aldrich, St. Louis, MI, United States). The membranes were then incubated for 1 h at room temperature with the horseradish-peroxidase-conjugated secondary goat anti-mouse antibodies (115-035-062; 1:5,000; in TTBS with 1% nonfat milk; Jackson ImmunoResearch Laboratories Inc, West Grove, PA, United States). SuperSignal West Pico Chemiluminescent Substrate (Thermo Fischer Scientific, Waltham, MA, United States) was used for chemiluminescent detection, with a CCD camera (LAS-4000; Fujifilm, Tokyo, Japan). Differential expression of STS was determined after normalization to GAPDH, using the ImageJ program.

The RL95-2 and KLE cell lines were seeded into six-well plates (92106; TPP, Sigma-Aldrich Chemie GmbH, Deisenhofen, Germany) at 3.5 × 10⁶ cells/well and 3.0 × 10⁵ cells/well, respectively. The next day, when they had reached 70% confluency, the cells were washed twice with Dulbecco’s phosphate-buffered saline (DPBS) and treated with different concentrations of E1, E1-S, DHEA or DHEA-S (10, 100, 500, 1,000 nM), dissolved in dimethylsulfoxide (DMSO) and medium without phenol red and FBS. The final concentration of DMSO was 0.05%. The cells were treated for 48 h, and then the medium was removed and stored at −80°C in glass tubes (6 ml tubes; 986492; Wheaton, VWR, Pennsylvania, United States) until the liquid chromatography–high-resolution mass spectrometry (LC-HRMS) analysis. For normalization purposes, the cells in individual wells were counted using an automated cell counter (TC20; Bio-Rad, CA, United States). Two independent experiments were performed.

A selective and sensitive LC-HRMS assay was used for quantification of steroid precursors and estrogen metabolites (i.e., AD, DHEA, DHEA-S, E1, E1-S, E2, E2-S, E2-glucuronide, estriol, testosterone). This system was validated according to the Q2 (R1) International Conference on Harmonisation guidelines, as described previously (Poschner et al., 2017).

The HPLC system (UltiMate 3000 RSLC-series; Thermo Fisher Scientific, Inc, Waltham, MA, United States) was run with a C18 column (Phenomenex Luna 3 µm C18 (2) 100 Å; 250 × 4.6 mm ID; Phenomenex, Inc, Torrance, CA, United States), and with a C18 guard column (Hypersil BDS; 5 μm, 10 × 4.6 mm ID; Thermo Fisher Scientific, Inc). The column temperature was maintained at 43°C, with 100 µl sample injected. The gradient elution used aqueous ammonium acetate buffer (10 mM, pH 5.0) as solvent A, and acetonitrile as solvent B. The gradient was as follows: 0.0–19.0 min, 25.0–56.3% B; 19.0–19.5 min, 56.3–90.0% B; 19.5–24.0 min, 90.0% B; 24.5–24.5 min, 90.0–25.0% B; 24.5–30.5 min, 25% B. The flow rate was 1.0 ml/min. The HPLC was coupled to a mass spectrometer (maXis HD ESI-Qq-TOF; Bruker Corporation, Bremen, Germany). Full-scan mass spectra were recorded from 150 m/z to 500 m/z.

The RL95-2 (passage, +10 to +15) and KLE (passage, +7 to +15) cells were plated into six-well plates at a cell density of 1.0 × 10⁶ cells/well and 1.0 × 10⁵ cells/well, respectively. After 24 h, the cells were washed with DPBS, and serum-free and phenol-red-free culture medium was added. The cells were then incubated with 2.3, 8.5, and 85 nM E1-S (in ethanol; final ethanol concentration, 0.25%). The effects of the STS inhibitor STX64 on E1-S metabolism was evaluated as follows: 30 min before addition of 2.3 nM E1-S, the cells were incubated with 10 nM STX64 (in anhydrous DMSO; final DMSO concentration, 0.25%). After 8, 24, 48, and 72 h of incubation, the cell culture medium was collected in microcentrifuge tubes (Eppendorf, Germany), and stored at −80°C until further processing. Three independent experiments were carried out, each performed in duplicate.

