Abstract
Enzymes of the aldo-keto reductase (AKR) and short-chain dehydrogenase/reductase superfamilies are involved in the reduction of compounds containing a ketone group. In most cases, multiple isoforms appear to be involved in the reduction of a compound, and the enzyme(s) that are responsible for the reaction in the human liver have not been elucidated. The purpose of this study was to quantitatively evaluate the contribution of each isoform to reduction reactions in the human liver. Recombinant cytosolic isoforms were constructed, i.e., AKR1C1, AKR1C2, AKR1C3, AKR1C4, and carbonyl reductase 1 (CBR1), and a microsomal isoform, 11β-hydroxysteroid dehydrogenase type 1 (HSD11B1), and their contributions to the reduction of 10 compounds were examined by extrapolating the relative expression of each reductase protein in human liver preparations to recombinant systems quantified by liquid chromatography–mass spectrometry. The reductase activities for acetohexamide, doxorubicin, haloperidol, loxoprofen, naloxone, oxcarbazepine, and pentoxifylline were predominantly catalyzed by cytosolic isoforms, and the sum of the contributions of individual cytosolic reductases was almost 100%. Interestingly, AKR1C3 showed the highest contribution to acetohexamide and loxoprofen reduction, although previous studies have revealed that CBR1 mainly metabolizes them. The reductase activities of bupropion, ketoprofen, and tolperisone were catalyzed by microsomal isoform(s), and the contributions of HSD11B1 were calculated to be 41%, 32%, and 104%, respectively. To our knowledge, this is the first study to quantitatively evaluate the contribution of each reductase to the reduction of drugs in the human liver.
SIGNIFICANCE STATEMENT To our knowledge, this is the first study to determine the contribution of aldo-keto reductase (AKR)-1C1, AKR1C2, AKR1C3, AKR1C4, carbonyl reductase 1, and 11β-hydroxysteroid dehydrogenase type 1 to drug reductions in the human liver by utilizing the relative expression factor approach. This study found that AKR1C3 contributes to the reduction of compounds at higher-than-expected rates.
Introduction
Most administered drugs are metabolized mainly in the liver and intestines by oxidation, reduction, hydrolysis, or conjugation. Reduction reactions account for 2.2% of the metabolic reactions of clinical drugs (Fukami et al., 2022). In particular, compounds containing ketone, nitro, and sulfoxide groups undergo reduction reactions. Although cytochrome P450s, which are major drug-metabolizing enzymes, catalyze not only oxidation reactions but also reduction reactions, the drugs reduced by cytochrome P450 are limited (Zangar et al., 2000). Most reduction reactions of the ketone groups in compounds are catalyzed by enzymes belonging to the aldo-keto reductase (AKR) and short-chain dehydrogenase/reductase (SDR) superfamilies. In humans, there are 15 AKR isoforms—AKR1A1, AKR1B1, AKR1B10, AKR1B15, AKR1C1, AKR1C2, AKR1C3, AKR1C4, AKR1D1, AKR1E2, AKR6A3, AKR6A5, AKR6A9, AKR7A2, and AKR7A3 (www.med.upenn.edu/akr)—and 75 SDR isoforms (Persson and Kallberg, 2013), including carbonyl reductase (CBR)-1 (SDR21C1), CBR3 (SDR21C2), CBR4 (SDR45C1), 11β-hydroxysteroid dehydrogenase type 1 (HSD11B1, SDR26C1), dehydrogenase/reductase member 4 (SDR25C2), and dicarbonyl/L-xylulose reductase (SDR20C1). Among them, 12 AKR isoforms—with the exception of AKR6A3, AKR6A5, and AKR6A9—and six SDR isoforms, CBR1, CBR3, CBR4, HSD11B1, dehydrogenase/reductase member 4, and dicarbonyl/L-xylulose reductase, have been considered to participate in the reduction of xenobiotics (Matsunaga et al., 2006).
In most cases, multiple isoforms appear to be involved in compound reduction. For example, doxorubicin has been reported to be reduced to doxorubicinol by AKR1A1, AKR1B1, AKR1B10, AKR1C3, AKR1C4, and CBR1 (Kassner et al., 2008), but the enzyme with the highest contribution to doxorubicin reduction in the human body has not been elucidated. Although Kishimoto et al. (1994) showed—in an inhibition study using several inhibitors—that acetohexamide reduction in erythrocytes is mainly catalyzed by CBR1, Ohara et al. (1995) reported that this reaction was not catalyzed by purified human CBR1. Dalmadi et al. (2003) revealed tolperisone reduction by human liver microsomes (HLM), but the responsible enzyme(s) were not identified. Thus, there was hardly any progression in the identification of the enzyme(s) responsible for each reduction reaction. Estimation of the contribution of each isoform to the reduction of compounds of interest is helpful for predicting interindividual differences in drug efficacy or drug-drug interactions.
Recently, we evaluated the mRNA levels of 12 AKR and 6 SDR isoforms in the human liver and demonstrated that AKR1C1, AKR1C2, AKR1C3, CBR1, and HSD11B1 were expressed at over 10% of the total reductase expression, and the sum of these five isoforms reached 64.5% (Amai et al., 2020). AKR1C4 is moderately expressed (5.6% of the total reductase expression levels) (Amai et al., 2020). Early studies demonstrated that AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 can catalyze drug reduction reactions (Ohara et al., 1995; Matsunaga et al., 2006; Meyer et al., 2013). Therefore, we sought to quantitatively evaluate the contributions of these isoforms to the reduction reactions of 10 representative compounds containing a ketone group in the human liver.
