Abstract
Despite several studies suggesting that CYP3A5 expression can influence the extent of hepatic CYP3A-mediated inhibition, a systematic in vitro-in vivo evaluation of this potential clinically important issue has not been reported. Using representative probes from two distinct CYP3A substrate subgroups (midazolam, erythromycin), the inhibitory potency of fluconazole was evaluated in pooled human liver microsomes (HLM) with a low or high specific CYP3A5 content, in recombinant CYP3A enzymes (rCYP3A), and in healthy volunteers lacking or carrying the CYP3A5*1 allele. Fluconazole was a slightly more potent inhibitor of CYP3A activity in CYP3A5-HLM than in CYP3A5+ HLM with midazolam (Ki of 15 and 25 μM, respectively) but not with erythromycin (IC50 of 70 and 54 μM, respectively). In comparison, fluconazole was a much more potent inhibitor of rCYP3A4 than rCYP3A5 with both midazolam (Ki of 7.7 and 54 μM, respectively) and erythromycin (IC50 of 100 and 350 μM, respectively). As predicted from HLM, with i.v. midazolam, the average (± S.D.) in vivo Ki (Ki,iv) was significantly higher in CYP3A5*1 carriers (24 ± 17 and 17 ± 8 μM for homozygous and heterozygous groups, respectively) than in noncarriers (13 ± 6 μM) (p = 0.02). With the erythromycin breath test, the average Ki,iv was not different between homozygous CYP3A5*1 carriers (30 ± 12 μM) and noncarriers (58 ± 53 μM). In conclusion, the effect of CYP3A5 on hepatic CYP3A-mediated inhibitory drug-drug interactions is substrate-dependent, and HLM, rather than rCYP3A, are the preferred in vitro system for predicting these interactions in vivo.
Adverse interactions between two or more medications have been a longstanding problem in clinical practice. Such interactions frequently result from one drug impairing the elimination of another, which can lead to increased systemic concentrations of the affected drug and the potential for an adverse or toxic reaction. The cytochrome P450 3A (CYP3A) subfamily, consisting primarily of CYP3A4 and CYP3A5 in adults, is believed to participate in the metabolism of more than half of therapeutic agents that undergo oxidation (Wilkinson, 2005). Taken together, inhibition of CYP3A-mediated metabolism is a common mechanism underlying numerous drug-drug interactions. In addition, the extent of inhibition of CYP3A-mediated drug clearance by a comedicant often varies between individuals, making it difficult to predict the magnitude and severity of the interaction.
Both CYP3A4 and CYP3A5 are expressed primarily in the liver and small intestine. Although more than 30 allelic variants in the CYP3A4 gene have been identified, low-variant frequencies, often combined with a lack of functional consequence, indicate a limited contribution by these variants to the large interindividual variation observed in CYP3A4 expression (Lamba et al., 2002; Wojnowski, 2004). In contrast, CYP3A5 expression is clearly polymorphic, and the frequency of robust expression of a functionally significant protein results from genetic mutations that vary among different ethnic groups (Lamba et al., 2002; Xie et al., 2004). For example, CYP3A5 protein has been detected readily in liver specimens obtained from 10 to 40% of European and North American Caucasians, 33% of Japanese individuals, and 55% of African Americans (Lamba et al., 2002). Although CYP3A4 generally represents the dominant CYP3A enzyme in the liver, CYP3A5 can represent more than 50% of total hepatic CYP3A content in some individuals (Kuehl et al., 2001; Lin et al., 2002).
Inheritance of at least one copy of the reference CYP3A5 allele (CYP3A5*1) confers CYP3A5 expression, whereas inheritance of two copies of a variant allele (CYP3A5*3) confers undetectable or very low CYP3A5 expression (Kuehl et al., 2001). Consistent with the varying frequencies of CYP3A5 expression, the frequency of CYP3A5*3 ranges from 85 to 95% in Caucasians, from 65 to 85% in Chinese and Japanese individuals, and from 27 to 55% in African Americans (Lamba et al., 2002; Xie et al., 2004). In addition to CYP3A5*3, two other variants, CYP3A5*6 and CYP3A5*7, are associated with reduced CYP3A5 expression and are more common in individuals of African origin (2–17%) compared with Caucasians and Asians (<2%) (Lamba et al., 2002; Xie et al., 2004). Accordingly, all three variants should be considered in any genotyping-based prediction of CYP3A5 function in populations of African origin (Hustert et al., 2001).
CYP3A4 and CYP3A5 share most of the same substrates (Huang et al., 2004), and in vitro investigations have indicated that CYP3A5 can be less susceptible to inhibition than CYP3A4. Early examples involved the antifungal agents ketoconazole and fluconazole (Gibbs et al., 1999). Using recombinant CYP3A enzymes and midazolam as the substrate, the Ki of ketoconazole toward CYP3A5 was 4-fold greater than that toward CYP3A4 (110 versus 27 nM); an even greater difference was observed for fluconazole (85 versus 9 μM). A weaker inhibitory potency toward CYP3A5 compared with CYP3A4 was reported subsequently for other clinically relevant inhibitors, including diltiazem, nicardipine, and mifepristone (Jones et al., 1999; McConn et al., 2004). These observations suggest that genotypic differences in susceptibility to CYP3A inhibition could explain in part the large interindividual variation in the extent of inhibition of drug clearance by drugs and other xenobiotics.
A caveat to the aforementioned inhibitory differences between the CYP3A enzymes is that only midazolam was used as the CYP3A probe. In vitro evidence suggests that the extent of inhibition of CYP3A activity varies with substrate (Kenworthy et al., 1999; Stresser et al., 2000; Wang et al., 2000). This variation is explained in part, at least for CYP3A4, by the existence of multiple binding domains within the enzyme active site (Hosea et al., 2000; Schrag and Wienkers, 2001). Midazolam, testosterone, and nifedipine appear to bind to different domains and are representative of three distinct substrate subgroups (Kenworthy et al., 1999). The use of at least two representative CYP3A substrates has been recommended for both the in vitro and in vivo evaluation of the inhibitory potential of drugs and other xenobiotics (Kenworthy et al., 1999; Yuan et al., 2002).