A deuterated internal standard of E2-d₂ was added to the cell culture medium samples following their thawing at room temperature. The lipophilic fraction containing the analytes of interest was extracted using solid-phase extraction (Strata-X polymer-based columns; Phenomenex, CA, United States). This method involved: column conditioning (1 ml methanol), column equilibration (1 ml water), sample loading, column drying (high vacuum, 10 min), and sample elution (1.5 ml methanol). The solvent was then evaporated off using a vacuum concentrator (Savant SPD 31 DDA-230; Thermo Fisher Scientific, Waltham, MA, United States), and reconstituted in a 100 μl 70% methanol/0.2 mM NH₄F in water. The samples were stored at −20°C until LC-tandem mass spectrometry (MS/MS) analysis.

An LC-MS/MS system used for detection and quantification of E1, E1-S, E2, and E2-S comprised of a Shimadzu Nexera XR system (Shimadzu Corporation, Kyoto Japan) coupled to a triple quadrupole system (Triple Quad 3 500; AB Sciex Deutchland GmbH, Darmstadt, Germany), operating with the Analyst 1.6 software (AB Sciex Deutchland GmbH, Darmstadt, Germany). Chromatographic separation of the estrogens was performed using a C18 column (Kinetex 2.6 μm XB; 100 × 4.6 mm; Phenomenex, Aschaffenburg, Germany) equipped with a guard column and cartridges (Securityguard C18; 4 × 3.0 mm; Phenomenex, Aschaffenburg, Germany). The mobile phases were 0.2 mM NH₄F, 5% methanol in water (A) and 0.2 mM NH₄F in methanol (B). Samples of 25 μl were injected via an autosampler (SIL-20 AC XR; Shimadzu Corporation, Kyoto, Japan). The mobile phase flow rate was 0.5 ml/min, and the gradient elution was as follows: 0.0–3.0 min, 70% A; 3.0–8.0 min, 70–4% A; 8.0–8.01 min, 4–70% A; 8.01–15.0 min, 70% A. The column temperature was maintained at 38°C. The MS/MS analysis was performed in negative ion mode with constant electrospray ionization conditions. The source-dependent parameters were as follows: curtain gas, 50; collision gas, 8; ion spray voltage, −4,500 V; source temperature, 600°C; ion source gas 1, 40; ion source gas 2, 80. All transitions were recorded using the scheduled multiple reaction monitoring (MRM) algorithm. The target scan time was set to 1 s, with an MRM detection window of 120 s. The resolution for the first and third quadrupole (Q1, Q3) was set as UNIT, with the pause between the mass ranges set at 5 ms.

The concentration of each steroid was calculated using the internal standard approach. The standard calibration curves were constructed from 12 calibration concentrations prepared in 70% methanol/0.2 mM NH₄F, to cover the range from 0.01 to 100 ng/ml. An internal standard of E2-d₂ was added to each sample at a final concentration of 1 ng/ml. The retention times and monitoring transitions for the analytes are given in Supplementary Table S1. The limits of detection and quantification were calculated as 3× and 10× the signal/noise ratio. The limit of detection for E1, E1-S, E2, and E2-S was 1 pg/ml. The limit of quantification for E1 and E1-S was 5 pg/ml, and for E2 and E2-S, 10 pg/ml.