Materials and Methods
Chemicals and Reagents
Acetohexamide, dihydro ketoprofen, doxorubicin hydrochloride, ketoprofen, haloperidol, loxoprofen sodium hydrate, and tolperisone hydrochloride were purchased from FUJIFILM Wako Pure Chemical (Osaka, Japan). Bupropion and naloxone hydrochloride were purchased from Sigma-Aldrich (St. Louis, MO). Doxorubicinol, lisofylline, reduced haloperidol, erythro-dihydro bupropion, and threo-dihydro bupropion were purchased from Toronto Research Chemicals (Toronto, Canada). Oxcarbazepine, hydroxyhexamide, and 6β-naloxol were purchased from LKT Laboratories (St. Paul, MN), Medchem Express (Tokyo, Japan), and Cayman Chemical (Ann Arbor, MI), respectively. trans-Hydroxyloxoprofen and cis-hydroxyloxoprofen were obtained from Daiichi Sankyo (Tokyo, Japan). Glucose-6-phosphate (G6P), G6P dehydrogenase, and β-nicotinamide adenine dinucleotide phosphate (NADP+) were obtained from Oriental Yeast (Tokyo, Japan). The rabbit anti-human HSD11B1 polyclonal antibody was purchased from Boster Biologic Technology (Pleasanton, CA). IRDye 680-labeled goat anti-rabbit IgG and Odyssey blocking buffer were purchased from LI-COR Biosciences (Cambridge, UK). Immobilon-P PVDF membranes were purchased from Millipore (Billerica, MA). Human liver cytosol [pooled human liver cytosol (HLC), n = 150 (male, n = 75; female, n = 75)] and human liver microsomes [pooled HLM, n = 50 (male, n = 25; female n = 25)] were purchased from Corning (Corning, NY). All other chemicals and solvents were of analytical grade or of the highest grade commercially available.
Synthesis of Dihydrotolperisone
To a stirred suspension of lithium aluminum hydride (80.8 mg, 2.13 mol) prepared in dry tetrahydrofuran (10 mL), tolperisone hydrochloride (300 mg, 1.06 mmol) was added at 0°C. After stirring at 0°C for 1 hour, complete consumption of tolperisone was confirmed by TLC. The reaction mixture was poured into ice-cold water. The mixture was extracted with ethyl acetate, and the extract was dried over Na2SO4. Na2SO4 was removed by filtration, and the filtrate was concentrated under reduced pressure to obtain a mixture of threo- and erythro-dihydrotolperisone as a colorless oil. The crude material was analyzed by reversed-phase high-performance liquid chromatography (HPLC) with a Cosmosil 5C18-mass spectrometry (MS)-II column (4.6 × 150 mm; Nacalai Tesque, Kyoto, Japan) at a flow rate of 1 mL/min with an isocratic mobile phase of 80% methanol in water with 0.05% triethylamine. Diastereomers 1 and 2 appeared at 6.4 and 6.9 minutes, respectively. The crude material was then purified by recycling a preparative HPLC system with a Cosmosil 5C18-MS-II column (20 × 250 mm; Nacalai Tesque) at a flow rate of 14 mL/min with an isocratic mobile phase of 85% methanol in water with 0.05% triethylamine. As HPLC purification was not sufficient to separate the diastereomers, the obtained materials were converted to hydrochloride salts and purified by washing with an organic solvent. The purified products were analyzed by 1H NMR spectroscopy with a JNM-ESC600 spectrometer (JEOL, Tokyo, Japan).
Fractions containing diastereomer 1 as the major compound were concentrated to afford colorless oil. The oil was dissolved in dichloromethane (DCM) (1.0 mL), and 4 M HCl in AcOEt (0.5 mL, 2.0 mmol) was added to the DCM solution. The mixture was concentrated to obtain a colorless solid. The solid was washed with DCM:hexane (1:4) to give the hydrochloride salt of dihydrotolperisone diastereomer 1 as a colorless solid (125 mg). The product was characterized by 1H NMR [(600 MHz, CD3OD) δ 7.25 (d, J = 7.8 Hz, 2H), 7.18 (d, J = 7.8 Hz, 2H), 4.77 (d, J = 4.2 Hz, 1H), 3.56 (br s, 2H), 3.34–3.25 (m, 3H), 3.18 (dd, J = 13.2, 4.8 Hz, 1H), 3.15–2.78 (m, 3H), 2.46–2.37 (m, 1H), 2.33 (s, 3H), 2.10–1.40 (m, 6H), 0.92 (d, J = 6.6 Hz, 3H)]. The product purity was 99% (confirmed by quantitative nuclear magnetic resonance using 1,3,5-trimethoxybenzene as the internal standard). Upon taking the purity into account, the yield of diastereomer 1 was 41%.
Fractions containing diastereomer 2 as the major compound were concentrated to obtain colorless oil. The oil was dissolved in DCM (1.0 mL), and 4 M HCl in AcOEt (0.5 mL, 2.0 mmol) was added to the DCM solution. The mixture was concentrated to obtain a colorless solid. The solid was washed with DCM/hexane (1/4) to obtain the hydrochloride salt of dihydrotolperisone diastereomer 2 as a colorless solid (101 mg). The product was characterized by 1H NMR [(600 MHz, CD3OD) δ 7.27 (d, J = 7.8 Hz, 2H), 7.19 (d, J = 7.8 Hz, 2H), 4.43 (d, J = 9.6 Hz, 1H), 3.86 (br d, J = 12.0 Hz, 1H), 3.55 (br d, J = 12.0 Hz, 1H), 3.18 (dd, J = 13.2, 6.0 Hz, 1H), 3.05–2.86 (m, 2H), 2.41–2.30 (m, 1H), 2.33 (s, 3H), 2.05–1.50 (m, 6H), 0.72 (d, J = 6.6 Hz, 3H)]. Product purity was 90% (confirmed by qNMR using 1,3,5-trimethoxybenzene as the internal standard). Upon taking the purity into account, the yield of diastereomer 2 was 30%.
Construction of Recombinant Human AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 To Be Expressed in Sf21 Cells
The expression systems for human AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 were constructed using a Bac-to-Bac Baculovirus Expression System (Invitrogen, Carlsbad, CA). The cDNAs of six enzymes were prepared by reverse-transcription polymerase chain reaction using total RNA purified from human liver samples obtained from the National Disease Research Interchange (Philadelphia, PA) through the Human and Animal Bridging Research Organization (Chiba, Japan). Reverse-transcription polymerase chain reaction products amplified with the primers listed in Table 1 were subcloned into the pFastBac1 vector. The pFastBac1 vectors were transformed into DH10Bac competent cells, followed by transposition of the inserts into bacmid DNA. The recombinant bacmid was transfected into Spodoptera frugiperda Sf21 cells. Other steps were performed in accordance with a method previously described (Fukami et al., 2010). Mock sample was also prepared by transfection with a nonrecombinant bacmid DNA. Protein concentration was determined using the method proposed by Bradford (1976) using γ-globulin as a standard.
