Visual Overview
Abstract
CYP3A4 is among the most abundant liver and intestinal drug-metabolizing cytochrome P450 enzymes, contributing to the metabolism of more than 30% of clinically used drugs. Therefore, interindividual variability in CYP3A4 activity is a frequent cause of reduced drug efficacy and adverse effects. In this study, we characterized wild-type CYP3A4 and 40 CYP3A4 variants, including 11 new variants, detected among 4773 Japanese individuals by assessing CYP3A4 enzymatic activities for two representative substrates (midazolam and testosterone). The reduced carbon monoxide–difference spectra of wild-type CYP3A4 and 31 CYP3A4 variants produced with our established mammalian cell expression system were determined by measuring the increase in maximum absorption at 450 nm after carbon monoxide treatment. The kinetic parameters of midazolam and testosterone hydroxylation by wild-type CYP3A4 and 29 CYP3A4 variants (Km, kcat, and catalytic efficiency) were determined, and the causes of their kinetic differences were evaluated by three-dimensional structural modeling. Our findings offer insight into the mechanism underlying interindividual differences in CYP3A4-dependent drug metabolism. Moreover, our results provide guidance for improving drug administration protocols by considering the information on CYP3A4 genetic polymorphisms.
SIGNIFICANCE STATEMENT CYP3A4 metabolizes more than 30% of clinically used drugs. Interindividual differences in drug efficacy and adverse-effect rates have been linked to ethnicity-specific differences in CYP3A4 gene variants in Asian populations, including Japanese individuals, indicating the presence of CYP3A4 polymorphisms resulting in the increased expression of loss-of-function variants. This study detected alterations in CYP3A4 activity due to amino acid substitutions by assessing the enzymatic activities of coding variants for two representative CYP3A4 substrates.
Introduction
CYP3A4 is the most abundant liver and intestinal drug-metabolizing cytochrome P450 enzyme, contributing to the metabolism of more than 30% of medications (Zanger and Schwab, 2013; Jackson et al., 2018; Niwa et al., 2020). Interindividual variability in CYP3A4 activity and expression influences drug responsiveness, efficacy, and adverse-effect rates (Lamba et al., 2002a). Besides the various factors that contribute to this variability, including CYP3A4 inducers and inhibitors and food-drug and drug-drug interactions, CYP3A4 genetic variations also affect drug responses (Racha et al., 2003; Huang et al., 2008; Elens et al., 2013; Hendriks et al., 2020). CYP3A5 genetic polymorphisms, including CYP3A5*3, have been widely investigated because of their high allele frequencies and clinical impact on the variability of CYP3A activity. CYP3A4 gene variations may be overlooked because of the higher impact of some CYP3A5 variants on activity. However, CYP3A4 gene variations should also be investigated to elucidate their influence on the combinational effects of CYP3A4 and CYP3A5 enzymatic activity and CYP3A in vivo activity (Fukushima-Uesaka et al., 2004; Lee et al., 2005; Zhou et al., 2017). For instance, several CYP3A4 variant alleles, including CYP3A4*1B (a promoter variant) and *22 (an intronic variant), reduce mRNA levels or cause splicing defects, although CYP3A4*6, *17, *20, and *26 are associated with diminished enzymatic activities by loss of function (Lamba et al., 2002a; Elens et al., 2013; Werk et al., 2014). The identification of these variants can improve patient safety and treatment efficacy; thus, increased dose-adjusted tacrolimus concentration levels are achieved in transplant patients carrying a loss-of-function genotype (CYP3A4*20) (Gómez-Bravo et al., 2018). Hence, CYP3A4 genotyping improves predictive outcomes.
The Human Cytochrome P450 Allele Nomenclature Database [https://www.pharmvar.org/gene/CYP3A4 (accessed September 21st, 2020)] contains records of CYP3A4 genetic polymorphisms (CYP3A4*2–*24, CYP3A4*26, and CYP3A4*28–*34). Moreover, several whole-genome sequencing (WGS) studies have identified CYP3A4 genetic variations, including c.1165C>T (Pro389Ser) and c.1423C>G (Leu475Val) (Apellaniz-Ruiz et al., 2015). A WGS project of the Tohoku Medical Megabank Organization, which targeted 4773 Japanese individuals, identified numerous genetic variations, including single nucleotide variations (SNVs), as shown in Table 1 [https://jmorp.megabank.tohoku.ac.jp (accessed August 30th, 2020)]. The identification of rare genetic variants, which may cause interindividual differences in drug efficacy and adverse reactions, facilitates the development of personalized medicine (Fujikura et al., 2015; Schärfe et al., 2017). CYP3A4 rare genetic variants have been implicated in high-dose simvastatin–related myopathy and affect serum voriconazole concentrations linked to adverse drug reactions (Ingelman-Sundberg et al., 2018). Although the CYP3A4 SNVs analyzed in this study are relatively rare (Table 1), their effects on CYP3A4 activity should be determined to improve treatment outcomes and manage adverse effects.