For analysis of E1-S transport, RL95-2 and KLE cells were seeded into six-well plates at 1.0 × 10⁶ cells/well and 1.2 × 10⁵ cells/well, respectively, in triplicates. Here, RL95-2 cells were between passages +12 and +14, and KLE cells between passages +28 and +32. When the cells reached 60% confluency, full growth medium was replaced with medium without FBS, for 24 h. The cells were then washed twice with warm DPBS (S5652; Sigma-Aldrich, St. Louis, MO, United States) and incubated for 30 min at 37°C in 1.9 ml/well transport buffer (125 mM NaCl, 4.8 mM KCl, 1.2 mM CaCl₂, 1.2 mM KH₂PO₄, 12 mM MgSO₄, 25 mM MES, 5.6 mM glucose, pH 5.5) with or without 10 µM inhibitor cyclosporine A (CsA, SML1018, Sigma-Aldrich, St. Louis, MI, United States) or 10 µM bromosulphophthalein (BSP, HY-D0217, MedChemExpress, Monmouth Junction, NJ, United States). After 30 min, steroid transport was initiated by addition of 100 µl transport buffer containing [³H]E1-S (NET203250UC; Perkin Elmer Inc, MA, United States) to a final concentration of 16 nM, and the cells were incubated for 2, 5, 15, and 30 min at 37°C. The medium was collected and the uptake was stopped with 2 ml ice-cold DPBS. After five washes with DPBS, the cells were lyzed by addition of 300 µ/well 1% Triton X-100, mixed for 30 min at 200 rpm at 4°C, and then frozen at −80°C. After thawing, the lyzed cell samples were collected and centrifuged at 11,000 × g for 15 min at 4°C. Finally, 250 µl of the cell lysates was mixed with 1.5 ml scintillation fluid (Quickszint Flow 302; Zinnsser Analytic, Frankfurt, Germany) and the radioactivity was determined in a scintillation counter (MicroBeta TriLux 1450; PerkinElmer, CT, United States). The [³H]E1-S concentrations in the individual samples were then back-calculated from the disintegrations per minute (dpm) and the specific isotope activity (49.19 Ci/mmol) with the conversion of 1 Ci = 2.22 × 10¹² dpm. The data were normalized to total protein concentrations, as determined using the protein assay kits (BCA protein assay; Pierce, Thermo Scientific), according to manufacturer instructions. The E1-S transport was studied as two to three independent experiments.

For evaluation of gene expression, RL95-2 and KLE cells were cultured independently on four different occasions; three times in duplicate, and once singly. The expression of the genes of interest and the reference genes was determined in each of these seven samples of three technical triplicates. The means for each independent experiment (n = 4) were considered in the statistical analysis. Western blotting was performed on protein samples from three biological replicates of each cell line. Cells for LC-HRMS and LC-MS/MS analyses were cultured in two and three independent experiments, respectively (n = 2 and 3), each in duplicate. E1-S uptake was performed independently two to three times, each time as triplicates.

Statistical evaluation was carried out using the GraphPad Prism software for Windows, version 8.0 (San Diego, CA, United States), with Mann–Whitney tests, Kruskal–Wallis tests followed by Dunn’s multiple comparisons tests or ANOVA followed by Tukey’s tests. Differences with p < 0.05 are considered as statistically significant. Unless stated otherwise, all of the data are shown as means ± standard deviation (SD).

After menopause, the ovaries cease to produce estrogens, and local E2 formation relies on the actions of transporters for the uptake of the inactive steroid precursors DHEA-S and E1-S into cells. We evaluated here the expression of 20 genes that encode 15 E1-S uptake and four E1-S efflux transporters (Figure 1, Supplementary Table S2). The great majority of the genes that encoded these E1-S uptake and efflux transporters were expressed in both the RL95-2 and KLE cells.

For the relative expression of the poorly differentiated (KLE) EC cells compared to the moderately differentiated (RL95-2) EC cells (i.e., KLE vs RL95-2 relative expression), differences were seen for several of the E1-S uptake transporters (Figure 1A). These showed higher expression for six E1-S uptake transporters (KLE vs RL95-2): SLCO1A2 (3433.6-fold), SLCO1B3 (2301.5-fold), SLCO1C1 (381.2-fold), SLCO3A1 (19.0-fold), SLC10A6 (5.5-fold), and SLC22A9 (5.0-fold). Two E1-S uptake transporters showed lower expression instead (KLE vs RL95-2): SLC22A11 (59.3-fold) and SLCO4A1 (3.4-fold) (Supplementary Table S2). No differences were seen for the E1-S efflux transporters (Figure 1B).