Immunoblot Analysis of AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 Proteins
SDS-PAGE and immunoblot analysis were performed by referring to methods reported previously (Honda et al., 2021). HLC and HLM (2 µg) as well as recombinant enzymes (rAKR1C1, rAKR1C2, rAKR1C4, and rCBR1:2 µg, rAKR1C3:5 µg, rHSD11B1:2.5 µg) were separated on 7.5% SDS-PAGE and transferred to an Immobilon-P transfer membrane. The membranes were probed with monoclonal mouse anti-human AKR1C1 antibody (R&D Systems, Minneapolis, MN), monoclonal mouse anti-human AKR1C3 antibody (Sigma-Aldrich), polyclonal sheep anti-human AKR1C4 antibody (R&D Systems), polyclonal rabbit anti-CBR1 antibody (Bethyl Laboratories, Montgomery, AL), polyclonal rabbit anti-human HSD11B1 antibody (Boster Biologic Technology, Wuhan, China), and the corresponding fluorophore-conjugated secondary antibody. The bands were visualized by using an Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE). HLM and recombinant HSD11B1 were deglycosylated using Endo H in accordance with a previously reported method (Muta et al., 2014) and subjected to SDS-PAGE and immunoblot analysis.
Quantification of the Protein Levels of Reductases in Human Liver Fractions and Expression Systems
Absolute protein levels of AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 in the HLC, HLM, and recombinant systems were determined using liquid chromatgraphy-tandem mass spectrometry (LCMS/ MS). The amino acid sequences and product ions used for the quantification of each reductase protein are listed in Table 2. The HLC, HLM, and recombinant samples were diluted with SDS and incubated at 100°C for 10 minutes. Fifty micrograms of the sample was carbamidomethylated with dithiothreitol, followed by treatment with iodoacetamide. Alkylated proteins were precipitated using a mixture of ice-cold methanol, chloroform, and water. The precipitates were dissolved in 8 M urea/Tris-EDTA solution, diluted with 0.1 M Tris-HCl (pH 8.0), and digested with a Trypsin/Lys-C mixture at an enzyme/substrate ratio of 1:100 at 37°C for 16 hours. The digested samples were mixed with stable isotope-labeled peptides and acidified with formic acid. The supernatants were evaluated using an liuquid chromatography system (Eksigent Ekspert NanoLC 400; SCIEX, MA) equipped with a ChromXP C18 column (3 µM, 500 µM I.D × 10 cm; SCIEX) connected to a mass spectrometer (TripleTOF 6600; SCIEX). Elution was performed using a mixture of solvent A, consisting of purified water and formic acid (1000:1, v/v), and solvent B, consisting of acetonitrile and formic acid (1000:1, v/v), as the mobile phase. MultiQuant software 3.0.3 (SCIEX) was used for data analysis. The levels of each protein in the pooled HLC, HLM, and expression systems were evaluated using two multiple reaction monitoring (MRM) modes. The relative expression factor (REF) for each reductase was calculated as follows:
The average REF values evaluated using two kinds of MRM modes were used to calculate the contribution.
Measurement of Reductase Activities for 10 Compounds Containing a Ketone Group
The reductase activities of 10 compounds (Fig. 1) were measured as follows: a typical incubation mixture (final volume of 0.2 mL) containing 100 mM potassium phosphate buffer (pH 7.4), enzyme sources (HLC, HLM, or recombinant reductases), and the substrate was prepared. Acetohexamide, bupropion, haloperidol, ketoprofen, and oxcarbazepine were dissolved in DMSO, and doxorubicin hydrochloride, loxoprofen sodium hydrate, naloxone hydrochloride, pentoxifylline, and tolperisone hydrochloride were dissolved in distilled water. The final concentration of DMSO in the incubation mixture was adjusted to 1%. The assay conditions were determined such that the protein concentration and incubation time were linear (Table 3). The substrate concentrations were set near the Michaelis constant (Km) values identified in this study (Table 3). After preincubation at 37°C for 2 minutes, reactions were initiated by the addition of an NADPH-generating system (5 mM G6P, 0.5 mM NADP+, 5 mM MgCl2, and 1 U/mL G6P dehydrogenase). After incubation at 37°C for the incubation time shown in Table 3, the reaction was terminated by the addition of 0.1 mL of ice-cold acetonitrile (the exception being naloxone, for which ice-cold methanol was used). The mixture was then centrifuged at 20,380g for 5 minutes.
For the measurement of reductase activities of nine compounds—except naloxone—the supernatant was subjected to LC-MS/MS. An LCMS-8040 (Shimadzu, Kyoto, Japan) equipped with an LC-20AD HPLC system was used. The mobile phase was 0.1% formic acid and acetonitrile for the detection of acetohexamide, bupropion, doxorubicin, haloperidol, oxcarbazepine, pentoxifylline, and tolperisone metabolites or 10 mM ammonium acetate (pH 5.6) and acetonitrile for the detection of ketoprofen and loxoprofen metabolites. The acetonitrile concentrations used are listed in Table 4. The column temperature was maintained at 40°C. Nitrogen was used as the nebulizer and drying gas. The operating parameters were optimized as follows: nebulizer gas flow, 3 L/min; drying gas flow, 15 L/min; desolvation line temperature, 300°C; and heat-block temperature, 400°C. The other analytical conditions are listed in Table 4. Each metabolite was monitored in the MRM mode. The analytical data were processed using LabSolutions (version 5.82.1; Shimadzu). In loxoprofen and bupropion reductions, trans- and cis-hydroxyloxoprofen and threo- and erythro-dihydro bupropion were quantified separately. For tolperisone reduction, two reduced metabolites, threo- and erythro-dihydrotolperisone, were separately quantified with confirmation of retention time using the metabolites synthesized above. Racemates of licarbazepine, lisofylline, hydroxy acetohexamide, doxorubicinol, reduced haloperidol, and dihydro ketoprofen were observed as single peaks.
