Tibolone Metabolism in Human Liver Is Catalyzed by 3α/3β-Hydroxysteroid Dehydrogenase Activities of the Four Isoforms of the Aldo-Keto Reductase (AKR)1C Subfamily

Tibolone (Livial) [[7α,17α]-17-hydroxy-7-methyl-19-norpregn-5(10)-en-20-yn-3-one] is used in the treatment of climacteric symptoms and the prevention of osteoporosis (Albertazzi et al., 1998; Moore, 1999). Its favorable effects on the breast and endometrium (Colacurci et al., 1998; Valdivia and Ortega, 2000; Volker et al., 2001; Gompel et al., 2002; Blok et al., 2003) suggest that it may be an alternative to a selective estrogen receptor modulator (Smith and O'Malley, 2004) and to traditional estrogen and progestogen combined therapy. Tibolone differs from estrogen and progestogen combined therapy, because it exerts tissue selective effects via site-specific metabolism (Kloosterboer, 2001; Kloosterboer and Ederveen, 2003). The agent has been assigned the acronym selective tissue estrogenic activity regulator to distinguish it from a selective estrogen receptor modulator, since its effects are not solely mediated by the estrogen receptor (Kloosterboer and Ederveen, 2003).

After oral administration, tibolone is quickly metabolized into 3α- and 3β-hydroxytibolone (presumably in the intestine and liver). These phase I metabolites have half-lives of 7 h, and their maximal plasma concentration can exceed that for tibolone by 10-fold (Timmer et al., 2002). Minor amounts of Δ⁴-isotibolone are also found, which are present for only 6 h in the circulation (Fig. 1). The major active circulating metabolite is the 3α-hydroxyderivative (Timmer et al., 2002; Vos et al., 2002), and over 75% tibolone and its metabolites are found to be in the inactive sulfated form. Of a total administered oral dose, 50% is found in the feces as 3α-hydroxy- and 17β-hydroxysulfates, suggesting that transformation by intestinal mucosa and bacteria may contribute to the metabolic profile (Vos et al., 2002).

Tibolone and its metabolites have distinctive properties in target tissues. Tibolone and its Δ⁴-isomer are androgen and progesterone receptor agonists, whereas the two 3-hydroxymetabolites are weak ligands for the estrogen receptors (de Gooyer et al., 2003). In breast cancer cell lines, tibolone and its 3-hydroxyderivatives inhibit estrogen sulfatase and the 17-ketosteroid reductase(s) (that convert inactive estrone into active 17β-estradiol) and activate 17β-hydroxysteroid oxidase(s) (that catalyze the inactivation of 17β-estradiol) (Chetrite et al., 1999; Pasqualini and Chetrite, 1999; Purohit et al., 2001; Gompel et al., 2002; van de Ven et al., 2002). These actions selectively deprive the estrogen receptors in breast tissue of their natural ligand. A similar phenomenon was seen in endometrial cell lines but not in osteoblast-like cells (de Gooyer et al., 2001). Differences either in sulfatase inhibition or regulation in breast versus bone cells may account for the bone-sparing effects of the drug. In addition, tissue levels of sulfated metabolites may also act as a drug reservoir.

In target tissues (e.g., uterus and vagina), the major metabolite was 3β-hydroxytibolone and not the 3α-hydroxy-derivative (Blom, 2001). The formation of the 3β-hydroxymetabolite and the Δ⁴-isomer was thought to be catalyzed by NAD(H)-dependent 3β-HSD/Δ^5-4-ketosteroid isomerase (3β-HSD/KSI). However, the formation of 3β-hydroxytibolone was NADPH-dependent and could not be inhibited by the 3β-HSD/KSI inhibitor epostane (Blom, 2001). Moreover, the formation of the Δ⁴-isomer was neither cofactor-dependent nor inhibitable by epostane (Blom, 2001). These results suggested routes other than 3β-HSD/KSI to these metabolites.

The four cytosolic NADPH-dependent enzymes of the aldo-keto reductase (AKR) 1C subfamily [i.e., human 20α-HSD (AKR1C1), human 3α-HSD type 3 (AKR1C2; also known as bile acid-binding protein), human 3α-HSD type 2 (AKR1C3; also known as human 17β-HSD type 5), and human 3α-HSD type 1 (AKR1C4) (Penning, 1997)] catalyze the reduction of the Δ⁵⁽¹⁰⁾-3-ketosteroid tibolone to yield 3α- and 3β-hydroxytibolone in vitro (Steckelbroeck et al., 2004b). The ratio of reductive 3α-HSD versus 3β-HSD activity varied significantly among the isoforms. With tibolone, AKR1C1 and AKR1C2 exclusively catalyzed the formation of 3β-hydroxytibolone, AKR1C3 showed weak 3β/3α-HSD activity, and AKR1C4 acted predominantly as a 3α-HSD (Table 1).