CYP3A4 allelic variants characterized in this study
Heterologous expression systems can be employed to evaluate the effects of CYP3A4 polymorphisms in vitro and complement in vivo analysis data. Performing in vivo assays is challenging because of low allele frequencies and increased patient burden. However, all CYP3A4 variant proteins should be evaluated using the same protocol and experimental conditions because methodological differences between studies may lead to inconsistent kinetic parameters for some variants (Hiratsuka, 2012). To date, characterization studies have used cDNA-based heterologous expression systems, including Escherichia coli, baculoviruses, and mammalian cell systems (Eiselt et al., 2001; Lee et al., 2005; Maekawa et al., 2010; Kumondai et al., 2018; Saito et al., 2018; Watanabe et al., 2018; Gutiérrez Rico et al., 2020). However, accurate assessments are limited by interspecies differences of post-translational modifications. Thus, mammalian cell expression systems should be preferred to evaluate the function of human CYP3A4 variants in drug metabolism (Clark and Pazdernik, 2016). Recently, an intronic CYP2D6 mutation was evaluated using a minigene expression system with transfected whole CYP2D6 sequences (<5 kb) (Zanger et al., 2020). Minigenes may aid the assessment of CYP3A4 levels. However, genomic DNA expression using minigenes, including the whole CYP3A4 gene (>20 kb), is challenging because transfection efficiency decreases with increased DNA length (Hornstein et al., 2016). Therefore, we used our heterologous expression system consisting of 293FT cells coexpressing CYP3A4, cytochrome P450 oxidoreductase (CPR), and cytochrome b5 (Kumondai et al., 2020).
Here, we functionally characterized wild-type CYP3A4 (CYP3A4.1) and 40 CYP3A4 allelic variants, including 11 novel variants identified in 4773 Japanese individuals, by using two representative CYP3A4 substrates (midazolam and testosterone). Notably, clinical and in vitro studies have assessed the functional differences and drug-drug interactions related to CYP3A4, including cytochrome P450 variants with substrate-specific enzymatic activities (Foti et al., 2010; Hiratsuka, 2016; Xiao et al., 2019). Moreover, carbon monoxide (CO)-difference spectroscopy and three-dimensional (3D) structure analysis were performed to gain insights into functional alterations caused by CYP3A4 variants.
Materials and Methods
Chemicals.
The following reagents were purchased from commercial sources: midazolam [PubChem Compound ID (CID) 4192] and flunitrazepam (Wako Pure Chemical Industries, Osaka, Japan), 1′-hydroxymidazolam (CID 107917) and 6β-hydroxytestosterone-d7 (Corning Incorporated, NY), testosterone (CID 6013) and 6β-testosterone (CID 65543) (Sigma-Aldrich, Steinheim, Germany), NADH and NADPH (Oriental Yeast, Tokyo, Japan), polyclonal anti-human CYP3A4 antibody (ab3572) (Abcam, Cambridge, UK), horseradish peroxidase–conjugated goat anti-rabbit IgG (ProteinSimple, Tokyo, Japan), dimanganese decacarbonyl (Sigma-Aldrich), and sodium cyanide and cytochrome c from horse heart (Nacalai Tesque, Kyoto, Japan). All other chemicals and reagents were of the highest quality commercially available.
Detection of CYP3A4 Sequence Alterations by Sanger Sequencing Analysis.
PCR amplification was conducted using peripheral blood leukocyte genome DNA isolated from whole blood using the Gentra Puregene Blood Kit (Qiagen, Hilden, Germany), as described previously (Nagasaki et al., 2015). Whole blood samples were obtained from Japanese subjects participating in the community-based cohort study conducted by the Tohoku Medical Megabank Organization, and written informed consent was obtained from all subjects before sample collection. Primers were used to amplify sequences containing each CYP3A4 exon (Supplemental Table 1). PCR amplification was performed with genomic DNA samples (10 ng), 2× AmpliTaq Gold 360 Master Mix (Applied Biosystems, Foster City, CA), and 0.5 µM of each primer in a total volume of 20 µl. Thermal cycling conditions included an initial denaturation step at 95°C for 10 minutes, followed by 30 cycles of denaturation at 95°C for 30 seconds, annealing at 60°C for 30 seconds, extension at 72°C for 30 seconds (for exons 2–12) or 1 minute (for exons 1 and 13), and a final extension at 72°C for 7 minutes. The PCR products were purified using columns and analyzed by Sanger sequencing utilizing the same primers for each exon as in the PCR except for the forward sequencing reaction of exon 1 (5′-CAGGCGTGGAAACACAATGGTGG-3′) and reverse sequencing reaction of exon 11 (5′-TGTGGATGACTGTAGTTTTC-3′). To determine the linkage between two SNVs identified in one subject, the respective PCR fragments were ligated into the pcDNA3.4 vector (ThermoFisher Scientific), and the ligation products were transformed into E. coli. Single colonies of the transformants were collected, and the plasmids were isolated and analyzed by Sanger sequencing. This analysis was performed in accordance with the Declaration of Helsinki, and prior written informed consent was obtained from all participants.