To our knowledge, a systematic in vitro-in vivo evaluation regarding inhibitor potency toward different CYP3A substrates in humans with different CYP3A5 genotypes has not been evaluated. Accordingly, the objectives of the current work were to determine the inhibitory potency of fluconazole toward two distinct probe substrates using pooled human liver microsomes (HLM) with known CYP3A5 protein content and in healthy volunteers with known CYP3A5 genotype. Midazolam and erythromycin were selected as the substrates because both can be given i.v. and safely to humans. Fluconazole was selected as the inhibitor based on the 9-fold difference in inhibitory potency between recombinant CYP3A enzymes (Gibbs et al., 1999), minimal plasma protein binding, metabolism representing a minor route of elimination, a lack of effect on the efflux transporter P-glycoprotein (P-gp) (Venkatakrishnan et al., 2000), and a long elimination half-life in vivo (∼30 h), which would provide a relatively stable concentration during the period of substrate elimination.
Materials and Methods
Materials and Chemicals. Western blotting reagents and materials (SDS, acrylamide/bis, ammonium persulfate, N,N,N′,N′-tetramethylethylenediamine, polyvinylidene difluoride membranes, and enhanced chemiluminescence reagents) were purchased from sources as described previously (Paine et al., 2005). Baculovirus insect cell-expressed human CYP3A4 and CYP3A5 (co-expressed with cytochrome P450 reductase but not cytochrome b5), the anti-CYP3A4 and -CYP3A5 antibodies (WB-3A4 and -3A5, respectively), and 1′-hydroxymidazolam (1′-OH MDZ) were purchased from BD Gentest (Woburn, MA). A panel of 10 African American liver microsomal preparations (race/ethnicity supplied by the manufacturer) was purchased from Xenotech, LLC (Lenexa, KS). The secondary antibody, goat anti-rabbit horseradish peroxidase-conjugated IgG, was purchased from Zymed Laboratories (South San Francisco, CA). Midazolam maleate, alprazolam, erythromycin, and NADPH were purchased from Sigma-Aldrich (St. Louis, MO). Fluconazole was purchased from MP Biomedicals (Irvine, CA). 15N3-Midazolam was a gift from Hoffman-La Roche (Nutley, NJ). [14C N-Methyl]erythromycin (specific activity, 48.8 mCi/mmol; radiochemical purity, 98.8%) was purchased from PerkinElmer (Shelton, CT). [3H]Formaldehyde (specific activity, 20 Ci/mmol; radiochemical purity, 99%) was purchased from American Radiolabeled Chemicals (St. Louis, MO). All the other chemicals were of electrophoresis or analytical grade where appropriate.
Characterization of HLM for CYP3A Protein Content and Susceptibility to Inhibition.Western blot analysis for CYP3A content. A panel of liver microsomes obtained from African American donors (n = 10) was characterized as described previously (Lyke et al., 2003) to generate two pools of HLM that contained either CYP3A4 or CYP3A5 as the major CYP3A enzyme. Briefly, microsomes were diluted in sample buffer as described (Paine et al., 2005) to yield a final concentration of 5 to 20 μg/60 μl. Reference standards were prepared similarly using the recombinant CYP3A enzymes (rCYP3A). The diluted microsomal preparations and reference standards were boiled (3 min), loaded onto 0.1% SDS/9% polyacrylamide gels (14 × 16 cm), and the proteins were separated by electrophoresis as described previously (Paine et al., 2005). The proteins were transferred overnight to polyvinylidene difluoride membranes at 4°C, after which the membranes were placed in blocking buffer [5% nonfat dry milk in phosphate-buffered saline containing 0.3% Tween 20 (PBS-T)] at room temperature. After 1 h, the blots were rinsed in PBS-T and then incubated with the anti-CYP3A4 (1:500) or anti-CYP3A5 (1:3000) antibody. After 1 (anti-CYP3A4) or 2 (anti-CYP3A5) h, the membranes were rinsed in PBS-T, incubated with the secondary antibody (1:500 × 1 h for CYP3A4; 1:3000 × 2 h for CYP3A5), and rinsed again. The proteins of interest were visualized by enhanced chemiluminescence using the Chemi-Doc imaging system (Bio-Rad, Hercules, CA). Integrated optical densities were obtained using the Bio-Rad software program Quantity One (version 4.1). Calibration curves were generated by plotting the integrated optical densities of the reference standards against the mass of rCYP3A loaded. The amount of CYP3A enzyme per well was calculated relative to the calibration curve. Specific content was calculated by dividing the amount of enzyme per well by the amount of total microsomal protein loaded. The three preparations with undetectable CYP3A5 were pooled and designated as “CYP3A5 nonexpressors.” The three preparations with the highest percentage of CYP3A5 content (with respect to total CYP3A content, CYP3A4 + CYP3A5) were pooled and designated as “CYP3A5 expressors.” Each pool was then analyzed for CYP3A4 and CYP3A5 content in the same manner as described for the individual preparations.
Inhibitory potency of fluconazole toward CYP3A catalytic activity. The inhibitory potency of fluconazole toward CYP3A activity (midazolam 1′-hydroxylation or erythromycin N-demethylation) was evaluated using the CYP3A5- and CYP3A5+ HLM. For comparison, similar experiments were conducted with rCYP3A. Midazolam and fluconazole were dissolved in methanol to yield concentrated solutions ranging from 1 to 8 mM and 0.5 to 40 mM, respectively. Erythromycin was dissolved in methanol to yield a 10 mM solution. To increase the sensitivity of the radiometric assay for determination of erythromycin N-demethylase activity (Lyke et al., 2003), [14C]erythromycin was purified to remove trace amounts of [14C]formaldehyde. In brief, a volume of the stock solution (in 100% ethanol) was diluted 1:10 in water, after which a saturated solution of sodium carbonate (0.5% v/v) was added. The diluted solution was then loaded onto a pre-equilibrated C18 Bond Elute cartridge (Varian Inc., Palo Alto, CA), and the impurities were eluted with 10% methanol. The cartridge was allowed to dry, and the purified [14C]erythromycin was eluted with 100% methanol. The volume of purified [14C]erythromycin was then adjusted with 100% methanol to achieve a radioactivity concentration of ∼2 μCi/ml.