AKR1C4 expression is liver-specific, whereas mRNA expression of the other three isoforms occurs in a tissue-specific pattern in humans (Penning et al., 2000). Consequently, the preference of AKR1C1 and AKR1C2 to form 3β-hydroxytibolone and the preference of AKR1C4 to form 3α-hydroxytibolone could explain why 3β-hydroxytibolone is the major metabolite in human target tissues and why 3α-hydroxytibolone is the major circulating metabolite (Steckelbroeck et al., 2004b). In this study, we investigated AKR1C isozymes in the hepatic metabolism of tibolone using human liver autopsy samples, human hepatoma cells (HepG2), and primary human hepatocytes. Reactions were phenotyped using AKR1C-selective inhibitors. We find that the pattern of tibolone metabolism is consistent with the known properties and expression patterns of AKR1C enzymes, that 3β-hydroxytibolone is the major hepatic metabolite, and that the liver-specific AKR1C4 is not the major source of circulating 3α-hydroxytibolone.

Materials and Methods

Chemicals. Tibolone (Org OD14), 3α-hydroxytibolone, and 3β-hydroxytibolone [[3α- or 3β,7α,17α]-3,17-dihydroxy-7-methyl-19-norpregn-5(10)-en-20-yn-3-one; Org 4094 and Org 30126, respectively]; Δ⁴-isotibolone [[7α,17α]-17-hydroxy-7-methyl-19-norpregn-4(5)-en-20-yn-3-one, Org OM38]; [16-³H]tibolone (32.8 Ci/mmol); [16-³H]3α-hydroxytibolone (38.0 Ci/mmol); and [16-³H]3β-hydroxytibolone (34.4 Ci/mmol) were obtained from N.V. Organon (Oss, Netherlands). Unlabeled Tibolone and its metabolites were judged ≥98.9% pure by HPLC and ¹H NMR were consistent with their assigned structures. [16-³H]Tibolone and its radioactive metabolites were found to have ≥95% radiochemical purity based on radiochromatography by either HPLC or TLC. Flufenamic acid was obtained from ICN Biomedical Inc. (Aurora, OH). 5β-Cholanic acid-3α,7α-diol was purchased from Steraloids, Inc. (Newport, RI). Phenolphthalein was obtained from Fisher Scientific International (Pittsburg, PA). Pyridine nucleotides were purchased from Roche Applied Science (Indianapolis, IN). TRIzol LS reagent, Superscript II preamplification system, and oligonucleotide primers were obtained from Invitrogen (Carlsbad, CA). All other reagents were of American Chemical Society grade or better.

Sources of Tissues and Cells. Human liver autopsy samples were obtained from the National Disease Research Interchange (Philadelphia, PA). Liver tissues from four female Caucasian donors aged 34, 74, 74, and 77 years, were used for this study. None of the donors suffered from liver disease, and the causes of death were cardiopulmonary arrest, chronic obstructive pulmonary disease, hypoglycemia, and cardiac arrest. The human hepatoma cell line HepG2 (HB8065) was obtained from the American Type Culture Collection (Manassas, VA). HepG2 cells were maintained in 6-cm dishes at 37°C and 5% CO₂ containing 5 ml of minimum essential medium (MEM) supplemented with 500 U of penicillin, 500 U of streptomycin, 2 mM L-glutamine, and 10% heat-inactivated charcoal-stripped fetal bovine serum (FBS). Primary cultures of human hepatocytes from single donors were obtained from Cambrex Bio Science Walkersville, Inc. (Walkersville, MD) and delivered on Matrigel coated 6-well plates. Hepatocytes were maintained for 24 to 48 h at 37°C and 5% CO₂ in 2 ml of the culture medium with supplements and growth factors recommended and supplied by the manufacturer. Experiments were performed with primary hepatocytes from three Caucasian donors: a 7-year-old boy, a 53-year-old man, and a 46-year-old woman. None of the donors suffered from liver disease. The study protocols were given expedited approval by the Institutional Review Board, “Committee on Studies Involving Human Beings, Derived Material or Data” at the University of Pennsylvania (approval date July 29, 2004).

Real-Time Reverse Transcription-PCR Analyses. Real-time reverse transcription (RT)-polymerase chain reaction analyses of AKR1C isoform mRNA expression were performed by adapting a method described previously (Stoffel-Wagner et al., 2003). Culture medium was removed from either one 10-cm dish of confluent HepG2 cells (passage 8) or one well of a primary culture of human hepatocytes (from an untreated well at the last experimental time point). Cells were washed with phosphate-buffered saline. Cells were lysed, and total RNA was extracted using 1 ml of the TRIzol reagent. The TRIzol step was repeated to eliminate traces of contaminating DNA. Normal human liver total RNA from a 51-year-old man was obtained from BD Bioscience CLONTECH (Palo Alto, CA). Reverse transcription of 1 μg of total RNA with Superscript II reverse transcriptase was performed according to the manufacturer's protocol. Normal human liver total cDNA obtained from a 59-year-old woman was provided by Prof. Dr. D. Klingmüller from the Department of Clinical Biochemistry of the University Hospital of Bonn (Bonn, Germany).