Construction of CYP3A4 cDNA Expression Vectors.
The wild-type CYP3A4 (CYP3A4*1) plasmid used in this study has been described in a previous report (Sakurai et al., 2005). A plasmid containing CYP3A4*1 cDNA was used as a template to generate CYP3A4 allelic variant constructs (CYP3A4*2–CYP3A4*5, CYP3A4*7–CYP3A4*21, CYP3A4*23, CYP3A4*24, Pro389Ser, Leu475Val, CYP3A4*28, CYP3A4*29, CYP3A4*31–CYP3A4*34, Thr138Ala, Met256Ile, Ile300Val, Thr323Ala, Met353Leu, Ile369Val, Ile369Asn, Ala370Val, Arg372Thr, Tyr399Ser, and Ala448Ser) using primer sets described in Supplemental Table 2 with the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) following the manufacturer’s instructions. All prepared constructs were confirmed by Sanger sequencing. Wild-type and variant CYP3A4 cDNAs were subcloned into the mammalian expression vector pcDNA3.4. A mock plasmid, including the pcDNA3.4 vector, was prepared as previously reported (Kumondai et al., 2018; Saito et al., 2018).
Expression of CYP3A4 Variants in 293FT Cells.
The 293FT cells (ThermoFisher Scientific) were cultured in Dulbecco’s modified Eagle’s medium (Nacalai Tesque) containing 10% fetal bovine serum at 37°C under 5% CO2. Cells were plated at a density of 2.2 × 106 cells per 100-mm dish; 24 hours after plating, the cells were transfected with a plasmid carrying CYP3A4 (9.6 µg), CPR (0.2 µg), and cytochrome b5 (0.2 µg) cDNA using 30 μl of 1.0 mg/ml of Polyethylenimine Max (Polysciences, Inc., Warrington, PA), according to previously described methods (Kumondai et al., 2020). After incubation for 12 hours at 37°C, 0.25 mM of 5-aminolevulinic acid hydrochloride (Nacalai Tesque) and 0.25 mM iron (II) sulfate heptahydrate (Wako) were added to the medium. After incubation for 48 hours post-transfection at 37°C, the cells were scraped from the plates, and the microsomal fractions were prepared as previously described (Kumondai et al., 2020). Protein concentrations were determined using the BCA Protein Assay Kit (ThermoFisher Scientific).
Western Blotting.
CYP3A4 levels were quantified with an immunoassay using the Wes system (ProteinSimple). Each well was loaded with 25 μg/ml microsomes. CYP3A4 was detected using a polyclonal anti-human CYP3A4 antibody (diluted 1:100) and horseradish peroxidase–conjugated goat anti-rabbit IgG. A total protein assay was performed to normalize each signal using 25 μg/ml microsomes following the manufacturer’s instructions. Recombinant human CYP3A4 Supersomes (Corning Incorporated) were used as the standard to quantify the CYP3A4 protein level.
CO-Difference Spectroscopy.
The cytochrome P450 holoprotein content was spectrophotometrically measured according to a previously reported method based on the established protocol by Omura and Sato (Kumondai et al., 2020).
Measurement of NADPH-Cytochrome c Reduction Activity and Microsomal Cytochrome b5 Content Determination.
The activity of CPR and the content of cytochrome b5 were spectrophotometrically evaluated as previously described (Kumondai et al., 2020).
Midazolam 1′-Hydroxylation.
CYP3A4 midazolam 1′-hydroxylation activity was determined using a previously described method with several modifications (El Mahjoub and Staub, 2000). The reaction mixture, a total volume of 100 μl, contained the following components: the microsomal fraction (50 μg), midazolam (1, 2, 4, 6, 8, 10, and 15 μM), and 100 mM potassium phosphate buffer (pH 7.4). After preincubation at 37°C for 3 minutes, reactions were initiated by adding 10 mM NADPH, and incubation continued at 37°C for 10 minutes. Reactions were terminated by adding 100 μl of acetonitrile containing 500 nM flunitrazepam as an internal standard. After protein removal by centrifugation at 15,400g for 10 minutes, the supernatant was injected into a liquid chromatography-tandem mass spectrometry (LC-MS/MS) system, as previously described (Kumondai et al., 2020).
Testosterone 6β-Hydroxylation Assay.