Ki of fluconazole using midazolam as substrate. Reaction mixtures consisted of 0.05 mg/ml microsomal protein or 20 pmol/ml rCYP3A, midazolam (1–8 μM), fluconazole (0–80 or 0–400 μM), and potassium phosphate buffer (0.1 M, pH 7.4). After a 5-min equilibration period at 37°C, the reactions were initiated with NADPH (final concentration, 1 mM) to yield a final volume of 0.5 ml. Reactions were quenched after 2 (microsomes) or 4 (rCYP3A) min with 1 ml of ice-cold acetonitrile. The resulting mixtures were spiked with internal standard (alprazolam, 30 pmol), vortex-mixed, and stored at -20°C pending analysis for 1′-OH MDZ by liquid chromatography/mass spectrometry as described previously (Paine et al., 2004). The amount of 1′-OH MDZ formed was linear with respect to the incubation time and amount of enzyme source. Initial estimates of the apparent Km and Vmax were derived from Eadie-Hofstee plots of the substrate concentration-velocity data in the absence of inhibitor. Initial estimates of the apparent Ki were derived from Dixon plots of the inhibitor concentration-velocity data. Kinetic parameters (Km, Vmax, Ki) were obtained from untransformed data by nonlinear least-squares regression using WinNonlin (version 4.1, Pharsight, Mountain View, CA). The appropriateness of the model (competitive, noncompetitive, or mixed-type inhibition for a unienzyme system) was assessed from visual inspection of the observed versus predicted data, randomness of the residuals, Akaike information criteria, and standard errors of the parameter estimates. Apparent intrinsic clearance (Clint) was calculated as the ratio of Vmax to Km.
IC50 of fluconazole using 14C-erythromycin as substrate. For technical and cost reasons, the IC50, rather than Ki, was determined. The highly sensitive radiometric assay for the determination of erythromycin N-demethylase activity has been described previously (Lyke et al., 2003) and was modified from a method by Riley and Howbrook (1997). The amount of [14C]erythromycin required for an experiment (i.e., 0.2 μCi/incubation) was removed from the diluted purified stock solution and combined with an appropriate volume of the cold erythromycin solution (10 mM) to achieve a total erythromycin content of 10 nmol/incubation. This batch substrate mix was evaporated to dryness under nitrogen and solubilized in 1 μl methanol/incubation. Potassium phosphate buffer (0.1 M, pH 7.4) was then added to achieve a concentration of 10 nmol erythromycin/50 μl, and the mixture was chilled on ice before adding to the reaction tubes. Reaction mixtures consisted of 0.125 mg/ml microsomal protein or 50 pmol/ml rCYP3A, erythromycin (50 μM, which approximates the Km), fluconazole (0–400 μM), and potassium phosphate buffer. After a 5-min equilibration period at 37°C, the reactions were initiated with NADPH (final concentration, 1 mM) to yield a final volume of 0.2 ml. Reactions were quenched after 10 min with 50 μl of 10% trichloroacetic acid and placed on ice. The mixtures were spiked with 50 μl of internal standard (consisting of 0.05 μCi/ml [3H]formaldehyde with 0.06 M cold formaldehyde and 0.5 mM cold erythromycin as trace carriers) and centrifuged, and the supernatants were transferred to pre-equilibrated Supelclean Envi-Carb solid-phase extraction tubes (3 ml, 0.25 g) (Sigma-Aldrich). The tubes were eluted with 2 volumes of water (2 × 0.5 ml), which was transferred to scintillation vials containing 20 ml of scintillation mixture. The vials were placed in a PerkinElmer TRI-carb 2900T liquid scintillation analyzer, and the radioactivity was counted using a dual 3H/14C program (QuantaSmart 1.1 software, PerkinElmer). The amount of [14C]formaldehyde formed was linear with respect to the incubation time and amount of enzyme source. Percent control activity was determined as the ratio of the amount of [14C]formaldehyde formed in the presence to that in the absence of fluconazole. Initial estimates of the apparent IC50 were derived from linear regression of the natural logarithm of [fluconazole] versus percent control activity data. The apparent IC50 was obtained from untransformed data by nonlinear least-squares regression using WinNonlin.
Human Volunteer Study.Subjects. All the subjects provided written, informed consent before participating in the clinical protocol, which was approved by the Institutional Review Board and the Clinical Research Advisory Committee at the University of North Carolina. To maximize the probability of identifying homozygous CYP3A5*1 carriers, only African American subjects were recruited. Healthy unrelated volunteers, self-identified as African American or black, provided mouthwash samples, from which genomic DNA was isolated using the QIAamp DNA blood mini kit (QIAGEN Inc., Valencia, CA). Roughly equal numbers of volunteers genotyped as CYP3A5*1/*1 (n = 6), “CYP3A5*1/*X”(n = 7), or “CYP3A5*X/*X”(n = 6) were enrolled in the pharmacokinetic study; in this context, “X” represents the CYP3A5*3, *6, or *7 allele. Individuals carrying two CYP3A5*1 alleles (CYP3A5*1/*1)orone CYP3A5*1 allele (CYP3A5*1/*X) were designated as CYP3A5 expressors. Individuals with two copies of the same defective allele (CYP3A5*X/*X) were designated as CYP3A5 nonexpressors. Individuals with two different defective alleles (CYP3A5*3/*6, CYP3A5*3/*7,or CYP3A5*6/*7) were excluded to avoid potential inference about the phenotype.
The participants, 13 women and 6 men, ranged in age from 18 to 45 years (average ± S.D., 24 ± 8 years) and in weight from 57 to 97 kg (average ± S.D., 76 ± 12 kg). Before enrollment, each subject presented for a screening visit that consisted of a medical history, physical examination, vital signs, and laboratory tests that included complete blood count and blood chemistries (blood urea nitrogen, serum creatinine, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, total bilirubin). All the women underwent a serum pregnancy test. None of the subjects was taking medications, prescription or nonprescription (including herbal products), known to alter CYP3A activity. All the subjects were instructed to refrain from consuming grapefruit-containing products beginning at least 1 week before and during the course of the study, as well as to refrain from caffeinated and alcoholic beverages the evening before admission to the University of North Carolina General Clinical Research Center.