To amplify the mRNA species, specific oligonucleotide primers crossing intron-exon boundaries were used. Real-time PCR reactions were carried out with the DNA Engine 2 Opticon (MJ Research, Inc., Waltham, MA) using the nonspecific DNA binding dye SYBR green I for detection of the PCR products. Real-time PCR was set up using 12.5 μl of the PCR master mix (QuantiTect SYBR Green PCR Kit; QIAGEN, Valencia, CA) supplemented with 10 pmol of the respective forward and reverse oligonucleotide primers and either the external standards or 25 ng of total cDNA to give a final volume of 25 μl. For quantification of AKR1C, serial dilutions of authentic full-length standard cDNA were used as external standards (Steckelbroeck et al., 2004a). For quantification of the low-copy housekeeping gene porphobilinogen deaminase (PBGD), serial dilutions of purified PCR product were used as external standards. Thirty-five amplification cycles were conducted, and acquisition of the fluorescent signal was adjusted to the end of each 72°C extension step. Initial differences in the amount of RNA subjected to RT were corrected by determining the mRNA expression levels of PBGD and the subsequent adjustment of the target gene data with regard to the mean PBGD expression level. Results represent mean values of experiments performed in triplicate and are given as femtogram of target gene RNA per nanogram of total RNA.

Expression and Purification of Recombinant Human AKR1C Isoforms. Homogeneous recombinant human AKR1C isoforms were purified as described previously (Burczynski et al., 1998) to yield enzymes of the following specific activities: 0.21 μmol of androsterone (75 μM) oxidized/min/mg (AKR1C4) and 2.1, 2.5, and 2.8 μmol of 1-acenaphthenol (1 mM) oxidized/min/mg (AKR1C1, AKR1C2, and AKR1C3, respectively). The homogeneous enzymes were stored in aliquots at –80°C.

Radiometric Determination of Recombinant AKR1C Enzyme Activity. Incubations (final volume of 200 μl) were conducted in 100 mM potassium phosphate buffer (pH 7.0) with 5% acetonitrile containing 0.091 μCi of ³H-labeled tibolone. The final substrate concentration (1 μM) was achieved by the addition of unlabeled compound. Purified AKR1C1-AKR1C4 (0.15–0.5 μg) was added, and the reactions were started by the addition of NADPH (2.3 mM final concentration). Assays were performed in the absence and presence of increasing inhibitor concentrations and incubated over time at 25°C. The reactions were terminated by the addition of 750 μl of ice-cold water-saturated ethyl acetate and extracted twice. The combined organic phases were evaporated to dryness, redissolved, and applied to LK6D Silica TLC plates (Whatman Inc., Clifton, NJ) subjected to thin-layer chromatography, and radiochromatograms were scanned with an automatic TLC-linear analyzer (Bioscan Imaging Scanner System 200-IBM with AutoChanger 3000; Bioscan, Washington, DC) as described previously (Steckelbroeck et al., 2004b). The extraction procedure and subsequent steps were optimized to keep procedural losses to a minimum. The percentage recovery of total radioactivity was >95% for the in vitro experiments. In addition, the coefficient of variation (CV) (i.e., the mean ± S.D. expressed as percentage of the mean) was found to be less than 2 to 8% for the replicates.

Computer-aided quantitative evaluation of radiosignals emitted from the plates permitted calculation of the relative amount of each radioactive steroid as a percentage of the total radioactivity recovered from a single TLC lane. Blank values were subtracted. Enzyme activity was expressed as nanomoles/minute/milligram, and a 100% value was assigned to the activity in the absence of enzyme inhibitor. IC₅₀ values were calculated from plots of percentage activity versus inhibitor concentration. The identity of radioactive steroids on the TLC plates was verified by the staining of cochromatographed reference standards. In addition, the experiments were replicated on a larger scale with unlabeled steroids and the products identified versus authenticated standards using LC/MS.

LC/MS Analysis of the Products of Tibolone Reduction Catalyzed by Recombinant AKR1C1-AKR1C4. Liquid Chromatography for LC/MS analysis was carried out using a Waters Alliance 2690 HPLC system (Waters Corporation, Milford, MA). A Synergi Polar RP column (250 × 4.6 mm i.d., 5 μm; Phenomenex, Torrance, CA) was employed. Solvent A was 5 mM aqueous ammonium acetate containing 0.01% trifluoroacetic acid, and solvent B was 5 mM ammonium acetate in acetonitrile containing 0.01% trifluoroacetic acid. The linear gradient was as follows: 70% solvent B at 0 min, 70% solvent B at 2 min, 85% solvent B at 20 min, 70% solvent B at 22 min, and 70% solvent B at 30 min with a flow rate of 1.0 ml/min. All separations were performed at ambient temperature. Mass spectrometry was conducted with an LCQ ion trap mass spectrometer (Thermo Finnigan, San Jose, CA) equipped with an atmospheric pressure chemical ionization source. The mass spectrometer was operated in the positive ion mode with a discharge current of 5 μA applied to the corona needle. Nitrogen was used as the sheath (40 units) and auxiliary (10 units) gas to assist with nebulization. The vaporizer and heated capillary temperature were set at 550 and 180°C, respectively. Full scanning analyses were performed in the range of m/z 150 to 450. Products of the reactions were identified based on their LC retention times and mass spectrum in the atmospheric pressure chemical ionization-positive ion mode relative to those observed with the authentic standards.