CYP3A4 testosterone 6β-hydroxylation activity was determined using a previously described method with several modifications (Nguyen et al., 2020). The reaction mixture, in a total volume of 100 μl, contained the following components: the microsomal fraction (50 μg), testosterone (5, 10, 20, 50, 100, 200, or 500 μM), and 100 mM potassium phosphate buffer (pH 7.4). After preincubation at 37°C for 3 minutes, reactions were initiated by adding 10 mM NADPH, and the incubation was continued at 37°C for 10 minutes. The reactions were terminated by adding 100 μl of acetonitrile containing 10 μM 6β-hydroxytestosterone-d7 as an internal standard. After protein removal by centrifugation at 15,400g for 10 minutes, the supernatant was injected into an LC-MS/MS system.
The 6β-hydroxytestosterone content was measured using the LC-MS/MS system in the positive ion detection mode at the electrospray ionization interface (QTRAP 6500 LC-MS/MS system; AB Sciex, MA). Separation by ultra-high-performance liquid chromatography was conducted using the Nexera Ultra-High-Performance Liquid Chromatography system (Shimadzu, Kyoto, Japan). Chromatographic separation was performed with a Kinetex C18 column (2.1 × 50 mm, 5-µg particle size; Phenomenex, Shimadzu) maintained at 40°C. Mobile phases were prepared with deionized water containing 0.1% formic acid as eluent A and acetonitrile containing 0.1% formic acid as eluent B. The flow rate was 300 µl/min. The gradient program was as follows: elution was initiated with 10% B, followed by a linear gradient to 60% B from 1 to 7 minutes, held at 100% B for 2 minutes, and then immediately returned to the initial conditions, which were maintained for 3 minutes until the end of the run. Quantification analyses using Analyst Software (Sciex) were performed in the selected reaction monitoring mode in which ion transitions from the precursor into a product ion were monitored: mass-to-charge ratio (m/z) 289.1→97.0 for testosterone (collision energy, 14 V), m/z 305.0→269.1 for 6β-hydroxytestosterone (collision energy, 12 V), and m/z 312.1→276.1 for 6β-hydroxytestosterone-d6 (collision energy, 4 V). The optimized mass spectrometry settings were as follows: entrance potential, 2.0 V; curtain gas, 25.0 psi; ion transfer voltage, 5500.0 V; temperature, 750.0°C; gas 1, 70.0 psi; gas 2, 30.0 psi; and collision gas, 12.0. Standard curves for 6β-hydroxytestosterone were constructed within the range of 100–30,000 nM using metabolite standards, with a quantification limit of 100 nM.
Data Analysis.
Kinetic data, including the Michaelis constant (Km), turnover number (kcat), and catalytic efficiency (kcat/Km) values, were determined using the Enzyme Kinetics Module of SigmaPlot 12.5 (Systat Software, Inc., Chicago, IL), a curve-fitting program based on nonlinear regression analysis. The normality of our data sets was initially assessed using the Shapiro-Wilk test. Statistical analyses for multiple comparisons were performed through variance analysis by Dunnett’s T3 test or the Kruskal-Wallis method (IBM SPSS Statistics version 22; International Business Machines, Armonk, NY). All values are expressed as the means ± S.D. of experiments performed in triplicate. Differences with P < 0.05 were considered statistically significant. All assays and samples were prepared in triplicate to permit statistical analysis.
3D Structural Modeling of CYP3A4.
The 3D structural modeling of CYP3A4 was based on the CYP3A4 X-ray structure of Sevrioukova and Poulos (2017) (Protein Data Bank code 5te8). Swiss-model was used to complement a part of the lacking structure in CYP3A4. After removing the substrate, midazolam or testosterone was docked with the CYP3A4 X-ray structure according to the CDOCKER protocol of Discovery Studio 2.5 (BIOVIA, CA). Docking iterations were conducted, taking into consideration the binding orientation and binding energies under the condition that the volume of the space was defined as 9, and the heme iron charge status was Fe3+. After replacing each substitution, structural optimization was conducted as previously described (Oda et al., 2005).
Results
We performed resequencing for the subjects carrying novel SNVs, except for a subject carrying c.1342G>T because this DNA sample was not available for further sequencing. For SNVs located within exons, WGS identified 10 novel variants; these findings were identical with the results obtained using Sanger sequencing. Two SNVs (c.389G>A and c.412A>G) carried by one subject were located in different alleles (Supplemental Fig. 1). Thus, 11 novel CYP3A4 variants, including the c.1342C>T variant, were identified among 4773 Japanese individuals with low allele frequencies (0.01%–0.05%).
CYP3A4 protein levels normalized by total protein concentrations were assessed by Western blotting with a polyclonal CYP3A4 antibody that recognized all CYP3A4 variants (Supplemental Figs. 2, A and B). The protein levels of CYP3A4.1 and the CYP3A4 variants did not differ significantly (Supplemental Fig. 2C). Moreover, the aspect ratio of CPR activity and cytochrome b5 content did not significantly vary among the analyzed CYP3A4 variants (Supplemental Fig. 3). Thus, we used single microsomes expressed per CYP3A4 variant for functional characterization.