Determination of CYP3A5 genotype. CYP3A5 genotype was determined based on sequencing assays for the CYP3A5*3 (Kuehl et al., 2001), CYP3A5*6 (14690G>A), and CYP3A5*7 (27131-32insT) variant alleles. Polymerase chain reaction (PCR) for CYP3A5*6 genotyping was performed using forward (5′-TTGCTGCATGTATAGTGGAAGG-3′) and reverse (5′-GTGTGAGGGCTCTAGATTGACA-3′) primers at concentrations of 400 nM, 50 ng of template DNA, and Ready-to-Go bead (puReTaq Ready-to-Go PCR beads, GE Healthcare, Little Chalfont, Buckinghamshire, UK) in a final volume of 25 μl, producing a 419-base pair fragment of the CYP3A5 gene. A 461-base pair amplicon was obtained by identical conditions for CYP3A5*7 genotyping using forward (5′-CTCCTCCACACATCTCAGTAGGT-3′) and reverse (5′-CATTTCCCTGGAGACTTGTACC-3′) primers. PCR amplification consisted of the following: after an initial denaturing step at 95°C for 5 min, amplification was performed for 35 cycles of denaturation (95°C for 30 s), annealing (55°C for 30 s), and extension (72°C for 60 s), followed by a final extension at 72°C for 5 min. PCR products were spin column-purified to remove unincorporated nucleotides and primers using the QIAquick PCR Purification Kit (QIAGEN Inc.) and sequenced for the forward and reverse direction on an ABI Prism 377Xl DNA Sequencer (Applied Biosystems, Foster City, CA) with the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit (PerkinElmer).
Study design. The clinical protocol consisted of two study phases scheduled on consecutive days. Subjects were admitted to the General Clinical Research Center the evening before the first phase. All the women underwent a repeat serum pregnancy test before initiation of other study procedures. All the subjects underwent an overnight fast before each study phase. On the morning of the first phase (day 1), an indwelling heparin-lock catheter was placed into an antecubital vein for serial blood collections and for administration of the erythromycin breath test (ERMBT) as described (Paine et al., 2002). In brief, 3 μCi (0.07 μmol) of [14C N-methyl]erythromycin (Metabolic Solutions, Nashua, NH) was diluted in 5 ml of 5% dextrose and given over 1 min. Before and 20 min after injection, the subject exhaled into a collection bag through a Quintron (Milwaukee, WI) disposable modified Haldane-Priestley tube. After injection of [14C N-methyl]erythromycin, a med-lock was placed into the opposite antecubital vein for administration of midazolam (1 mg) (Bedford Laboratories, Bedford, OH). Midazolam was given 30 min after [14C N-methyl]-erythromycin injection, and then the med-lock was discontinued. Blood (5 ml) was drawn through the indwelling catheter into EDTA-containing Vacutainer tubes (Becton-Dickinson, Rutherford, NJ) just before midazolam administration and 5, 10, 15, 45, 60, 120, 180, 240, 360, 480, 600, and 720 min thereafter. Plasma was separated from blood cells by centrifugation (3500g, 15 min, 4°C) and stored at -80°C pending analysis for midazolam.
The second study phase (day 2) began the following morning, when the subject was given a single p.o. dose (400 mg) of fluconazole (Diflucan, Pfizer, New York, NY). Ninety minutes after fluconazole was given, the ERMBT was administered (at approximately the same time as on the previous day). Midazolam was given 30 min after [14C N-methyl]erythromycin injection (i.e., 120 min after fluconazole). Blood (10 ml) was collected through the indwelling catheter just before fluconazole administration, just before midazolam administration, and at 5, 10, 15, 45, 60, 120, 180, 240, 360, 480, 600, and 720 min after midazolam administration. Plasma was separated from blood cells by centrifugation (3500g, 15 min, 4°C) and stored at -80°C pending analysis for fluconazole and midazolam. During both study phases, meals and snacks, which were devoid of grapefruit-containing products and caffeinated beverages, were provided after the 240-min (4-h) blood collection with respect to midazolam administration. Vital signs (blood pressure, pulse, respirations, temperature) were monitored periodically until the last blood collection, after which the subject was discharged.
Analytic Procedures.14CO2 in breath following ERMBT administration. The collected breath was bubbled, via a peristaltic pump, into a scintillation vial containing 4 ml of a CO2 trapping solution (benzethonium hydroxide/absolute ethanol, 1:1) and the blue indicator dye thymolphthalein (1%). When 2 mmol of CO2 became trapped, as indicated when the blue dye turned clear, scintillation fluid (10 ml) was added, and 14C specific activity was measured by scintillation counting. Results were expressed as the percentage of administered 14C exhaled over the 1st hour postinjection, as estimated by a single breath collection at 20 min (Wagner, 1998).