To measure naloxone reductase activity, the supernatant was evaluated using an HPLC system (Hitachi, Tokyo, Japan) equipped with an L-2200 autosampler (Hitachi), L-2130 pump (Hitachi), L-2350 column oven (Hitachi), L-2400 UV detector (Hitachi), and D-2500 Chromatographic Integrator (Hitachi). The mobile phase was 20% methanol/50 mM potassium phosphate (pH 6.0), and the column used was Wakopak eco-ODS (4.6 × 150 mm, 5 μm, Wako Pure Chemical). The eluent was monitored at a wavelength of 215 nm. Column temperature was set at 35°C, and the flow rate was 1.0 mL/min. Only 6β-naloxol was detected under the condition in which 6α-naloxol and 6β-naloxol were separated.
Kinetic Analyses
Kinetic analyses were performed using the following ranges of substrate concentrations: acetohexamide, 0–1.5 mM or 0–100 µM; doxorubicin, 0–600 µM; haloperidol, 0–500 µM; ketoprofen, 0–500 µM; oxcarbazepine, 0–50 µM; pentoxifylline, 0–1.0 mM or 0–400 µM; bupropion, 0–200 µM; loxoprofen, 0–500 µM or 0–250 µM; naloxone, 0–250 µM; and tolperisone, 0–400 µM. The kinetic parameters, including Km, Vmax, and intrinsic clearance (CLint), were determined using Prism 5 software (GraphPad, San Diego, CA).
Calculation of the Contribution of Each Reductase to Activity in HLC or HLM
The substrate concentrations (15 µM oxcarbazepine, 0.5 µM pentoxifylline, 20 µM acetohexamide, 5 µM loxoprofen, 15 µM doxorubicin, 1.5 µM naloxone, 10 µM haloperidol, 45 µM bupropion, 1 µM ketoprofen, and 10 µM tolperisone) to calculate the contribution of each reductase were determined based on the estimated hepatic maximum concentration of the compound (Cinlet,u,max) calculated using the following equation: where fu is the unbound fraction in the blood, Cmax is the maximum concentration in the blood, ka is the first-order rate constant for gastrointestinal absorption, fa is the fraction absorbed from the gastrointestinal tract into the portal vein, and QH is the hepatic blood flow rate. The pharmacokinetic data used in this study were obtained from previous studies (Chiang and Hawks, 2003; Kang et al., 2011; Elger et al., 2013) or interview forms.
The contribution of each isoform to reductase activity in the human liver cytosol or microsomes was calculated by applying REF. REF was determined as the ratio of the abundance of each reductase in the liver cytosol or microsomes to that of the recombinant enzymes. Using REF and measured reductase activity of the recombinant enzyme (Vmeasured), the predicted activity of human liver reductase (Vpredicted) was estimated as follows:
The contribution of each isoform to reductase activity in the liver cytosol (Vliver cytosol) and microsomes (Vliver microsomes) was calculated using the following equation:
Statistical Analysis
Statistical significance between two or multiple groups was determined using a two-tailed Student’s t test or a two-tailed Tukey’s honestly significant difference, respectively. P < 0.05 was considered significant.
Results
Reductase Activities for Compounds Containing a Ketone Group by HLC and HLM
Kinetic analyses of the reductase activities of 10 compounds containing a ketone group were performed by HLC and HLM (Figs. 2 and 3). The obtained kinetic parameters are listed in Table 5. In the reduction reactions of six compounds, acetohexamide, doxorubicin, haloperidol, ketoprofen, oxcarbazepine, and pentoxifylline (Fig. 2), reduced metabolites were detected as racemates in single peaks. In reductase activities for five out of six compounds, except for ketoprofen (Fig. 2, A–C, E, and F), HLC showed significantly higher (3.5–13.8-fold) CLint values than HLM, whereas, with respect to ketoprofen reductase activity, HLM showed a significantly higher CLint value (15-fold) than HLC (Fig. 2D). The activities of acetohexamide and pentoxifylline exhibited biphasic kinetics (Fig. 2, A and F), whereas those of doxorubicin, haloperidol, oxcarbazepine, and pentoxifylline exhibited monophasic kinetics (Fig. 2, B–E). In the reduction reactions of the remaining four compounds, bupropion, loxoprofen, naloxone, and tolperisone (Fig. 3), diastereomers of their reduced metabolites were separately detected. In the reductase activity for bupropion (Fig. 3, A and B), HLC and HLM showed significantly higher (20- and 60-fold, respectively) CLint values of threo-dihydro bupropion formation than those of erythro-dihydro bupropion formation. In threo-dihydro bupropion formation (Fig. 3A), HLM showed a 52-fold higher CLint value than HLC, and their activities exhibited monophasic kinetics. In the reductase activity for loxoprofen (Fig. 3, C and D), HLC and HLM showed significantly higher (7.1- and 8.8-fold, respectively) CLint values with respect to trans-hydroxyloxoprofen formation than those of cis-hydroxyloxoprofen formation. In trans-hydroxyloxoprofen formation (Fig. 3C), HLC showed a significantly (9.1-fold) higher CLint value than HLM, and their activities exhibited biphasic kinetics (Fig. 3C). With respect to the reductase activity of naloxone (Fig. 3, E and F), only HLC showed 6β-naloxol formation, and the activities exhibited monophasic kinetics. For tolperisone reductase activity (Fig. 3, G and H), the peaks with faster and later retention times in the LC-MS/MS analysis were regarded as diastereomers 1 and 2, respectively. HLC and HLM showed significantly higher (35- and 28-fold, respectively) CLint values for diastereomer 2 formation than for diastereomer 1 formation. For diastereomer 2 formation (Fig. 3H), HLM showed a 24-fold higher CLint value than HLC, and their activities exhibited monophasic kinetics.