Radiometric Determination of Ketosteroid Reductase Activity in Subcellular Fractions of Human Liver and HepG2 Cells. All steps of tissue or cell preparation were carried out at 4°C. Thawed liver autopsy tissues (250 mg) were homogenized in 2.5 ml of ice-cold 50 mM Tris-HCl (pH 7.4) containing 250 mM sucrose, 1 mM EDTA, and 1 mM 2-mercaptoethanol (Tris-HCl-SEM). Eight 10-cm dishes of confluent HepG2 cells were washed with phosphate-buffered saline, harvested, and homogenized in ice-cold Tris-HCl-SEM (500 μl/dish). To obtain a cell-free supernatant, the homogenates in each case were centrifuged at 800g for 10 min to remove cell debris and the nuclear fraction. Aliquots of the supernatants were mixed with 30% glycerol and stored at –80°C to be used as the “whole homogenate fraction.” The remainder was centrifuged at 100,000g for 60 min to obtain the soluble supernatant as the “cytosolic fraction,” which was mixed with 30% glycerol and stored at –80°C. The resulting pellet was resuspended in ice-cold Tris-HCl-SEM by sonication and centrifuged at 100,000g for 60 min to eliminate traces of contaminating cytosolic proteins. The supernatant was discarded, and the resulting pellet was washed twice and then resuspended in Tris-HCl-SEM, mixed with 30% glycerol, and finally stored at –80°C. It was designated the “membrane fraction.” An aliquot of each fraction was used for protein determination.

Incubations (final volume of 200 μl) were conducted in Tris-HCl-SEM with 5% acetonitrile containing 5 mM MgCl₂ and 0.091 μCi of ³H-labeled tibolone. Final substrate concentration (1 μM) was achieved by the addition of unlabeled compound. The respective cell fraction (containing 5–15 μg of liver proteins or 10 μg of HepG2 proteins) was added, and the reaction was started by the addition of NADPH (1 mM final concentration). Assays were incubated in duplicate for 60 min (liver samples) or 30 min (HepG2 samples) at 37°C. The reactions were terminated, steroids were extracted, organic extracts were subjected to thin-layer chromatography, and radiochromatograms were analyzed as described above. The identity of the analytes was established by cochromatography with the standards authenticated by LC/MS. The recovery of the radioactivity in the organic phase accounted for in excess of 80% of the radioactivity. Blank values were subtracted, and enzyme activity was expressed as nanomoles/hour/milligram of protein. In addition, the CVs were found to be within 5% for all replicates. Levels of glucose-6-phosphatase (membrane marker enzyme) and fructose-1,6-diphosphatse (cytosolic marker enzyme) were determined in subcellular liver fractions to validate the fractionation method.

Cell Culture Experiments. HepG2 cells (4 × 10⁶) were incubated in 5 ml of fresh medium containing 10 μM tibolone (including 0.91 μCi of ³H-labeled tibolone in 15 μl of dimethyl sulfoxide) and 10% heat-inactivated charcoal-stripped FBS. In the inhibition experiments, 100 μM flufenamic acid was also added and the concentration of dimethyl sulfoxide was unchanged. Assays were performed in duplicate dishes, and aliquots (500 μl) of the medium were withdrawn twice from each dish at 0, 1, 3, 6, and 9 h and extracted twice with 2.5 ml of water-saturated ice-cold ethyl acetate. The combined organic phases were evaporated to complete dryness. Organic extracts were subjected to TLC, and radiochromatograms were analyzed as described above. The identity of the analytes was established by cochromatography with the standards authenticated by LC/MS. The recovery of the radioactivity in the organic phase accounted for in excess of 90% of the radioactivity. No radioactivity was present in the aqueous phase. In addition the CVs were found to be within 5% for all replicates.

Primary hepatocytes were incubated in 2.5 ml of fresh medium containing 1 μM tibolone, 3α-hydroxytibolone, or 3β-hydroxytibolone (including 1.14 μCi of the ³H-labeled compounds). In the inhibition experiments, an additional 100 μM concentration of the respective inhibitor (phenolphthalein, the NSAID flufenamic acid, or the bile acid 5β-cholanic acid-3α,7α-diol) was added. The effects of each inhibitor were measured in duplicate dishes, and aliquots (200 μl) of the medium were withdrawn twice from each dish at 0, 0.5, 1, 2, and 4 h, diluted with 200 μl of ice-cold H₂O, and extracted twice with 1 ml of water-saturated ice-cold ethyl acetate. Organic phases were combined, evaporated to complete dryness, and analyzed by TLC and autoradiography as described above. The identity of the analytes was established by cochromatography with the standards authenticated by LC/MS. Polar metabolites were quantified by measuring the radioactivity that remained in the aqueous phase after ethyl acetate extraction by liquid scintillation counting. Radioactivity was counted twice in 50 μl of the extracted aqueous phases as automatically quench-corrected disintegrations per minute with a TriCarb 2100 (Packard-Instrument; PerkinElmer Life and Analytical Sciences Inc., Boston, MA). The total radioactivity in the organic and aqueous phases accounted for 80% of the radioactivity. Less than 1% of the radioactivity was retained within the cells, and the remainder was sequestered by the Matrigel matrix.