Midazolam in plasma following i.v. administration. A 0.5-ml aliquot of plasma was diluted with 0.5 ml of nanopure water, after which 50 μl of internal standard (15N3-midazolam, 1 μg/ml in methanol) was added. After adding 1 ml of saturated sodium borate, pH 10, the mixture was extracted with 5 ml of toluene/methylene chloride (7:3), shaken for 20 min, and centrifuged (900g,20 min). The organic layer was collected and evaporated to dryness under nitrogen at 47°C. The residues were reconstituted with 50 μl of methanol, vortex-mixed gently, and then transferred to high-performance liquid chromatography (HPLC) vials containing 50 μl of nanopure water. A 10-μl aliquot was injected onto an HPLC system consisting of a Shimadzu (Columbia, MD) LC-10AD pump, DGU-14A degasser, SCL-10A controller, and SIL-10AD autosampler. The HPLC system was coupled to a Micromass Quattro II triple quadrupole mass spectrometer (Manchester, UK) operated in the positive ion electrospray atmospheric pressure ionization mode. Analytes were separated with a Zorbax Eclipse XDB-C8 column (5 μm, 2.1 × 50 mm) (Agilent Technologies, Palo Alto, CA) equipped with a Phenomenex (Torrance, CA) SecurityGuard C8 guard column. The mobile phase, at a flow rate of 0.3 ml/min, consisted initially of 45% methanol/water and 0.1% acetic acid. After 0.5 min, methanol was increased linearly to 60% over 1.5 min, held for 0.5 min, increased to 90% over 0.5 min, held for 5 min, then decreased back to 45% and held for 2 min. Ions with m/z ratios of 325.9 and 328.9 were monitored for midazolam and 15N3-midazolam, respectively. Midazolam was quantified by comparing peak area ratios (midazolam/15N3-midazolam) with those from a standard curve prepared with known amounts of midazolam using Micromass Masslynx data analysis software. The interday coefficient of variation in the assay was 6.7 and 5.0% at 2 and 50 ng/ml midazolam, respectively.
Fluconazole in plasma following p.o. administration. Perchloric acid/water (1:1) (15 μl) was added to plasma (0.2 ml) on ice, and the mixtures were vortex-mixed and centrifuged (15,000g for 10 min at 4°C). The supernatant (115 μl) was transferred to microfuge tubes containing 24 μl of potassium phosphate buffer (2 M, pH 12), vortex-mixed, kept on ice for 40 min, and then centrifuged (15,000g for 10 min at 4°C). The supernatant was transferred to HPLC vials, and 25 μl was injected onto the HPLC system, which consisted of a Hewlett Packard series 1050 UV detector and autosampler and an Agilent series 1100 pump and degasser. Fluconazole was eluted using an Eclipse XDB-C8 column (5 μm, 4.6 × 150 mm). Mobile phase A consisted of 40% methanol/40% acetonitrile/20% water, and mobile phase B consisted of 40% methanol/60% water. The initial conditions were 100% B at a flow rate of 0.7 ml/min, followed by a linear gradient to 100% A over 8 min. At 9 min, the flow rate was increased to 1.4 ml/min and held for 13 min, after which the flow rate was returned to 0.7 ml/min and 100% B over 1 min. Fluconazole was detected by UV absorbance at 260 nm. The interday variation in the assay was 3.4% at 5.5 μg/ml.
Pharmacokinetic analysis and in vitro-in vivo prediction of CYP3A inhibition. The pharmacokinetics of midazolam were evaluated using a noncompartmental approach (WinNonlin). The area under the plasma concentration versus time curve (AUC) was determined using the log-linear trapezoidal method, and systemic Cl was calculated as the ratio of dose to AUC. The terminal elimination half-life was obtained from the linear terminal slope of the semilog plot. The in vivo Ki (Ki,iv) for the inhibition of midazolam metabolism by fluconazole was calculated according to the following equation (Tucker, 1992; Yao et al., 2003): where Clint,i and Clint,c denote the Clint of the substrate under inhibited and control conditions, respectively; fm denotes the fraction of the clearance mediated by the inhibited pathway; [I] denotes the average plasma concentration of the inhibitor (competitive or noncompetitive); and Ki,iv is the in vivo Ki. If midazolam is assumed to be metabolized entirely by a single enzyme (i.e., fm = 1), eq. 1 simplifies to: The average inhibitor concentration after midazolam administration was calculated as AUC120–840min/720 min. The high unbound fraction of fluconazole in plasma (∼89%) was not considered in the calculation. Clint,c and Clint,i were determined by rearrangement of the well stirred model for hepatic clearance: where Qp and fu denote hepatic plasma flow and the plasma unbound fraction of midazolam (0.02), respectively. Qp was calculated using the measured hematocrit for each subject and the following equation: (0.02 l/min/kg × subject weight) × (1 - hematocrit). In addition, the mean Clint in each genotype group was calculated from the mean blood clearance after conversion of plasma clearance to blood clearance using a midazolam blood-to-plasma ratio of 0.6 (Gorski et al., 1998). Equation 2 was also used to determine the Ki,iv of fluconazole toward the ERMBT, assuming erythromycin N-demethylation is mediated exclusively by CYP3A (McGinnity et al., 1999) and that the ERMBTc/ERMBTi ratio approximates the blood Clint,c/Clint,i ratio for the N-demethylation pathway. The inhibitor concentration at the end of the 20-min breath collection period (i.e., 110 min after the fluconazole dose) was used in the calculation. The ratio of in vitro/in vivo Ki values, RKi, was calculated from each set (midazolam and erythromycin) of experimentally derived inhibition parameters. For erythromycin, the in vitro Ki was assumed to be equal to the IC50 for a noncompetitive inhibition mechanism (Kenworthy et al., 1999), which was based on inhibition data involving fluconazole and the related CYP3A substrate cyclosporine (Omar et al., 1997).
Statistical Analysis. The in vitro enzyme kinetic values are presented as mean ± S.E. of the parameter estimates. All the statistical comparisons were made using SAS (version 9.1, SAS Institute Inc., Cary, NC). The associations between CYP3A5 genotype and the following four outcomes were assessed by regression analysis: inhibition of midazolam clearance by fluconazole, inhibition of the ERMBT by fluconazole, the Ki,iv of fluconazole toward midazolam, and the Ki,iv of fluconazole toward the ERMBT. To stabilize the variance in the outcomes and satisfy the statistical assumptions required for analysis of variance, the logarithm of each outcome was used as the dependent variable. The assumptions of normality and homoscedasticity were verified from residual plots; with all the datasets, no evidence was found that contradicted model assumptions. The midazolam AUC and the ERMBT were adjusted for the logarithm of control values to appropriately account for baseline differences in these variables among subjects. Relationships between midazolam Cl and the ERMBT under control and inhibited conditions were evaluated from Spearman correlation coefficients (rs). For all the analyses, a p value <0.05 was considered significant.