Expression of AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 Proteins in Expression Systems
To evaluate the catalytic potency of human AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1, expression systems were constructed for these isoforms. Their expression was examined using immunoblot analysis (Fig. 4). Analysis using anti-AKR1C1 antibody revealed a clear band for rAKR1C1. A clear band for rAKR1C2 and a faint band for rAKR1C4 (Fig. 4A) were also observed, which could be due to the high amino acid identity between AKR1C1 and AKR1C2 (97.8%) or AKR1C4 (82.7%). A band was also observed in the HLC (Fig. 4A). In the analysis using the anti-AKR1C3 antibody, a clear band was observed only in case of rAKR1C3 among the expression systems (Fig. 4B). A clear band was observed in the HLC, and a faint band was observed in the HLM (Fig. 4B). Analysis using an anti-AKR1C4 antibody revealed a clear band in rAKR1C4, but bands were also detected in case of rAKR1C1 and rAKR1C2, and faintly in case of rAKR1C3 (Fig. 4C). The band intensity of the HLC was stronger than that of the HLM. Analysis using an anti-CBR1 antibody revealed three bands for rCBR1 (Fig. 4D). This could be due to modification with 2-oxocarbonyl acid, which does not affect CBR1 activity (Wermuth et al., 1993). Three bands were also observed in the HLC and faintly in the HLM. By analyzing the anti-HSD11B1 antibody, two bands were observed only for rHSD11B1 among the expression systems. A clear band was observed in the HLM. Although a band was also observed in HLC with different mobilities than HLM (Fig. 4E), it would be a nonspecific band because HSD11B1 is a microsomal protein. The mobilities of the bands differed between the HLM and rHSD11B1. HSD11B1 has three potential N-glycosylation sites (asparagine residues at 123, 162, and 207), according to NetNglyc 1.0 (https://services.healthtech.dtu.dk/service.php?NetNGlyc-1.0). Deglycosylation using Endo H caused the band to shift to a higher mobility, which was matched between HLM and rHSD11B1 (Fig. 4F). Therefore, the difference in mobility between HLM and rHSD11B1 may be due to the difference in N-glycans between HLM and Sf21 cells. Therefore, the expression of each isoform in the constructed expression system was confirmed.
Absolute Expression of Each Reductase Isoform in HLC, HLM, and Expression Systems
To estimate the contribution of AKR1C1, AKR1C2, AKR1C3, AKR1C4, CBR1, and HSD11B1 to the reduction reaction of each compound by HLC or HLM, their expression in the HLC, HLM, and expression systems was quantified by LC-MS/MS targeting the specific peptide of each reductase (Table 6). Unfortunately, AKR1C1 levels were not evaluated because a specific peptide of AKR1C1 could not be determined due to its high amino acid identity with AKR1C isoforms, especially AKR1C2. Among the expression systems of reductases—except AKR1C1—AKR1C3 showed the highest expression efficiency (Table 6). Although AKR1C isoforms are recognized as cytosolic enzymes (Breyer-Pfaff and Nill, 2004), AKR1C2, AKR1C3, and AKR1C4 have also been detected in HLM. CBR1 (recognized as a cytosolic enzyme) and HSD11B1 (recognized as a microsomal enzyme) were specifically detected in HLC and HLM, respectively. In HLC, the expression of AKR1C3 was the highest (159 pmol/mg), followed by that of AKR1C2 (64 pmol/mg), AKR1C4 (51 pmol/mg), and CBR1 (22 pmol/mg). In HLM, the expression of HSD11B1 (73 pmol/mg) was 5.2-fold higher than that of AKR1C3 (14 pmol/mg). The calculated REF values (average) of AKR1C2, AKR1C3, AKR1C4, and CBR1 in HLC were 0.24, 0.14, 0.18, and 0.13, respectively, and those of HSD11B1 in HLM were 0.29. These values were used to calculate the contribution of each reductase to the reduction reactions of the 10 compounds in HLC or HLM.
Evaluation of Reductase Activities by Recombinant Enzymes and Contribution of Each Reductase to the Reduction Reactions of 10 Compounds in HLC or HLM
To determine the enzyme responsible for the reduction reactions of the 10 compounds in HLC or HLM, the reductase activities of the recombinant reductases were measured at substrate concentrations close to the Km values in HLC or HLM. Kinetic analyses were performed using recombinant reductases with relatively high activity. To allow comparison with the kinetics of HLC or HLM, the activities of recombinant reductases in the kinetic analyses are shown after normalization using REF (Fig. 5). Finally, the contribution of each reductase to the reduction reactions of the 10 compounds by HLC or HLM was calculated at substrate concentrations corresponding to the estimated hepatic maximum concentrations. In the following sections, the reduction reactions of acetohexamide, doxorubicin, haloperidol, loxoprofen, naloxone, oxcarbazepine, and pentoxifylline, which are mainly catalyzed by enzymes in HLC, were analyzed, and those of bupropion, ketoprofen, and tolperisone, which are mainly catalyzed by enzymes in HLM, were subsequently analyzed.
Acetohexamide Reduction
Acetohexamide reductase activity at 100 µM substrate concentration in HLC was 8.6-fold higher than that in HLM (Fig. 5A). All recombinant reductases showed significantly (P < 0.001) higher activities than the mock; in particular, rAKR1C3 and rCBR1 showed relatively high activities (Fig. 5A). The activities of HLC at substrate concentrations from 1 to 100 µM were fitted to the Michaelis-Menten equation, whereas the activities of rAKR1C3 were fitted to the substrate inhibition equation, and those of rCBR1 increased linearly (Fig. 5A). The Km value of rAKR1C3 (10 ± 0.34 µM) was lower than that of the high affinity component in HLC (100 ± 19 µM) (Table 5 and 7). The contributions of AKR1C3 and CBR1 to acetohexamide reductase activity by HLC at a substrate concentration of 20 µM were 78% and 15%, respectively. These results suggest that AKR1C3 mainly catalyzes the reduction of acetohexamide in the liver. The remaining 7% activity of HLC might be attributed to AKR1C1, AKR1C2, and AKR1C4.