Comparisons of these RL95-2 and KLE cells with the model of normal endometrium, HIEEC cells, showed several differences in the expression of the E1-S uptake transporters (Figure 2, Supplementary Table S3). For RL95-2 vs HIEEC cells, the relative expression was higher for SLC22A11 (1199.8-fold). Then for KLE vs HIEEC, this was higher for SLCO1B3 (754.9-fold) and SLC22A9 (380.3-fold). For the efflux transporters, for KLE vs HIEEC cells, the relative expression was lower for ABCG2 efflux transporter (11.9-fold).

This gene expression analysis thus initially indicated that both of these EC cell lines can carry out E1-S uptake. The increased relative expression of six of these E1-S uptake transporters in the poorly differentiated KLE EC cells vs the moderately differentiated RL95-2 EC cells, and the higher expression of two of these genes for KLE cells vs HIEEC cells (over RL95-2 cells vs HIEEC cells) imply that more sulfated steroid precursors can enter KLE cells.

After the steroid precursors enter the cells, these can be further transformed through multiple steps into the most potent estrogen, E2. This can occur via the sulfatase and aromatase pathways, where E2 can be formed from E1-S and from DHEA-S (via androstenedione), respectively. The expression of 16 genes of these pathways was examined here for the RL95-2 and KLE model cell lines.

Both RL95-2 and KLE cells showed 10⁵-fold–10⁶-fold higher expression of STS compared to CYP19A1 (Figure 1C). This suggests that these cell lines form estrogens only through the sulfatase pathway. Additionally, these data indicated that RL95-2 and KLE cells differ in their E2 formation. KLE vs RL95-2 relative expression was lower for the sulfatase pathway STS (9.3-fold), as also for the aromatase pathway HSD3B1 (76.2-fold), HSD3B2 (4.5-fold), and AKR1C3 (50.7-fold). For AKR1C3 lower protein levels were seen in KLE cells as compared to RL95-2 (Supplementary Figures S1A, S2).

The genes that encode the HSD17B enzymes were also differentially expressed. The greatest difference for KLE vs RL95-2 cells was lower relative expression for HSD17B2 (1303.7-fold), the enzyme product of which has the highest catalytic efficiency for inactivation of E2 to E1. Similarly, this was lower for another oxidative HSD17B, HSD17B4 (1.8-fold). On the other hand, this was higher for HSD17B14 (7.9-fold), which also catalyzes oxidation of E2 to E1, but with lower catalytic efficiency. Additionally, KLE vs RL95-2 relative expression was lower for HSD17B12 (2.4-fold), which encodes an enzyme for reduction of E1 to E2. The gene that encodes the major reductive HSD17B1 enzyme was not differentially expressed. For KLE cells, these alterations in gene expression of the reductive and oxidative HSD17B enzymes would shift the balance between E1 and E2 towards E2 formation.

Differences in KLE vs RL95-2 relative expression of genes for the sulfotransferase enzymes were also seen. The relative expression here was lower for the gene that encodes DHEA sulfotransferases, SULT2B1 (17.7-fold), while expression of SULT2A1 was only detected in KLE cells. Higher relative expression was instead seen for the gene encoding E1 and E2 sulfotransferase, SULT1E1 (13.6-fold). This differential expression of these genes suggests that KLE cells followed more sulfation of DHEA (higher relative expression of SULT2A1), E1 and E2 (higher relative expression of SULT1E1) compared to RL95-2 cells.