Results
Specific CYP3A4 and CYP3A5 Content in Pooled HLM. Based on the specific contents of CYP3A4 and CYP3A5 in the individual liver microsomal preparations (data not shown), the pooled preparation designated as CYP3A5 expressors (CYP3A5+) contained a low amount, whereas the pool designated as CYP3A5 nonexpressors (CYP3A5-) contained a high amount of CYP3A4 immunoreactive protein (Fig. 1). Conversely, the CYP3A5+ HLM contained a high amount, whereas the CYP3A5-HLM contained essentially no detectable CYP3A5 immunoreactive protein. Specific protein content in each pair of replicates varied <15%. CYP3A4 and CYP3A5 content in the CYP3A5+ HLM averaged 13 and 58 pmol/mg, respectively; CYP3A4 content in the CYP3A5-HLM averaged 133 pmol/mg. Accordingly, CYP3A5 represented 82 and 0% of total CYP3A (CYP3A4 + CYP3A5) content in the CYP3A5+ and CYP3A5-HLM, respectively.
Inhibition Kinetics of Fluconazole toward CYP3A Catalytic Activity. With midazolam as the substrate, and in the absence of fluconazole, 1′-OH MDZ formation by each pool of HLM was consistent with classic Michaelis-Menten kinetics for a unienzyme system, as exemplified from linear, monophasic Eadie-Hofstee plots (data not shown). Despite the contribution of CYP3A5 to midazolam 1′-hydroxylation, monophasic plots were evident for both HLM pools because CYP3A4 and CYP3A5 exhibit similar Km values for midazolam 1′-hydroxylation (Gibbs et al., 1999). The apparent Km and Vmax values for 1′-OH MDZ formation were similar between the two pools (2.7 ± 0.2 μM and 2.2 ± 0.08 nmol/min/mg, respectively, for the CYP3A5-HLM; 2.9 ± 0.2 μM and 2.9 ± 0.1 nmol/min/mg, respectively, for the CYP3A5+ HLM). Therefore, the Clint was similar between the two pools (0.8 and 1.0 μl/min/mg for the CYP3A5- and CYP3A5+ HLM, respectively). For both HLM, the simple noncompetitive inhibition model best described the fluconazole-midazolam interaction. The Ki of fluconazole was slightly higher in the CYP3A5+ compared with the CYP3A5-HLM (Fig. 2 and Table 1).
1′-OH MDZ formation by each rCYP3A was consistent with classic Michaelis-Menten kinetics for a unienzyme system, and the simple linear mixed-type (rCYP3A4) or competitive (rCYP3A5) inhibition model best described the data. The Km and Vmax values obtained from rCYP3A5 were approximately 2.5-fold higher than those obtained from rCYP3A4 (0.9 ± 0.1 μM and 1.7 ± 0.1 pmol/min/pmol, respectively, for rCYP3A4; 2.4 ± 0.2 μM and 5.3 ± 0.1 pmol/min/pmol, respectively, for rCYP3A5). Therefore, the corresponding Clint values were similar between the two enzymes (2.0 and 2.2 μl/min/pmol for rCYP3A4 and rCYP3A5, respectively). The Ki obtained from rCYP3A5 was approximately 7-fold higher than that obtained from rCYP3A4 (Table 1), consistent with the previous report (Gibbs et al., 1999). The Ki obtained from the rCYP3A were roughly concordant (within 2-fold) with the corresponding values obtained from HLM.
With erythromycin as the substrate, the IC50 for fluconazole was slightly lower in the CYP3A5+ HLM compared with the CYP3A5-HLM (Fig. 3; Table 1). In contrast, the IC50 obtained from rCYP3A5 was 3.5 times higher than that obtained from rCYP3A4 (Table 1). The IC50 obtained from rCYP3A4 was similar to that with CYP3A5-HLM, whereas the IC50 obtained from rCYP3A5 was 6.5-fold higher than that obtained from CYP3A5+ HLM (Table 1).
Influence of CYP3A5 Genotype on the Extent of Inhibition of CYP3A Activity in Healthy Volunteers. To assess whether CYP3A5 expression influences the extent of CYP3A inhibition by fluconazole in vivo, a human volunteer study was conducted utilizing the same CYP3A probes. Because both probe drugs were given i.v., the interaction with fluconazole was assumed to reflect primarily inhibition of hepatic CYP3A.
Fluconazole was rapidly absorbed, reaching a peak concentration within 2 to 3 h after p.o. administration. Importantly, because midazolam and the ERMBT were administered 1.5 to 2 h after fluconazole was administered, the fluconazole concentration was near or at its peak at the time the CYP3A probes were given (Fig. 4). With midazolam as the probe, in the absence of fluconazole, the AUC and Cl of midazolam were similar among the three genotypic groups (Table 2) (p > 0.05). In the presence of fluconazole, the AUC of midazolam was significantly increased within each group (p < 0.0001); the percent increase was slightly greater in the CYP3A5*X/*X group (49 ± 3%) compared with CYP3A5*1 carriers (36 ± 15 and 40 ± 5% in the CYP3A5*1/*1 and CYP3A5*1/*X groups, respectively) (Table 2; Fig. 4). Correspondingly, midazolam Cl was significantly decreased by fluconazole within each genotypic group (p < 0.001), with the percent decrease being greatest in the CYP3A5*1 noncarriers (Table 2). The mean Ki,iv for fluconazole toward midazolam differed significantly among the three genotypic groups (p = 0.02), with the highest Ki,iv observed in the CYP3A5*1/*1 group (Fig. 5A). The mean (± S.D.) Ki,iv for the *1/*1, *1/*X, and *X/*X groups was 24 ± 17, 17 ± 8, and 13 ± 6 μM, respectively. When the average blood clearance data were used, the corresponding Ki,iv values were 21, 20, and 16 μM in the three genotype groups, respectively, showing a similar genotype effect as observed using the plasma clearance.