Results

Expression of AKR1C Isoforms in Fresh Liver Biopsies, in HepG2 Cells, and in Primary Human Hepatocytes. To relate tibolone metabolism in liver autopsy samples, in HepG2 cells, and in primary hepatocytes to AKR1C expression, AKR1C transcript levels were measured by quantitative real-time RT-PCR. Due to RNA degradation, AKR1C expression could not be measured in the respective liver autopsy samples. As a substitute, analysis was performed on fresh liver biopsies of a 51 year-old man and a 59-year-old woman (Fig. 2). In both samples, the rank order of mRNA expression was similar: AKR1C4 > AKR1C1 » AKR1C2 > AKR1C3. We found that human hepatoma cells (HepG2) lacked significant expression of AKR1C4, whereas AKR1C1-AKR1C3 was expressed in the following rank order: AKR1C1 ≥ AKR1C2 ≥ AKR1C3. A different expression pattern was determined in primary human hepatocytes obtained from a 7-year-old boy and a 47-year-old woman. Expression levels of AKR1C1-AKR1C3 were higher, and the expression level of AKR1C4 was lower relative to the biopsy samples. In the hepatocyte samples, the rank order of mRNA expression level was AKR1C1 > AKR1C2 ∼ AKR1C3 » AKR1C4.

Inhibition of Recombinant AKR1C Catalyzed Reduction of Tibolone by Phenolphthalein, the NSAID Flufenamic Acid, and the Bile Acid 5β-Cholanic Acid-3α,7α-diol. Inhibitors of tibolone reduction catalyzed by human recombinant AKR1C isoforms were identified so that the reactions in human liver autopsies and human liver cells could be phenotyped. These studies identified three useful inhibitors. Phenolphthalein inhibited all four enzymes with the highest affinity for AKR1C4 (Fig. 3A). The IC₅₀ values for the reduction of tibolone were 1.27 μM for AKR1C1, 5.80 μM for AKR1C2, 0.90 μM for AKR1C3, and 0.40 μM for AKR1C4. The NSAID flufenamic acid was a potent inhibitor of AKR1C1-AKR1C3-catalyzed reduction of tibolone and a weak inhibitor of AKR1C4 (Fig. 3B). The IC₅₀ values were 0.98 μM for AKR1C1, 0.63 μM for AKR1C2, 1.65 μM for AKR1C3, and 1080 μM for AKR1C4. Our experiments also confirmed the specificity of the bile acid 5β-cholanic acid-3α,7α-diol for AKR1C2 (Fig. 3C). The IC₅₀ values were 0.21 μM for AKR1C2 and 74.4 μM for AKR1C3 and were not detectable for AKR1C1 and AKR1C4.

Authentication of Metabolite Identity Formed by Recombinant AKR1C1-AKR1C4 by LC/MS. Unlabeled tibolone was incubated with each of the four recombinant AKR1C isoforms in the presence of NADPH until the reaction reached completion. The organic soluble products were extracted and analyzed by LC/MS. Products were identified based on their retention times and ions relative to those observed with the authentic synthetic standards as follows: tibolone [MH⁺ – H₂O; m/z = 295]; 3α- and 3β-hydroxytibolone [MH⁺ – 2H₂O; m/z = 279]; and Δ⁴-isotibolone [MH⁺; m/z = 313]. It was found that AKR1C1 and AKR1C2 formed 3β-hydroxytibolone exclusively; AKR1C3 gave a mixture of 3α-hydroxytibolone:3β-hydroxytibolone (1:5.4, area ratio on LC/MS chromatogram), and AKR1C4 gave a mixture of 3α-hydroxytibolone:3β-hydroxytibolone (2.8:1, area ratio on LC/MS chromatogram) (Fig. 4).

NADPH-Dependent Metabolism of Tibolone in Subcellular Fractions of Liver. Subcellular fractions of four human liver autopsy samples were prepared to determine whether cytosolic 3α/3β-HSD activities would reduce tibolone into its active 3-hydroxymetabolites. The activities of subcellular marker enzymes were measured to validate the fractionation method. Distribution of the marker enzyme glucose-6-phosphatase clearly verified that the membrane fraction was enriched, and distribution of the marker enzyme fructose-1,6-diphosphatase verified that the cytosolic fraction was enriched with soluble protein (Fig. 5D).

Conversion of tibolone into 3α/3β-hydroxytibolone and Δ⁴-isotibolone was observed in the subcellular liver fractions (Fig. 5). 3α-HSD activity was found in the cytosolic as well as the membrane fractions (Fig. 5A). In the whole homogenate and the cytosolic fraction, conversion of tibolone into 3α-hydroxytibolone was potently inhibited by phenolphthalein, whereas flufenamic acid and 5β-cholanic acid-3α,7α-diol were without effect. In contrast, membrane-associated 3α-HSD activity was inhibited by these three compounds. Conversion of tibolone into Δ⁴-isotibolone was highest in the membrane fraction and was decreased by phenolphthalein (Fig. 5C). High 3β-HSD activity was exclusively found in the cytosolic fractions (Fig. 5B). In both the whole homogenate and the cytosolic fraction, conversion of tibolone into 3β-hydroxytibolone was potently inhibited by flufenamic acid and 5β-cholanic acid-3α,7α-diol, whereas phenolphthalein had a small effect. Formation of 3β-hydroxytibolone was favored over 3α-hydroxytibolone by the ratio of 4:1.