Doxorubicin Reduction
Doxorubicin reductase activity at a substrate concentration of 150 µM by HLC was 9.2-fold higher than that by HLM (Fig. 5B). Among the recombinant enzymes, rAKR1C3 and rCBR1 showed significantly (P < 0.001) higher activities than the mock (Fig. 5B). The activities of rAKR1C3 and rCBR1 were fitted to the Michaelis-Menten equation (Fig. 5B). The Km values of rAKR1C3 (47 ± 3.6 µM) and rCBR1 (200 ± 21 µM) were lower and higher, respectively, than those of HLC (Tables 5 and 7). The contributions of AKR1C3 and CBR1 to doxorubicin reductase activity by HLC at a substrate concentration of 15 µM were 51% and 58%, respectively. These results suggest that AKR1C3 and CBR1 predominantly contribute to doxorubicin reduction in the liver to the same extent.
Haloperidol Reduction
Haloperidol reductase activity at a 50 µM substrate concentration by HLC was 2.3-fold higher than that by HLM (Fig. 5C). Among the recombinant enzymes, rCBR1 showed prominently high activity, and rHSD11B1 also showed significantly (P < 0.001) higher activity than the mock (Fig. 5C). As HSD11B1 is a microsomal reductase, subsequent experiments were performed with a focus on CBR1. The activities of rCBR1 were fitted to the Michaelis-Menten equation (Fig. 5C). The Km value of rCBR1 (77 ± 6.4 µM) was similar to that of HLC (85 ± 9.4 µM) (Tables 5 and 7). The contribution of CBR1 to haloperidol reductase activity by HLC at a substrate concentration of 10 µM was 97%, suggesting that CBR1 is the only enzyme responsible for haloperidol reduction in the liver.
Loxoprofen Reduction
Loxoprofen reductase activity at a 50 µM substrate concentration by HLC was 25-fold higher than that by HLM (Fig. 5D). Among the recombinant enzymes, rCBR1 showed the highest activity, followed by rAKR1C3 and rAKR1C4 (Fig. 5D). The activity of HLC at substrate concentrations ranging from 1 to 500 µM exhibited biphasic kinetics (Fig. 3C). The activities of CBR1 linearly increased, the activities of rAKR1C3 fitted to the substrate inhibition equation, and those of rAKR1C4 fitted to the Michaelis-Menten equation (Fig. 5D). The Km values of rAKR1C3 (140 ± 17 µM) and rAKR1C4 (38 ± 3.5 µM) were lower than those of HLC (250 ± 80 µM) (Table 7). The contributions of AKR1C3, AKR1C4, and CBR1 to loxoprofen reductase activity in HLC at 5 µM were 54%, 34%, and 31%, respectively. AKR1C3, AKR1C4, and CBR1 contributed to loxoprofen reduction in the liver.
Naloxone Reduction
Naloxone reductase activity at a substrate concentration of 20 µM was observed by HLC but not by HLM (Fig. 5E). Among the recombinant enzymes tested, rAKR1C4 showed the highest activity (Fig. 5E). The activities of rAKR1C4 fitted to the Michaelis-Menten equation (Fig. 5E). The Km value of rAKR1C4 (31 ± 7.4 µM) was similar to that of HLC (37 ± 6.8 µM) (Tables 5 and 7). The contribution of AKR1C4 to naloxone reductase activity by HLC at a substrate concentration of 1.5 µM was 150%. Although there seemed to be some errors in the calculation of the contribution, it was suggested that AKR1C4 is the major enzyme catalyzing naloxone reduction in the liver.
Oxcarbazepine Reduction
Oxcarbazepine reductase activity at a substrate concentration of 5 µM by HLC was 24-fold higher than that by HLM (Fig. 5F). rAKR1C1, rAKR1C2, rAKR1C3, rAKR1C4, and rCBR1 showed significantly (P < 0.01) higher activity than those of the mock cells (Fig. 5F). The activities of rAKR1C1, rAKR1C2, and rAKR1C3 fitted the Michaelis-Menten equation, and those of rAKR1C4 and rCBR1 increased linearly (Fig. 5F). The Km value of rAKR1C1 (5.5 ± 0.41 µM) was lower than that of HLC (15 ± 3.6 µM). The Km value of rAKR1C2 (20 ± 6.6 µM) was close to that of HLC, and that of rAKR1C3 (56 ± 20 µM) was higher than that of HLC (Tables 5 and 7). The contributions of AKR1C2, AKR1C3, AKR1C4, and CBR1 to oxcarbazepine reductase activity by HLC at a substrate concentration of 15 µM were 25%, 22%, 12%, and 12%, respectively. The remaining 29% of HLC activity may be attributed to AKR1C1.
Pentoxifylline Reduction
The pentoxifylline reductase activity at 100 µM substrate concentration by HLC was 4.4-fold higher than that by HLM (Fig. 5G). rAKR1C1, rAKR1C2, and rAKR1C4 showed significantly (P < 0.001) higher activity than that of the mock control (Fig. 5G). The activities of rAKR1C1, rAKR1C2, and rAKR1C4 fitted to the Michaelis-Menten equation (Fig. 5G). The Km value of rAKR1C1 (36 ± 3.4 µM) was lower than that of HLC (45 ± 4.9 µM), and that of rAKR1C2 (160 ± 9.3 µM) and rAKR1C4 (140 ± 2.1 µM) was higher than that of HLC (Table 7). The contributions of AKR1C2 and AKR1C4 to pentoxifylline reductase activity by HLC at 0.5 µM were 56% and 14%, respectively. The remaining 30% can be attributed to AKR1C1.
Bupropion Reduction
Bupropion reductase activity at a 50 µM substrate concentration by HLM was 19-fold higher than that by HLC (Fig. 5H). Among the recombinant enzymes, rHSD11B1 showed the highest activity, whereas rAKR1C1, rAKR1C3, and rHSD11B1 showed only marginal activity (Fig. 5H). The activities of rHSD11B1 fitted to the Michaelis-Menten equation (Fig. 5H). The Km value of rHSD11B1 (20 ± 2.7 µM) was similar to that of HLM (30 ± 17 µM) (Tables 5 and 7). The contribution of HSD11B1 to the bupropion reductase activity by HLM at 45 µM was 41%. It was suggested that microsomal enzyme(s) other than HSD11B1 would also participate in the reduction of bupropion.