With the ERMBT as the probe, complete datasets were not available for two of the volunteers in the *1/*X group. In the absence of fluconazole, the mean (± S.D.) ERMBT results were 2.40 ± 0.49, 1.95 ± 0.45, and 2.54 ± 0.76 for the *1/*1, *1/*X, and *X/*X groups, respectively. The mean values for each group were not significantly different from each other (p > 0.05). In the presence of fluconazole, the mean ERMBT result was significantly decreased (p < 0.05) in all the groups (to 1.32 ± 0.19, 1.49 ± 0.27, and 1.48 ± 0.60, respectively), but the extent of inhibition did not differ among the three groups (p = 0.14). Likewise, the corresponding Ki,iv did not differ among the groups (p = 0.49) (Fig. 5B). The mean (± S.D.) Ki,iv for the *1/*1, *1/*X, and *X/*X groups was 30 ± 12, 77 ± 72, and 58 ± 53 μM, respectively. As has been reported previously (Kinirons et al., 1999; Masica et al., 2004), a correlation was not evident between midazolam Cl (whether normalized to body weight) and the ERMBT in both the presence and absence of fluconazole (rs ≤ 0.27, p ≥ 0.28).
In Vitro-in Vivo Extrapolations. Results from HLM and the in vivo study were used to calculate RKi values for the inhibition of midazolam intrinsic metabolic clearance and the ERMBT by fluconazole. Assuming that the CYP3A5*1/*1 group had the highest hepatic CYP3A5 content, coupled with the incomplete set of ERMBT values for the CYP3A5*1/*X group, only the *1/*1 group was used as the in vivo correlate to the CYP3A5+ HLM. With midazolam, the RKi was 1.04 and 1.15 for the CYP3A5+ and CYP3A5-groups, respectively. With erythromycin, the RKi was 1.8 and 1.2 for the CYP3A5+ and CYP3A5-groups, respectively.
Discussion
A growing body of in vitro data has shown that CYP3A5 is less susceptible (more resistant) than CYP3A4 to inhibition by some therapeutic agents (Gibbs et al., 1999; Jones et al., 1999; McConn et al., 2004; Wang et al., 2005). Because CYP3A5 can represent a significant portion of total hepatic CYP3A in some individuals (Kuehl et al., 2001; Lin et al., 2002) and CYP3A5 genotype can significantly influence the disposition of some CYP3A substrates in vivo (e.g., saquinavir, tacrolimus, sirolimus, alprazolam) (Andersson et al., 2005; Mouly et al., 2005; Le Meur et al., 2006; Park et al., 2006), differential inhibition of the two major CYP3A isoforms could explain in part the large interindividual variation observed in the extent of CYP3A-mediated drug-drug interactions. Despite the wealth of in vitro and in vivo data suggesting that CYP3A5 expression can influence the extent of CYP3A inhibition and/or the clearance of some substrates, to our knowledge, there has been no systematic in vitro-in vivo evaluation of this clinically relevant issue. Accordingly, using two distinct CYP3A substrates, the inhibitory effects of fluconazole on hepatic CYP3A activity were examined both in pooled HLM with a high or low CYP3A5 protein content and in healthy volunteers with known CYP3A5 genotype.
The inhibitory potency of fluconazole toward midazolam 1′-hydroxylation in the CYP3A5+ HLM, in which CYP3A5 content was approximately 4-fold greater than CYP3A4 content, was slightly less than that in the CYP3A5-HLM (Ki of 25 and 15 μM, respectively). In contrast, a much greater difference was observed between the recombinant enzymes (Ki of 53 and 7 μM for rCYP3A5 and rCYP3A4, respectively). This discrepancy between the two enzyme systems is attributed in part to the CYP3A5+ HLM containing CYP3A4 at an appreciable fraction (20%) of total CYP3A, which attenuated the relative resistance of CYP3A5 to inhibition. A simulation utilizing results with the same recombinant enzymes and with CYP3A5 representing 80% of total CYP3A content predicted a Ki of 60 μM (Gibbs et al., 1999). The additional discrepancy (60 versus 25 μM) could be caused by artificial differences in the microenvironment, e.g., lipid content, cytochrome P450 reductase levels, and the presence or absence of cytochrome b5 and other proteins that may influence the binding and effect of a CYP3A inhibitor. Unlike with midazolam, the inhibitory potency of fluconazole toward erythromycin N-demethylation in the CYP3A5+ HLM was slightly higher than that in the CYP3A5-HLM (IC50 of 54 and 70 μM, respectively). However, like with midazolam, the inhibitory difference was less profound than that with recombinant enzymes (IC50 of 350 and 100 μM for rCYP3A5 and rCYP3A4, respectively). Assuming that HLM represented the more clinically relevant system, the corresponding data were used to predict the effect of CYP3A5 genotype on the hepatic CYP3A-mediated interaction between fluconazole and the two probe substrates in vivo.
As predicted by HLM, the (inferred) presence of CYP3A5 lowered the inhibitory potency of fluconazole toward midazolam systemic clearance in healthy volunteers. The average Ki,iv for homozygous CYP3A5*1 carriers was 85% higher than that for the noncarriers and was comparable with the ∼70% difference observed in HLM. The ratio of the in vitro Ki to the Ki,iv (RKi) was at or near unity for both the CYP3A5- and CYP3A5+ groups (1.0 and 1.2, respectively), underscoring the excellent predictability of the midazolam-fluconazole interaction by HLM. The significant effect of CYP3A5 genotype on the extent of inhibition of midazolam clearance by fluconazole substantiated an earlier study reported by Yu et al. (2004), who observed a similar difference in the extent of inhibition by itraconazole in healthy Korean individuals genotyped as CYP3A5*1/*1, *1/*3,or *3/*3.