NADPH-Dependent Metabolism of Tibolone in Subcellular Fractions of HepG2 Cells. To determine whether cytosolic 3α/3β-HSD activities catalyze the conversion of tibolone into its active 3-hydroxymetabolites within HepG2 cells, we investigated the NADPH-dependent reduction of tibolone in subcellular fractions of this cell line. Conversion of tibolone into 3β-hydroxytibolone was again observed, whereas 3α-HSD activity was not detectable (Fig. 6). 3β-HSD activity was enriched in the cytosolic fraction and inhibited by flufenamic acid. Minor conversion of tibolone into Δ⁴-isotibolone was observed in the membrane fraction and was not blocked by the AKR1C inhibitors.

Inhibition of the Time-Dependent Reduction of Tibolone in HepG2 Cells by the NSAID Flufenamic Acid. Human HepG2 cells expresses significant amounts of AKR1C1-AKR1C3 but do not express 3β-HSD/KSI isoforms and provide a system to study the contribution of the AKR1C isoforms to the in vivo formation of 3β-hydroxytibolone (Steckelbroeck et al., 2004a) (Fig. 2). The metabolism of tibolone (10 μM final concentrations) in HepG2 cell culture with MEM plus 10% heat-inactivated charcoal-stripped FBS was investigated. Time-dependent formation of almost equal amounts of 3β-hydroxytibolone and Δ⁴-isotibolone was observed (Fig. 7A). The formation of both metabolites was slightly inhibited by flufenamic acid. Strong conversion of tibolone into Δ⁴-isotibolone was also observed in control incubations without cells (Fig. 7B). This phenomenon could be clearly attributed to the presence of FBS in the culture medium (data not shown). Consequently, additional experiments with reduced amounts of FBS were performed. Optimal results were achieved when HepG2 cells were first preincubated for 3 h with or without the inhibitor in fresh MEM containing 1% FBS and then incubated in fresh MEM containing 1% FBS and tibolone (in the presence or absence of the inhibitor). In these experiments, only minor amounts of Δ⁴-isotibolone were formed and the high conversion of tibolone into 3β-hydroxytibolone was potently inhibited by flufenamic acid (Fig. 7C).

Inhibition of the Time-Dependent Reduction of Tibolone in Primary Hepatocytes by the NSAID Flufenamic Acid, Phenolphthalein, and the Bile Acid 5β-Cholanic Acid-3α,7α-diol. Human primary hepatocytes were also used to determine the contribution of all four AKR1C isoforms to the in vivo metabolism of tibolone. Similar results were observed in primary hepatocytes from a 7-year-old boy, a 53-year-old man, and a 46-year-old woman. The results obtained with hepatocytes from the 46-year-old female donor are shown (Fig. 8). Tibolone was rapidly metabolized by the hepatocytes and was completely consumed within 4 h. By the first hour, 40% tibolone was converted into 3α- and 3β-hydroxytibolone (Fig. 8A). Formation of 3β-hydroxytibolone was favored, and the ratio of 3β-hydroxytibolone to 3α-hydroxytibolone was 4:1. Both metabolites underwent further conversion into polar metabolites, which were (at later time points) detected as nonmigrating metabolites in TLC analyses and as radioactive metabolites that remain in the aqueous phase after organic extraction (Fig. 8D). The formation of Δ⁴-isotibolone was minor. By 4 h, the bulk of tibolone and/or its metabolites were conjugated.

Treatment with the three AKR1C inhibitors showed varying effects on the metabolism of tibolone in primary hepatocytes. The NSAID flufenamic acid inhibited the reduction of tibolone into 3β-hydroxytibolone (Fig. 7A), and as a consequence, the formation of 3α-hydroxytibolone was increased. Treatment with phenolphthalein inhibited the reduction of tibolone into the 3α-hydroxymetabolite. Significantly higher amounts of 3β-hydroxytibolone and, to a lesser extent, Δ⁴-isotibolone were now observed. Phenolphthalein was also a potent inhibitor of the subsequent formation of polar metabolites (Fig. 8D), which could account for the apparent higher than expected formation of 3β-hydroxytibolone and Δ⁴-isotibolone. Treatment with the bile acid 5β-cholanic acid-3α,7α-diol also inhibited the formation of 3β-hydroxytibolone but to a lesser extent than flufenamic acid.

The metabolism 3α/3β-hydroxytibolone in primary hepatocytes was investigated to determine i) whether conversion of these compounds into polar metabolites is inhibited by phenolphthalein, ii) whether metabolism of either compound is favored, and iii) whether these compounds undergo oxidation back to tibolone (with the subsequent formation of Δ⁴-isotibolone). 3α-Hydroxytibolone was rapidly metabolized by 3 h with 85% of the compound being converted to polar metabolites (Fig. 8B). Minor amounts of 3β-hydroxytibolone, tibolone, and the Δ⁴-isotibolone were observed. The metabolism of 3α-hydroxytibolone was potently inhibited by phenolphthalein, which resulted in a corresponding decrease in polar metabolites (Fig. 8, B and E). Some 3β-hydroxytibolone (15%) also accumulated because of inhibition of subsequent conjugation. The formation of 3β-hydroxytibolone was also blocked by flufenamic acid and 5β-cholanic acid 3α,7α-diol, suggesting the following reaction sequence for the epimerization of the steroid: initial oxidation of 3α-hydroxytibolone by AKR1C4 to yield tibolone, which is then reduced by AKR1C1 and AKR1C2 to yield 3β-hydroxytibolone.