In our study, CYP3A4 exhibited high activity with no significant solvent effects. The holoprotein content of CYP3A4 variants expressed in 293FT cells was assessed by CO-difference spectroscopy (Fig. 1). The reduced CO-difference spectra of microsomal fractions of CYP3A4.1 and 31 CYP3A4 variants (CYP3A4.2–.5, .7, .9, .10, .14–.16, .18, .19, .23, .24, Pro389Ser, Leu475Val, .28, .29, .31–.34, Thr138Ala, Met256Ile, Ile300Val, Thr323Ala, Met353Leu, Ile369Val, Ile369Asn, Ala370Val, and Ala448Ser) were determined by measuring the increase in maximum absorption at 450 nm after CO treatment. The remaining nine variants (CYP3A4.8, .11–.13, .17, 488Frameshift, .21, Arg372Thr, and Tyr399Ser) had no detectable holoprotein content based on the lack of a significant increase in their maximum absorption at 450 nm.
Representative CO-difference spectra of CYP3A4 variant proteins expressed in 293FT cells. All assays and measurements were performed in triplicate using a single microsomal preparation. Mock 1 indicates transfection with 10 μg mock plasmids. Mock 2 indicates transfection with 9.6 μg mock plasmids, 0.2 μg CPR plasmids, and 0.2 μg cytochrome b5 plasmids.
We selected midazolam and testosterone as substrates because midazolam 1′-hydroxylation and testosterone 6β-hydroxylation have long been recognized as CYP3A member-mediated reactions. First, we confirmed that the formation of 1′-hydroxymidazolam or 6β-hydroxytestosterone was linear for up to 10 minutes of incubation by using 50 μg microsomal proteins and 15 µM midazolam or 500 µM testosterone, respectively. Both metabolites were linear in the presence of 0–50 μg microsomal proteins during 10 minutes of incubation. The kinetic parameters were calculated by Michaelis-Menten curves of midazolam 1′-hydroxylation and testosterone 6β-hydroxylation for wild-type CYP3A4 and CYP3A4 variants (Fig. 2; Table 2). However, because of the insufficient metabolite production, we could not determine the kinetic parameters of midazolam 1′-hydroxylation for 11 variants (CYP3A4.8, .11–.13, .17, 488Frameshift, .21, Ile369Asn, Ala370Val, Arg372Thr, and Tyr399Ser). We found that 15 variants had significantly lower catalytic efficiency values (P < 0.05) than CYP3A4.1, whereas there were no significant differences in Km and kcat values among these CYP3A4 variants. Conversely, the Km values of testosterone 6β-hydroxylation were significantly higher for CYP3A4.7, .16, .33, and Thr138Ala than for CYP3A4.1 (P < 0.05). Moreover, the kcat and catalytic efficiency values were significantly lower for CYP3A4.12 and Leu475Val than for CYP3A4.1 (P < 0.05). Insufficient metabolite production prevented us from determining the kinetic values for nine variants (CYP3A4.8, .11, .13, .17, 488Frameshift, .21, Ile369Asn, Ala370Val, and Arg372Thr). Lastly, we calculated the correlation between the midazolam 1′-hydroxylation and the testosterone 6β-hydroxylation catalytic efficiency values. As shown in Fig. 3, there was a significant correlation between these values among the different variants (R2 = 0.724, P < 0.001).
Michaelis-Menten curves for CYP3A4 variants. Determined kinetic parameters (Km, kcat, and catalytic efficiency) of midazolam 1′-hydroxylation and testosterone 6β-hydroxylation. All assays and measurements were performed in triplicate using a single microsomal preparation.
Kinetic parameters of midazolam 1′-hydroxylation and testosterone 6β-hydroxylation by microsomes from 293FT cells expressing wild-type CYP3A4 and variant CYP3A4 proteins
The data represent the means ± S.D. of the three independently performed catalytic assays. All assays and measurements were performed in triplicate using a single microsomal preparation. Mock 1 indicates transfection with 10 μg mock plasmids. Mock 2 indicates transfection with 9.6 μg mock plasmids, 0.2 μg CPR plasmids, and 0.2 μg cytochrome b5 plasmids.
Correlation between the catalytic efficiency ratios for midazolam 1′-hydroxylation and testosterone 6β-hydroxylation among CYP3A4 variants. Midazolam 1′-hydroxylation catalytic efficiency ratios are plotted on the horizontal axis, and testosterone 6β-hydroxylation catalytic efficiency ratios are plotted on the vertical axis. The numbers correspond to known CYP3A4 variants, and the amino acid substitutions indicate the newly identified variants.