Ketoprofen Reduction
Ketoprofen reductase activity at a 100 µM substrate concentration by HLM was sixfold higher than that by HLC (Fig. 5I). Among the recombinant enzymes, rHSD11B1 and rCBR1 showed prominent activity (Fig. 5I). The activity of rHSD11B1 fitted to the substrate inhibition equation (Fig. 3I). The Km value of rHSD11B1 (120 ± 45 µM) was higher than that of HLM (53 ± 4.4 µM) (Tables 5 and 7). The contribution of HSD11B1 to ketoprofen reductase activity by HLM at 1 µM was 32%, suggesting that microsomal enzyme(s) other than HSD11B1 would also participate in ketoprofen reduction. The activity of rCBR1 linearly increased at substrate concentrations ranging from 1 to 500 µM, which was consistent with that of HLC. The contribution of CBR1 to ketoprofen reductase activity by HLC at a substrate concentration of 1 µM was 84%.
Tolperisone Reduction
Tolperisone reductase activity at a 50 µM substrate concentration by HLM was 27-fold higher than that by HLC (Fig. 5J). rHSD11B1 showed prominent activity, and rCBR1 showed marginal activity (Fig. 5J). The activities of rHSD11B1 were fitted to the Michaelis-Menten equation (Fig. 5J). The Km value of rHSD11B1 (35 ± 8.0 µM) was lower than that of HLM (61 ± 3.9 µM) (Tables 5 and 7). The contribution of HSD11B1 to tolperisone reductase activity in HLM at 10 µM was 104%, suggesting that HSD11B1 is the only enzyme that catalyzes tolperisone reduction in the liver.
Discussion
The reduction reaction, which is a phase I reaction, is involved in the bioactivation and inactivation of clinical drugs. Although multiple AKR and SDR enzymes often participate in drug reduction reactions, the contribution of each enzyme to the reaction in the human liver has never been quantitatively evaluated. To understand the reasons underlying the contribution of interindividual variability in drug efficacy and susceptibility to drug toxicity, knowledge of the responsibility of each enzyme for metabolism of the concerned drugs would be useful. In this study, the REF approach was used to clarify the contribution of each AKR or SDR isoform to the reduction reactions of 10 compounds that are known to be metabolized via reduction reactions in the human liver.
First, kinetic analyses of 10 compounds were performed (Figs. 2 and 3). In acetohexamide, doxorubicin, haloperidol, loxoprofen, naloxone, oxcarbazepine, and pentoxifylline reduction, HLC showed higher activity than HLM, whereas in bupropion, ketoprofen, and tolperisone reduction, HLM showed higher activity than HLC. The reduction efficiencies of some compounds were compared between HLC and HLM, and the subfractions showing higher activity were consistent with previous reports of haloperidol (Inaba and Kovacs, 1989), pentoxifylline (Lillibridge et al., 1996), oxcarbazepine (Malátková et al., 2014), and bupropion (Connarn et al., 2015). For doxorubicin, Skarka et al. (2011) claimed that microsomal enzymes were also significantly involved in the reduction reaction, because CLint value by HLM (1.0 µL/min per mg) was not much lower than that by HLC (2.1 µL/min per mg). In our study, the CLint value by HLC (2.2 ± 0.29 µL/min per mg) was 4.6-fold higher than that by HLM (0.48 ± 0.025 pmol/min per mg, Table 5). In the liver, the cytosolic protein content is approximately fivefold higher than the microsomal protein content (Tabata et al., 2004); therefore, in either case, cytosolic enzymes would play a significant role in doxorubicin reduction in the liver.
To calculate the contribution of each isoform to the reduction reactions in the liver, the hepatic protein levels of each isoform were quantified by proteomics using LC-MS/MS. Considering the relative protein content between cytosolic and microsomal proteins as described above (Tabata et al., 2004), the rank order of the protein level of each isoform in the human liver would be as follows: AKR1C3, AKR1C2, AKR1C4, CBR1, and HSD11B1 (AKR1C1 could not be quantified in this study). Our previous study revealed that the rank order of mRNA levels in the human liver is as follows: AKR1C2, AKR1C3, AKR1C1, CBR1, HSD11B1, and AKR1C4 (Amai et al., 2020). As AKR1C4 appears to be highly expressed at the protein level as compared with the mRNA level in the human liver, posttranslational regulation may significantly contribute to its expression.
Acetohexamide reduction was catalyzed by cytosolic enzymes rAKR1C3 and rCBR1, and the contributions of AKR1C3 and CBR1 to the activity in HLC were calculated to be 78% and 15%, respectively (Fig. 5A). Ohara et al. (1995) evaluated the decrease in NADPH at 1 mM acetohexamide and reported that CBR1 purified from the human liver did not catalyze acetohexamide reduction. The inconsistency with our results may be due to the differences in the evaluation methods. As for doxorubicin reduction, it has been reported, by inhibition and correlation analyses using human cytosol samples, that CBR1 rather than AKR1C3 would be a major contributor (Kassner et al., 2008). However, the present study demonstrated that AKR1C3 and CBR1 equally contributed to the reduction in liver function (Fig. 5B). Reduced metabolites of doxorubicin are associated with cardiotoxicity induced by doxorubicin (Olson and Mushlin, 1990; Forrest et al., 2000). As there are 14- and eightfold interindividual variabilities in AKR1C3 and CBR1 at the mRNA level (Amai et al., 2020), individuals with high expression of both isoforms would be at a high risk of developing adverse effects of doxorubicin. Loxoprofen reduction was catalyzed by rAKR1C3, rAKR1C4, and rCBR1, and their contributions to the activity in HLC were calculated to be 54%, 34%, and 31%, respectively (Fig. 5D). Ohara et al. (1995) reported that CBR1 is responsible for the reduction, but they did not analyze whether AKR1C3 has the ability to catalyze it. The present study demonstrated that AKR1C3 largely contributes to the reduction of acetohexamide, doxorubicin, and loxoprofen, which is consistent with the findings of previous studies.