With erythromycin as the probe, the RKi value was near unity for the CYP3A5-group but not for the CYP3A5+ group (1.2 versus 1.8). A potential explanation for this moderate in vitro-in vivo discrepancy is that the ERMBTc/ERMBTi ratio did not accurately reflect the Clint,c/Clint,i ratio of erythromycin N-demethylation. In addition, the in vitro Ki estimated from the IC50 may have differed from the true Ki because of the limited IC50 dataset. Although the efflux transporter P-gp has been reported as an important determinant of the ERMBT (Kurnik et al., 2006), P-gp is unlikely to contribute to this in vitro-in vivo discrepancy, as fluconazole appeared not to alter P-gp function (Venkatakrishnan et al., 2000). Despite the modest underprediction of the magnitude of the ERMBT-fluconazole interaction, the data did predict the lack of influence of CYP3A5 genotype on the interaction, substantiating HLM as the preferred enzyme system to predict the inhibitory effect in vivo. Indeed, it is virtually impossible to predict the combined inhibitory effect toward CYP3A4 and CYP3A5 in vivo using recombinant enzymes because the relative expression of each enzyme varies considerably between individuals (Lamba et al., 2002; Lin et al., 2002), and the Clint between the two enzymes for different substrates can vary (Huang et al., 2004; McConn et al., 2004).
Based on the variable expression of CYP3A4 and CYP3A5, and the difference in the inhibitory potency of fluconazole between the two enzymes, a large interindividual variation in the extent of inhibition of affected substrates would be expected in vivo in individuals expressing CYP3A5. Results with midazolam support this contention. For example, two of the study participants in the CYP3A5*1/*1 group appeared more resistant to fluconazole inhibition than others in the group, with a Ki,iv (45 μM) that was similar to the Ki obtained with rCYP3A5 (54 μM), suggesting that CYP3A5 was the major CYP3A enzyme expressed in these individuals. Conversely, one participant in this group had a Ki,iv near the Ki obtained with rCYP3A4 (7 μM), suggesting that CYP3A4 was the major CYP3A enzyme expressed in this individual despite having the CYP3A5*1/*1 genotype. Participants with the CYP3A5*X/*X genotype exhibited the lowest degree of interindividual variation with respect to the extent of inhibition of midazolam clearance (1.2-fold), which is expected with metabolic contribution from a single dominant CYP3A enzyme in all the subjects. The CYP3A5*1/*1 group showed the highest degree of variation (3.0-fold), whereas the heterozygous group showed intermediate variation (1.5-fold) relative to the two homozygous groups. These observations and the closeness of the RKi values to unity are consistent with the significant but modest effect of CYP3A5 genotype on the extent of inhibition of midazolam clearance.
As also predicted by the CYP3A5-HLM, fluconazole was a more potent inhibitor of midazolam clearance than the ERMBT in CYP3A5*1 noncarriers, as exemplified by the 4.5-fold difference in the respective Ki,iv values (13 versus 58 μM), and showing that the potency of a CYP3A inhibitor can be substrate-dependent in vivo. Because the two probes were given within 30 min of each other, and by the i.v. route (to avoid a significant contribution by intestinal CYP3A), any variation in inhibitor concentrations and/or CYP3A expression levels as a potential source for this substrate difference was minimized. Moreover, the tracer dose of [14C]erythromycin given was too low to cause significant inhibition of CYP3A enzymes. The presence of CYP3A5 attenuated this apparent substrate dependence both in vitro and in vivo.
Results from the in vitro-in vivo extrapolations showed that the more complex in vitro system (HLM versus recombinant enzyme) was superior in predicting the magnitude of the in vivo interaction and the effect of CYP3A5 genotype. It follows that human hepatocytes may serve as an even better system to prospectively evaluate potential in vivo drug-drug interactions of this nature, as the enzyme microenvironment should be less disturbed during preparation of hepatocytes than during the preparation of HLM. Hepatocytes would also allow testing of the potential contribution of drug transporters to changes in hepatic clearance.
Although the modest effect of CYP3A5 genotype on the extent of inhibition of midazolam clearance by fluconazole is probably not clinically important, a significant clinical effect on the clearance of other CYP3A substrates, especially those with a narrow therapeutic window, cannot be excluded. For example, the clearance of the immunosuppressant tacrolimus has been shown to be significantly influenced by CYP3A5 genotype (Andersson et al., 2005; Thervet et al., 2005), and interactions between tacrolimus and fluconazole, as well as other, more potent azole antifungal agents (ketoconazole and itraconazole), have been documented (Venkatakrishnan et al., 2000). More recently, the hepatic clearance of the chemotherapeutic agent vincristine was estimated to be 5-fold greater in CYP3A5 high expressors compared with CYP3A5 low expressors based on results from HLM with known CYP3A4 and CYP3A5 content and CYP3A5 genotype (Dennison et al., 2007). Whether CYP3A5 genotype significantly influences the hepatic clearance of vincristine in vivo, as well as the extent of inhibition of both tacrolimus and vincristine by fluconazole and other azole antifungals in vivo, awaits further investigation.
In summary, the current work showed that the extent of hepatic CYP3A inhibition by fluconazole in vitro and in vivo varies with substrate and that inheritance of the CYP3A5*1 allele was significantly associated with reduced CYP3A inhibition by fluconazole with midazolam, but not with erythromycin, as the probe. Based on these in vitro-in vivo extrapolations, HLM, rather than recombinant enzymes, appear to be the preferred in vitro system for predicting hepatic CYP3A-mediated drug-drug interactions. Additional studies with CYP3A substrates whose metabolism is influenced significantly by CYP3A5 genotype (e.g., tacrolimus, vincristine), as well as sensitive to inhibition by fluconazole and other CYP3A inhibitors, are warranted to determine the clinical importance of differences in the susceptibility of CYP3A4 and CYP3A5 to inhibition.
Footnotes
-
This work was supported by the National Institutes of Health (M01 RR00046, R01 GM38149, R01 GM63666, P01 GM32165, U01 GM61393, and U01 GM61374).
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.107.018382.
-
ABBREVIATIONS: HLM, human liver microsome(s); P-gp, P-glycoprotein; 1′-OH MDZ, 1′-hydroxymidazolam; rCYP3A, recombinant CYP3A; PBS-T, phosphate-buffered saline containing 0.3% Tween 20; Clint, intrinsic clearance; PCR, polymerase chain reaction; ERMBT, erythromycin breath test; HPLC, high-performance liquid chromatography; AUC, area under the plasma concentration versus time curve.
- Received August 17, 2007.
- Accepted October 19, 2007.
- The American Society for Pharmacology and Experimental Therapeutics