Comparing RL95-2 cells to HIEEC cells, relative expression was lower for CYP19A1 (16.1-fold), and higher for STS (73.1-fold) (Figure 2, Supplementary Table S3). This suggests that for RL95-2 cells, E1 and E2 can be formed via the sulfatase pathway. Several other genes involved in the aromatase and the sulfatase pathways also showed higher relative expression here: HSD17B2 (43,060.2-fold), HSD3B1 (209.0-fold), and SULT2B1 (185.8-fold). Other genes with higher relative expression in RL95-2 cells vs HIEEC cells were HSD17B8 (65.4-fold) and HSD3B2 (19.9-fold). This higher expression of HSD3B1 and HSD3B2 suggests that in RL95-2 cells, DHEA can be metabolized to androstenedione, although DHEA can also be sulfated due to higher expression of SULT2B1. Higher expression of several genes for oxidative HSD17B enzymes suggest that for RL95-2 cells compared to HIEEC cells, the balance between E1 and E2 will be shifted towards E1.

For KLE vs HIEEC cells, relative expression was higher for HSD17B1 (204.6-fold), which encodes reductive HSD17B, SULT1E1 (1020.7-fold), and SULT2A1, which was expressed in KLE cells but not in HIEEC cells. These data indicate that for KLE cells, more E2 can be formed, but conjugation to E2-S would prevail.

The actions of estrogens are mediated through their binding to and activation of nuclear estrogen receptors and membrane-bound receptors. Here, expression was evaluated for ESR1 and ESR2, which encode nuclear receptors ERα and ERβ, respectively, and the three GPER transcript variants, as v2 and v3 plus v4 (Hevir-Kene and Rižner, 2015) (Figure 1D).

Both ESR1 and ESR2 were expressed in RL95-2 and KLE cells, with no difference between these. For KLE vs RL95-2 relative expression, this was lower for GPER v2 (2189.7-fold) and GPER v3 plus v4 (80.9-fold). When RL95-2 and KLE cells were compared to the control cell line, HIEEC cells, there were no significant differences in gene expression of estrogen receptors ESR1, ESR2 and GPER (Figure 2, Supplementary Table S3). The expression of ESR1 surmounted expression of ESR2 in all three cell lines RL95-2, KLE and HIEEC for 3-fold, 3-fold and 67-fold, respectively (Supplementary Table S4).

These data suggest that in moderately differentiated EC (RL95-2 cells), estrogens might act via nuclear and membrane bound receptors, while in poorly differentiated EC (KLE cells), estrogens might preferentially activate the nuclear receptors.

For phase I metabolism inside cells estrogens can be metabolized to 2-, 4- or 16α-hydroxy-estrogens by the actions of different CYP enzymes. Catechol estrogens (2- or 4-OH-E1/E2) can be oxidized to semiquinones or quinones, which are associated with DNA damage. To avoid cell damage, these catechols and quinones are further conjugated and detoxified in phase II metabolism, by the COMT, SULT, UGT, GSTP1, NQO1, and NQO2 enzymes (Rižner, 2013; Hevir-Kene and Rižner, 2015). The expression of 13 genes that encode phase I and II metabolic enzymes was evaluated in these model cell lines (Figure 1E).

The relative expression of genes involved in phase I metabolism for KLE vs RL95-2 cells was higher for CYP1B1 (17.1-fold), which encodes a 4-hydoxylase, but lower for CYP1A1 (10.6-fold) and CYP3A7 (70.1-fold), which encode a 2-hydroxylase and a 16α-hydroxylase, respectively.

For expression of the eight genes of phase II metabolism that were evaluated, KLE vs RL95-2 relative expression was higher for SULT1E1 (13.6-fold), while expression of UGT2B7 was only detected in KLE cells. All of the other genes that encode detoxification-associated enzymes showed lower relative expression for KLE vs RL95-2 cells, as SULT2B1 (17.7-fold), NQO1 (12.7-fold), NQO2 (6.6-fold), COMT (6.3-fold), and GSTP1 (2.2-fold), with SULT1A1 not expressed in KLE cells. This was confirmed at the protein level for soluble COMT in KLE as compared to RL95-2 cells (Supplementary Figures S1B, S3). These changes imply that more 4-hydroxy-estrogens can be formed in KLE cells, along with more genotoxic 3,4-quinones, as a result of lower expression of genes that encode phase II metabolic enzymes.