We used 3D structural modeling to assess the molecular properties of CYP3A4 variants and their interactions with midazolam. As shown in Fig. 4, A and B, the residue substitution in CYP3A4.8 and CYP3A4.12 abolished the interaction with heme. The midazolam docking analysis indicated that the Thr138Ala substitution variant caused a hydrophobic interaction between Arg130 and Ser134 via the adjacently formed hydrophobic interaction between Ala138 and Ser134 (Fig. 4C). In CYP3A4.17, the substitution Ser189 located at the F-G loop lacked a hydrophobic interaction with Leu249 (Fig. 4D). CYP3A4.21 harbored its substitution in the I-helix, which formed a hydrophobic interaction between Cys319 and Val489 (Fig. 4E). Additionally, the substitution Ala448Ser abolished numerous hydrophobic interactions with heme (Fig. 4F). The substitution Tyr399Ser caused the formation of a hydrogen bond with Met371 (Fig. 4G). We further analyzed Ile369 and Val370, as shown in Fig. 4, H–M. The Ile369Val substitution had an additional hydrophobic interaction with midazolam and heme, affecting the distance between heme and the active site (Fig. 4H). Furthermore, differences in the midazolam and heme interactions due to the amino acid substitutions (Fig. 4, I, K, and L), as well as substrate-specific differences in the heme-surrounding interactions (Fig. 4, H, J, K, and M), were observed.
Diagram pairs showing the partial crystal structure of CYP3A4.1 (left image) and CYP3A4 variants (right image) for CYP3A4.8 (A), CYP3A4.12 (B), Thr138Ala (C), CYP3A4.17 (D), CYP3A4.21 (E), Ala448Ser (F), Tyr399Ser (G), Ile369Val coordinated with midazolam (H), Ile369Asn (I), Ile369Val coordinated with testosterone (J), CYP3A4.33 coordinated with midazolam (K), Ala370Val (L), and CYP3A4.33 coordinated with testosterone (M). Pink line, hydrophobic interactions. Green line, conventional hydrogen bonds. Gray line, carbon-hydrogen bonds. MDZ, midazolam; TST, testosterone.
Discussion
Several CYP3A4 genetic variations may contribute to interindividual differences in drug efficacy and adverse-effect rates (Lamba et al., 2002a). Several studies have identified CYP3A4 variants and characterized the associated in vitro enzyme kinetics, including previous studies of CYP3A4 variants among the Han Chinese population using numerous substrates (Supplemental Table 3) (Fang et al., 2017; Xu et al., 2018; Li et al., 2019a,b; Lin et al., 2019a,b; Yang et al., 2019; Zhou et al., 2019a; Chen et al., 2020). However, the underlying causes for functional differences produced by the amino acid alterations are not fully understood and require further research. This study characterized 40 CYP3A4 variants, including 11 novel variants identified among 4773 Japanese individuals, by assessing CO-difference spectra, enzymatic activities, and 3D structural modeling analysis.
A single individual had two allelic variants (CYP3A4.8 and Thr138Ala, Supplemental Fig. 1), each displaying significantly decreased CYP3A4 activity for midazolam and testosterone. These substitutions were located near the vital CPR-cytochrome P450 electron transfer site, where they could affect the enzymatic activities (Hasemann et al., 1995; Bridges et al., 1998). Accordingly, an increase in the maximum absorption at 450 nm after CO treatment was observed for the Thr138Ala variant (Fig. 1) as a result of interactions involving Ala138 that caused hydrogen bonds between Arg130 and Ser134, which could affect electron transfer because Arg130 located in the C-D loop plays an essential role in electrostatic interactions (Fig. 4C). Interestingly, CYP3A4.8 did not have an absorption peak at 450 nm after CO treatment (Fig. 1), reflecting a failed heme incorporation and resulting in zero CYP3A4.8 holoprotein content that could be based on disrupted interactions with heme (Fig. 4A). Although it may be an infrequent case, carriers of two allelic variants that are predicted to be poor metabolizers may be at an increased risk of adverse drug reactions when treated with drugs metabolized by CYP3A4.