A previous study by Eyles and Pond (1992) demonstrated that only S-(-)–reduced haloperidol is stereoselectively produced from haloperidol in the human liver, although it is not specified whether the metabolite detected in the present study was in the S-form or not. Haloperidol reduction was only observed in rCBR1 (Fig. 5C). This is consistent with a study by Breyer-Pfaff and Nill (2000), who found, using purified enzymes, that the main catalyst for haloperidol reduction is CBR1. As for naloxone reduction, Ohara et al. (1995) reported—by using purified enzymes—that CLint of 6β-naloxol formation from naloxone by AKR1C4 was significantly higher (2.5- or 2.7-fold) than that by AKR1C1 or AKR1C2. The results of the present study demonstrated that naloxone reduction by AKR1C1 and AKR1C2 was scarce, and activity in the human liver was almost entirely explained by AKR1C4 (Fig. 5E). This inconsistency may be explained by differences in evaluation methods, such as measurement of the decrease in NADPH (Ohara et al., 1995) or the production of a metabolite (in this study).
Oxcarbazepine and pentoxifylline reductions were catalyzed by multiple cytosolic isoforms (Fig. 5, F and G), but the sum of their contributions to the activities in HLC did not reach 100%. Jin and Penning (2006) demonstrated that the amino acid residues residing in the substrate binding sites of AKR1C1 and AKR1C2 are almost the same, supported by a high amino acid identity (97.8%, 316 of 323 amino acids). If the contributions of AKR1C1 to the reduction reactions of oxcarbazepine and pentoxifylline are equal to those of AKR1C2 (25% and 56%, respectively), the sum of the contributions of reductases reach 96% and 126%, respectively. Thus, the remaining contribution is likely to be AKR1C1. Although Malátková et al. (2014) reported that rCBR3 can catalyze oxcarbazepine reduction, its contribution to the human liver is negligible because its expression is 230–700-fold lower than that of AKR1C1, AKR1C2, AKR1C3, AKR1C4, or CBR1 at the mRNA level (Amai et al., 2020). Malátková et al. (2014) reported that in HLC, CLint of the S-licarbazepine formation was 3.7-fold higher than that of R-licarbazepine formation, owing to the lower Km value, and that AKR1C1, AKR1C2, and CBR1 preferentially produced S-licarbazepine, whereas AKR1C3 and AKR1C4 mainly produced R-licarbazepine. Concordantly, the present study showed that AKR1C1 and AKR1C2 had lower Km values than AKR1C3, although licarbazepine was detected as a racemate (Table 5). Nicklasson et al. (2002) reported that the area under the curve of S-lisofylline was 38-fold higher than that of R-lisofylline in six healthy volunteers taking 600 mg pentoxifylline, suggesting that the lisofylline detected in the present study might largely be in the S-form.
Bupropion, ketoprofen, and tolperisone were reduced by microsomal reductase HSD11B1 (Fig. 5, H–J). In this study, CLint of threo-dihydro bupropion formation in HLM was 60-fold higher than that of erythro-dihydro bupropion formation (Table 5), which was consistent with the results of a study by Connarn et al. (2015), who showed that the CLint of threo-dihydro bupropion formation was 9.5-fold higher than that of erythro-dihydro bupropion formation in HLM. Meyer et al. (2013) also demonstrated that threo-dihydro bupropion is preferentially produced in HLM, and recombinant HSD11B1 expressed in HEK293 cells showed the formation. For ketoprofen, it has been reported that recombinant HSD11B1 catalyzes the reduction reaction, although the kinetic parameters have not been calculated (Hult et al., 2001). Here, the role of HSD11B1 in tolperisone reduction in the human liver was clarified (Fig. 5J). HSD11B1 showed different contribution rates to bupropion, ketoprofen, and tolperisone reductions in HLM of 41%, 32%, and 104%, respectively (Fig. 5, H–J). Therefore, microsomal enzymes other than HSD11B1 may also catalyze bupropion and ketoprofen reduction reactions. Another possible reason for the low contribution rates is that the structural difference in N-glycans of HSD11B1 expressed in HLM and Sf21 cells (Fig. 4) affects the specific activity of HSD11B1 for bupropion and ketoprofen reduction. Instead of the REF value, the application of the relative activity factor (Crespi, 1995) may be more reliable for calculating the contribution of each enzyme to the reduction reactions in HLM. To make this feasible, specific substrates must be identified.
To our knowledge, this is the first study to determine the contribution of each reductase to drug reduction in the human liver and to provide a novel finding that AKR1C3 contributes to the reduction reaction of compounds at rates higher than expected.
Authorship Contributions
Participated in research design: Ichida, Fukami, A. Watanabe, K. Watanabe, Nakajima.
Conducted experiments: Ichida, Amai, Suzuki, Obuchi, Zhang.
Contributed new reagents or analytical tools: Ichida, Amai, Mishiro.
Performed data analysis: Ichida, Fukami, Amai, Takano, Nakajima.
Wrote or contributed to the writing of the manuscript: Ichida, Fukami, Nakano, Nakajima.
Footnotes
- Received July 20, 2022.
- Accepted October 11, 2022.
This work was supported by a TaNeDS Funding Program from Daiichi Sankyo (Tokyo, Japan) and in part by a Grant-in-Aid for Scientific Research (C) from the Japan Society for the Promotion of Science (16K08367).
Wataru Obuchi, Zhengyu Zhang, Akiko Watanabe, and Kengo Watanabe are employees of Daiichi Sankyo Co., Ltd.
Abbreviations
- AKR
- aldo-keto reductase
- CBR
- carbonyl reductase
- CLint
- intrinsic clearance
- DCM
- dichloromethane
- G6P
- glucose-6-phosphate
- HLC
- human liver cytosol
- HLM
- human liver microsomes
- HPLC
- high-performance liquid chromatography
- HSD11B1
- 11β-hydroxysteroid dehydrogenase type 1
- Km
- Michaelis constant
- LC-MS/MS
- liquid chromatography-tandem mass spectrometry
- MRM
- multiple reaction monitoring
- REF
- relative expression factor
- SDR
- short-chain dehydrogenase/reductase
- Vpredicted S-V, substrate concentration vs catalytic velocity
- predicted activity of human liver reductase
- Copyright © 2022 by The American Society for Pharmacology and Experimental Therapeutics