3β-Hydroxytibolone was also rapidly metabolized (>90%) by 3 h with the majority of the compound being converted to polar metabolites. Small amounts (<4%) of 3α-hydroxytibolone, tibolone, and Δ⁴-isotibolone were observed. Flufenamic acid and 5β-cholanic acid-3α,7α-diol failed to inhibit the appearance of these minor metabolites. By contrast, phenolphthalein almost completely blocked 3β-hydroxytibolone metabolism and largely eliminated the formation of these minor metabolites; with this inhibitor, there was a concomitant block in the formation of polar metabolites. These data are consistent with the epimerization of 3β-hydroxytibolone to 3α-hydroxytibolone catalyzed by AKR1C4 and the ability of phenolphthalein to block the conjugation of free 3β-hydroxytibolone.

Discussion

Tibolone is bioactivated via Phase 1 metabolism to produce 3α- and 3β-hydroxymetabolites that are in part responsible for its tissue-selective actions (Kloosterboer and Ederveen, 2003). However, the enzymes responsible for the in vivo formation of these metabolites have remained elusive. Candidate enzymes are the human AKR1C isoforms that catalyze the reduction of the Δ⁵⁽¹⁰⁾-3-ketosteroid tibolone to yield 3α- and 3β-hydroxytibolone in different ratios in vitro (Steckelbroeck et al., 2004b). Importantly, these enzymes work solely as ketosteroid reductases in vivo due to the potent inhibition of their oxidase activity by NADPH (Rižner et al., 2003; Steckelbroeck et al., 2004a,b). We now show that tibolone is metabolized in human liver homogenates, in human HepG2 cells, and in human primary hepatocytes to yield predominantly 3β-hydroxytibolone. The formation of this metabolite is favored based on knowing the in vitro activity and expression patterns of the AKR1C isoforms.

In whole liver, the rank order of mRNA expression was AKR1C4 > AKR1C1 » AKR1C2 > AKR1C3 and implicates these enzymes in the conversion of tibolone into its 3α- and 3β-hydroxymetabolites. In HepG2 cells, there was no detectable expression of AKR1C4 mRNA that tibolone would be reduced solely into its 3β-hydroxymetabolite. In primary hepatocytes, all four AKR1C isoforms were expressed but the levels of AKR1C1-AKR1C3 mRNA were higher than AKR1C4, which predicts that formation of 3β-hydroxytibolone would be favored. To test these predictions, hepatic metabolism studies were performed in the presence of AKR1C inhibitors to phenotype the reactions. The bile acid 5β-cholanic acid-3α,7α-diol was used to inhibit AKR1C2. Flufenamic acid was used as a potent inhibitor of AKR1C1, AKR1C2, and AKR1C3, and phenolphthalein was used as a selective inhibitor for AKR1C4 (Higaki et al., 2003).

Human liver autopsy subcellular fractions metabolized tibolone into 3α- and 3β-hydroxytibolone and Δ⁴-isotibolone, where the formation of 3β-hydroxytibolone was favored. NADPH-dependent formation of the 3α-hydroxymetabolite was observed in the cytosol and the membrane fraction (Fig. 5A). Potent and exclusive inhibition of this reaction by phenolphthalein implicates AKR1C4 as the cytosolic source of the 3α-HSD activity. Based on the high level of AKR1C4 mRNA expression observed in fresh liver biopsies (Fig. 1), it was anticipated that the formation of the 3α-hydroxytibolone metabolite would be greater. This discrepancy maybe due to differences in i) the stability of the AKR1C isoforms in the autopsy samples, ii) translation efficiency of the different AKR1C mRNAs, or iii) k_cat/K_M values of the AKR1C enzymes that compensate for differences in expression levels. Short-chain dehydrogenase/reductases might account for the modest reduction of tibolone to 3α-hydroxytibolone in liver microsomes, but their involvement is ruled out because these are mainly NAD⁺-dependent oxidative enzymes (Biswas and Russell, 1997; Gough et al., 1998; Wang et al., 1999; He et al., 2000; Chetyrkin et al., 2001; He et al., 2003; Liden et al., 2003).

In human liver autopsy samples, high NADPH-dependent formation of 3β-hydroxytibolone was exclusively enriched in the cytosolic fraction (Fig. 3B) and was potently inhibited by 5β-cholanic acid-3α,7α-diol and by flufenamic acid, whereas phenolphthalein was almost without effect. These results implicate AKR1C2 as the isoform responsible, because it has the highest k_cat/K_M value for the reduction of tibolone of all of the AKR1C isoforms (Table 1) (Steckelbroeck at al., 2004b). Thus, the expression of AKR1C2 is crucial for the hepatic conversion of tibolone into the active 3β-hydroxymetabolite.

Isomerization of tibolone was localized to the liver membrane fraction (Fig. 3C). However, fast nonenzymatic conversion of tibolone into Δ⁴-isotibolone occurs in the presence of acid or base (S. Steckelbroeck and T. M. Penning, unpublished data) (Perera et al., 1980) or in the presence of heat-inactivated FBS (Fig. 6B) and this activity may be nonenzymatic. A fraction of this activity was inhibited by phenolphthalein, leaving open the possibility that an unknown isomerase may contribute to this reaction.