Comparisons of RL95-2 and KLE cells with HIEEC cells for expression of genes encoding phase I metabolic enzymes revealed some significant differences (Figure 2, Supplementary Table S3). RL95-2 vs HIEEC relative expression was higher for CYP1A1 (127.6-fold), while for KLE vs HIEEC cells, this was higher for CYP1A2 (87.2-fold), CYP1B1 (81.1-fold) and CYP3A5 which was expressed in KLE but not in HIEEC cells. On the other hand, KLE cells showed lower relative expression for CYP3A7 (93.2-fold). These data suggest that more 2-hydroxyestrogens can be formed in RL95-2 vs HIEEC cells, and more 4-hydroxyestrogens in KLE vs HIEEC cells.

Genes of phase II metabolism were also differentially expressed between each of these EC cell lines and HIEEC cells (Figure 2, Supplementary Table S3). RL95-2 cells showed higher relative expression of SULT2B1 (185.8-fold) and NQO2 (14.1-fold), as for KLE cells for SULT1E1 (1020.7-fold) and UGT2B7 which was expressed in KLE cells but not in HIEEC cells.

The changes in the expression of genes that encode uptake and efflux transporters, estradiol biosynthetic enzymes, and phase I and II metabolic enzymes (Figure 1; summarized in Figure 3) were then integrated and taken forward. These indicated that for KLE vs RL95-2 cells, the KLE cells would initially have increased uptake of E1-S (i.e., higher relative expression of SLC10A6, SLC22A9, SLCO1C1, SLCO1A2, SLCO3A1, SLCO1B3) and decreased formation of E1 (i.e., lower relative expression of STS and lower protein levels of STS (Figure 4C). This can then be combined with increased activation of E1 to E2 (i.e., lower relative expression of HSD17B2) and increased formation of sulfated estrogens (i.e., higher relative expression of SULT1E1 and nonsignificantly higher protein levels (Figure 1, Supplementary Figures S1, S4). Indeed, we also experimentally confirmed increased E1-S uptake in KLE vs RL95-2 cells here (Figures 4A,B), which is associated with higher expression of several genes encoding OATP transporters, especially SLCO1A2, SLCO1B3 and SLCO1C1 in KLE, as also confirmed by inhibition of E1-S uptake by bromosulphophthalein, a known inhibitor of OATP1B1, OATP1B3, OATP1A2, OATP2B1 (König et al., 2006). Increased E1-S uptake in KLE cells in the presence of cyclosporine A, which also inhibits E1-S efflux transporters (Dantzic et al., 2018), imply that these efflux ABC transporters importantly regulate E1-S uptake in KLE cells. Furthermore, for KLE cells, the active estrogens formed from E1-S via the sulfatase pathway would be expected to act via nuclear receptors ERα and ERβ (i.e., lower relative expression of GPER, compared with RL95-2 cells).

Additionally, again for KLE vs RL95-2 cells expression data suggest that the KLE cells may show lower levels of 2-hydroxyestrogens and 16α-hydroxyestrogens (i.e., lower relative expression of CYP1A1, CYP3A7), but more 4-hydroxycatechols (i.e., higher relative expression of CYP1B1), which can be oxidized to carcinogenic 3,4-quinones. In KLE cells, lower detoxification levels of catechols (i.e., lower relative expression of COMT and lower S-COMT protein levels, Supplementary Figure S1) and quinones (i.e., lower relative expression of NQO1, NQO2, GSTP1) would also indicate higher levels of DNA adducts, compared to RL95-2 cells.

A similar summarized comparison between RL95-2 and HIEEC cells (Figure 2, Supplementary Figure S6) suggests enhanced hydrolysis of E1-S to E1 (i.e., increased levels of STS), hindered reduction of E1-S to E2 (i.e., higher expression of HSD17B2) and hydroxylation of E1 at the C2 position to form 2-hydroxyestrogens (i.e., higher expression of CYP1A1).