The midazolam and testosterone hydroxylation activities were slightly decreased or absent in 11 variants (CYP3A4.8, .11–.13, .17, 488Frameshift, .21, Ile369Asn, Ala370Val, Arg372Thr, and Tyr399Ser). Excluding Ile369Asn and Ala370Val, the nine remaining variants did not exhibit absorption peaks at 450 nm after CO treatment (Fig. 1), indicating failed heme incorporation. These findings suggest that the remaining nine CYP3A4 variants lacked quantifiable enzymatic activity. Similarly, the zero holoprotein content for variants CYP3A4.8 and CYP3A4.12 may be explained by disrupted interactions with heme (Fig. 4, A and B). The lack of activity exhibited by 488Frameshift was due to extensive changes in its molecular conformation caused by the frameshift mutation. The substitutions in CYP3A4.11 and .13, which were located in the conserved glutamine-arginine-arginine motif and K region critical for maintaining the structure among cytochrome P450 isoforms, could affect holoprotein stability, as previously reported (Eiselt et al., 2001). The substitution in CYP3A4.17 located in the E-helix was associated with slightly decreased enzymatic activity. Ser189 lacked a hydrophobic interaction with Leu249 (Fig. 4D), resulting in conformational changes in the active site cavity around the F-G loop, affecting the main active site channel as well as holoprotein stability (Benkaidali et al., 2019). CYP3A4.21 harbored a substitution in the I-helix, which may affect the stability and flexibility of the β-sheet by forming a hydrophobic interaction between Cys319 and Val489 (Fig. 4E). However, other substitutions, such as Ala448Ser, which apparently affected the hydrophobic interactions with heme, did not result in a complete lack of an absorption peak at 450 nm (Fig. 4G); however, the Ala448Ser substitution decreased the enzymatic activities. Thus, it is important to evaluate CO-difference spectroscopy results in combination with the findings of 3D structural modeling to improve our understanding of the consequential impact of amino acid substitutions.
In addition, a novel substrate recognition site (SRS) map for mammalian cytochrome P450 enzymes was developed based on the 3D crystal structures and docking of 868 substrates (Zawaira et al., 2011). Among 40 variants characterized in this study, the following variants were classified as listed regions based on the SRS map: CYP3A4.4 and .29, SRS-1; CYP3A4.2 and .5, SRS-2; CYP3A4.18 and Ile300Val, SRS-4; CYP3A4.12, .33, Ile369Val, Ile369Asn, Ala370Val, and Arg372Thr, SRS-5; and Leu475Val, SRS-6. For instance, the substitution Ile369Val appeared to directly affect the enzyme’s interactions with midazolam, resulting in an increased distance between heme and the active site (Fig. 4H). Substitutions found in SRS-5, including Ile369 and Ala370, had the most significant effect on CYP3A4 activity (Sevrioukova, 2017; Sevrioukova and Poulos, 2017). Interestingly, interactions between the heme group and assayed substrates differed depending on amino acid substitutions, including Ile369Val, Ile369Asn, CYP3A4.33 (Ala370Ser), and Ala370Val, which ultimately could cause the observed differences in enzymatic activities (Fig. 4, H, I, K, and L). Interestingly, variants carrying amino acid substitutions around these sites showed a weaker correlation with the decreasing catalytic efficiencies for midazolam and testosterone than CYP3A4.1, as shown in Fig. 3. The differences in the interactions between amino acids (Ile369 and Ala370) and heme may play a critical role in the enzymatic activity (Fig. 4, H, J, K, and M). The Tyr399Ser substitution also affected the SRS-5 region by forming a hydrogen bond between Ser399 and Met371 (Fig. 4G). Thus, amino acid substitutions in SRS regions interfered with CYP3A4 activities, but the extent of the resulting differences varied among substrates. Although our study detected a correlation between midazolam and testosterone activities, in the context of previous data (Supplemental Table 3), the degree of CYP3A4 activity reduction is apparently highly substrate-specific.
The effects of CYP3A4*16 and CYP3A4*18 on several clinically used drugs have been evaluated because of their relatively high allele frequencies (1.4%–5.0% and 1.3%–10%, respectively) (Lamba et al., 2002b; Yamamoto et al., 2003; Fukushima-Uesaka et al., 2004). However, their enzymatic activities also appear to be substrate-dependent (Supplemental Table 3). In addition to SNVs in the coding region, intron variants, such as CYP3A4*16B and *18B, should also be evaluated in vivo because several promoter and intronic variants, including CYP3A4*1B and *22, exhibit decreased activities due to reduced mRNA expression (Murayama et al., 2002; Elens et al., 2013). Notably, CYP3A5*3 and other CYP3A5 allelic variants play important clinical roles because the effects of CYP3A4 and CYP3A5 on the metabolism of numerous drugs overlap (Fukushima-Uesaka et al., 2004). Moreover, microRNA-27b was identified as a potential regulator of CYP3A4 mRNA level, underscoring the fact that other regulatory factors, and not only CYP3A4 allelic variants, affect CYP3A4 mRNA and protein levels (Pan et al., 2009; Ekström et al., 2015). Thus, future studies should establish a dosing algorithm based on interindividual differences in the in vivo activity of CYP3A.