In HepG2 subcellular factions, tibolone was reduced to 3β-hydroxytibolone by a cytosolic 3β-HSD activity that was potently inhibited by flufenamic acid. Minor formation of Δ⁴-isotibolone was again observed in the membrane fraction, but 3β-HSD/KSI was not responsible because this enzyme is not expressed in HepG2 cells (Steckelbroeck et al., 2004a).

In HepG2 cell culture experiments, isomerization of tibolone was noted (Fig. 7A); however, this was attributed to the presence of heat-inactivated FBS in the medium (Fig. 7B). In optimized assays with reduced amounts of heat-inactivated FBS, only minor amounts of Δ⁴-isotibolone were formed and a strong time-dependent conversion of tibolone into 3β-hydroxytibolone was again observed, which was potently inhibited by flufenamic acid (Fig. 7C). AKR1C2 is implicated as being responsible because of its expression level and high catalytic efficiency. The conversion of tibolone to 3β-hydroxytibolone was quantitative, and no conjugation was detected. HepG2 cells have high glucuronyltransferase activity but low sulfotransferase activity (Song et al., 1998; Narasaka et al., 2000; Duanmu et al., 2002), and this supports the observation that the 3-hydroxymetabolites of tibolone are not glucuronidated (Sandker et al., 1994; Vos et al., 2002). No formation of 3α-hydroxytibolone was observed, consistent with the lack of AKR1C4 expression in HepG2 cells.

In primary cultures of human hepatocytes, rapid metabolism of tibolone into 3α- and 3β-hydroxytibolone was observed (Fig. 8A). Formation of Δ⁴-isotibolone was minor. Within 1 h, 40% tibolone was converted into the two 3-hydroxymetabolites and the formation of the 3β-hydroxymetabolite was favored 4-fold. These metabolites disappeared at later time points because of their conversion into polar metabolites (Fig. 8D). Tibolone and its 3-hydroxymetabolites are efficiently converted into sulfate conjugates via sulfotransferases (Vos et al., 2002). Human hydroxysteroid sulfotransferase SULT2A1 and human estrogen sulfotransferase SULT2E1 (and to a minor extent human cholesterol sulfotransferase SULT2B1b) catalyze this reaction in vitro (Falany et al., 2004) and are expressed in human hepatocytes (Song et al., 1998; Narasaka et al., 2000; Duanmu et al., 2002).

The formation of 3β-hydroxytibolone in hepatocytes was inhibited by flufenamic acid and (to a lesser extent) by 5β-cholanic acid-3α,7α-diol, implicating both AKR1C2 and AKR1C1 in this reaction. The formation of 3α-hydroxytibolone was potently and exclusively inhibited by phenolphthalein implicating AKR1C4 in its formation. The 3α-hydroxymetabolite was elevated in the presence of flufenamic acid and 5β-cholanic acid-3α,7α-diol due to a redirection of tibolone metabolism. In the presence of phenolphthalein, substantially more 3β-hydroxytibolone accumulated in hepatocyte cultures than could be expected, due to AKR1C4 inhibition alone, and suggested that inhibition of sulfotransferase also occurred (Fig. 8, D–F).

Metabolism of 3α-hydroxytibolone and 3β-hydroxytibolone revealed that these compounds were rapidly (>90%) conjugated to water-soluble metabolites. Conjugation of both diastereomers was inhibited by phenolphthalein confirming that this agent blocks sulfonation. Moreover, epimerization of both diastereomers (4–8%) was observed (Fig. 8, B and C). The reaction sequence involves oxidation to tibolone followed by a reduction reaction that ultimately yields the opposing isomer. Inhibition of the low oxidation/epimerization of the 3β-diastereomer was only observed with phenolphthalein, which indicates that phenolphthalein-sensitive AKR1C4 activity was responsible for the ultimate formation of the 3α-isomer as previously reported (Table 1) (Steckelbroeck et al., 2004b). The stronger oxidation/epimerization of the 3α-diastereomer was inhibited by all three inhibitors and presumably reflects the role of AKR1C4 to form tibolone from 3α-hydroxytibolone and the role of AKR1C1-AKR1C3 to then reduce tibolone into 3β-hydroxytibolone.

In conclusion, 3β-hydroxytibolone is the major metabolite of tibolone in human liver samples, HepG2 cells, and human hepatocytes and this reaction is due to AKR1C2 and AKR1C1. AKR1C2 is the dominant source because of its more favorable k_cat/K_m (Table 1). AKR1C1 and AKR1C2 are also expressed in peripheral target tissues (Penning et al., 2000) and are likely responsible for the extrahepatic formation of 3β-hydroxytibolone. Hepatic 3α-HSD activity was lower than expected and lower than the observed 3β-HSD activity. Consequently, AKR1C4 and hence the liver may not be the major source of circulating 3α-hydroxytibolone. These results contradict the assumption that 3α-hydroxytibolone (the major circulating Phase I metabolite) is predominantly formed via hepatic 3α-HSD activity (Timmer et al., 2002; Vos et al., 2002). We conclude that an unknown extrahepatic source of 3α-HSD (e.g., intestinal or enterobacterial) may play an important role in the formation of 3α-hydroxytibolone after oral administration of the drug.