For the comparison between KLE and HIEEC cells (Figure 2, Supplementary Figure S7), this would support increased E1-S uptake (i.e., high relative expression of SLCO1B3, SLC22A9), but also enhanced activation of E1 to E2 and conjugation to E2-S, or formation of 2-hydroxy or 4-hydroxy-estrogens (i.e., higher relative expression of HSD17B1, SULT1E1, CYP1A2, CYP1B1).

To evaluate the formation of active estrogens in the RL95-2 and KLE cells, the metabolism of 10 nM, 100 nM, 500 nM, and 1000 nM DHEA-S, DHEA, E1-S, and E1 was studied following their addition to the cells, with their products separated and quantified by LC-HRMS (Figure 5).

These data showed more efficient DHEA-S metabolism in RL95-2 cells compared to KLE cells. In RL95-2 cells, DHEA-S was metabolized to DHEA and lower levels of androstenedione. In KLE cells, DHEA-S was only metabolized to DHEA at the highest, non-physiological, DHEA-S concentration (i.e., 1,000 nM), which is in line with the lower levels of STS seen previously for KLE cells (Figure 4).

In contrast, metabolism of DHEA was more efficient in KLE cells, where DHEA-S was the major product, followed by androstenedione. Here, DHEA-S production can be explained by expression of SULT2A1. In RL95-2 cells, DHEA was instead mainly metabolized to androstenedione, with very low levels of testosterone. Here, in RL95-2 cells, the lower DHEA-S production is supported by the higher relative expression of HSD3B1, HSD3B2, and AKR1C3. These data confirm that DHEA-S and DHEA cannot serve as precursors for estrogen biosynthesis in these cells, as no estrogens were formed here from these steroid precursors.

Incubations with E1-S revealed that both EC cell lines can take up and metabolize the E1-S steroid precursor, to form active E2. For KLE vs RL95-2 cells, lower levels of E1 and E2 were seen, with E2 formed only at the supraphysiological 1000 nM E1-S. This can be explained by higher relative expression of STS in RL95-2 cells. In contrast, incubation of EC cell lines with E1 showed that in KLE cells, there is greater metabolism of E1 to E2 and E1-S, which would be due to lower expression of HSD17B2 and higher expression of SULT1E1, compared to RL95-2 cells. In RL95-2 cells, formation of very low levels of estriol was seen after addition of E1-S and E1, which is in line with higher relative expression of CYP3A7, and for KLE cells, E2-glucuronides were detected as a result of UGT2B7 activity (Figure 1).

To better simulate the physiological conditions, the metabolism of E1-S in these EC cell lines was also studied at lower E1-S concentrations: 2.3, 8.5, and 85 nM (Figure 6, Supplementary Figure S8). After 72 h for KLE vs RL95-2 cells, metabolism of 2.3 nM E1-S resulted in lower levels of E1, E2 and E2-S; for KLE cells here, E1 and E2 were formed, but E2-S was the major product. The profiles of the metabolites remained the same at the higher, 8.5 and 85 nm E1-S concentrations (Supplementary Figure S8).

When comparing the EC cell lines to HIEEC cells, more E1 and E2 were formed from E1-S for RL95-2 cells, which is in line with higher levels of STS, and more E2-S was formed for KLE cells, which can be explained by higher relative expression of HSD17B1 and SULT1E1 (Figure 2, Supplementary Figures S6, S7). The importance of the sulfatase pathway for formation of E2 from E1-S was additionally confirmed using a specific and potent STS inhibitor, STX64; this shows an IC₅₀ of 8 nM at 20 μM E1-S (Malini et al., 2000), and was used at 10 nM STX64. In RL95-2 cells, this addition of STX64 in combination with 2.3 nM E1-S resulted in an almost complete block of the sulfatase pathway, with significantly lower levels of E1 and E2 formed (Figure 6). In KLE cells, where less E1 was formed from E1-S, addition of STX64 affected the levels of E1 and E2, although a significant difference was only seen for E1-S (Figure 6).