CYP3A4 can contribute to the catalysis of several sites in the same compound. For example, midazolam is metabolized by CYP3A4 to the main metabolite 1′-hydroxymidazolam and the secondary metabolite 4-hydroxymidazolam (Christensen et al., 2009). According to a previous report, the enzymatic activities of CYP3A4.16 and .18 for these two midazolam metabolites decreased to similar levels (Maekawa et al., 2010). Moreover, even with the amino acid sequence differences between CYP3A4 and CYP3A5 (83% homology), 4-hydroxymidazolam formation was several times lower than 1′-hydroxymidazolam, suggesting that single amino acid substitutions in CYP3A4 may not dramatically influence the catalysis of different primary products dependent on changing mutation sites (Niwa et al., 2014). However, a detailed study encompassing all midazolam metabolites is needed to fully understand the differences in the metabolic profiles associated with CYP3A4 genetic polymorphisms.
Cytochrome P450 variant characterization by assessing the enzymatic activities for all metabolites and using all possible substrates may prove essential to evaluate the impact caused by amino acid substitutions and to improve treatment outcomes. However, previously developed computational calculations have been successful in predicting the activities of over 90% of representative cytochrome P450 variants (Zhou et al., 2019b). Hence, using computational prediction to assess the effect of amino acid substitutions on enzymatic cytochrome P450 activities may be an acceptable surrogate approach, considering the challenging nature of in vitro and in vivo analyses.
In summary, we determined the in vitro metabolic activities of wild-type CYP3A4 and 40 CYP3A4 variant proteins using our recently established heterologous expression system. These data reveal the functional effects caused by CYP3A4 genetic polymorphisms, including 11 novel variants identified among 4773 Japanese individuals, and have been generated using a single microsomal protein preparation per CYP3A4 variant to reveal their qualitative features. We recognize the limitations associated with a single biologic replicate, and additional biologic replicates would provide greater confidence in estimates of intrinsic clearance for use in future applications. Reduced activity or inactive variants caused by CYP3A4 allelic variations located in the coding region may be the underlying cause of an altered therapeutic response leading to poor treatment outcomes. However, our assay that uses a cDNA-based expression system cannot evaluate the influence of splicing variants. Therefore, further studies including additional biologic replicates as well as the use of minigenes to assess the potential functional consequences of variants affecting splice sites on catalytic activity are required prior to further downstream applications such as in vitro-in vivo extrapolation. In the future, CYP3A4 genetic polymorphisms should be used in combination with other factors affecting the in vivo activity of CYP3A4, such as CYP3A4 inducers and inhibitors and CYP3A5 activity, to aid the implementation of personalized medicine, individualized dosing regimens, and predictive risk panels. Our current findings should provide guidance in advancing these efforts.
Acknowledgments
We thank the Biomedical Research Core at the Tohoku University Graduate School of Medicine for their technical support. We would like to thank Mizuguchi (Graduate School of Pharmaceutical Sciences, Osaka University) for gifting the plasmid vector. We would like to thank Editage (www.editage.com) for English language editing.
Authorship Contributions
Participated in research design: Kumondai, Hiratsuka.
Conducted experiments: Kumondai, Gutiérrez Rico, Hishinuma, Ueda, Saito, Abe, Maekawa.
Contributed new reagents or analytic tools: Saigusa, Kinoshita, Maekawa, Mano, Hirasawa, Hiratsuka.
Performed data analysis: Kumondai, Saito, Tadaka, Nakayoshi, Oda, Hiratsuka.
Wrote or contributed to the writing of the manuscript: Kumondai, Gutiérrez Rico, Hiratsuka.
Footnotes
- Received September 28, 2020.
- Accepted December 24, 2020.
M.H. was supported by grants from Japan Agency for Medical Research and Development (AMED) [Grant 20kk0305009], Takahashi Industrial and Economic Research Foundation, and Smoking Research Foundation. M.K. was supported by Japan Society for the Promotion of Science [Grant 19J10744] and the Pharmaceutical Society of Japan [Grant N-170603]. S.S., K.K., S.T., and D.S. were supported by grants from AMED [Grant JP20km0105001 and JP20km0105002]. This research was also supported in part by the Tohoku Medical Megabank Project: Promoting Public Utilization of Advanced Research Infrastructure, and the Sharing and Administrative Network for Research Equipment funded by the Ministry of Education, Culture, Sports, Science and Technology (MEXT). K.K. and S.T. were supported by grants for the Facilitation of R&D Platform for AMED Genome Medicine Support from AMED (Grant Number JP20km0405001).
The authors declare no conflict of interest.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- CO
- carbon monoxide
- CPR
- cytochrome P450 oxidoreductase
- 3D
- three-dimensional
- LC-MS/MS
- liquid chromatography-tandem mass spectrometry
- m/z
- mass-to-charge ratio
- PCR
- polymerase chain reaction
- SNV
- single nucleotide variation
- SRS
- substrate recognition site
- WGS
- whole-genome sequencing
- Copyright © 2021 by The American Society for Pharmacology and Experimental Therapeutics
