Effect of CYP3A5 Expression on the Inhibition of CYP3A-Catalyzed Drug Metabolism: Impact on Modeling CYP3A-Mediated Drug-Drug Interactions

Abstract

The purpose of this study was to determine the impact of CYP3A5 expression on inhibitory potency (K_i or IC₅₀ values) of CYP3A inhibitors, using recombinant CYP3A4 and CYP3A5 (rCYP3A4 and rCYP3A5) and CYP3A5 genotyped human liver microsomes (HLMs). IC₅₀ ratios between rCYP3A4 and rCYP3A5 (rCYP3A5/rCYP3A4) of ketoconazole (KTZ) and itraconazole (ITZ) were 8.5 and 8.8 for midazolam (MDZ), 4.7 and 9.1 for testosterone (TST), 1.3 and 2.8 for terfenadine, and 0.6 and 1.7 for vincristine, respectively, suggesting substrate- and inhibitor-dependent selectivity of the two azoles. Due to the difference in the IC₅₀ values for CYP3A4 and CYP3A5, nonconcordant expression of CYP3A4 and CYP3A5 protein can significantly affect the observed magnitude of CYP3A-mediated drug-drug interactions in humans. Indeed, the IC₅₀ values of KTZ and ITZ for CYP3A-catalyzed MDZ and TST metabolism were significantly higher in HLMs with CYP3A5*1/*1 and CYP3A5*1/*3 genotypes than in HLMs with the CYP3A5*3/*3 genotype, showing CYP3A5 expression–dependent IC₅₀ values. Moreover, when IC₅₀ values of KTZ and ITZ for different HLMs were kinetically simulated based on CYP3A5 expression level and enzyme-specific IC₅₀ values, a good correlation between the simulated and the experimentally measured IC₅₀ values was observed. Further simulation analysis revealed that both the K_i ratio (for inhibitors) and V_max/K_m ratio (for substrates) between CYP3A4 and CYP3A5 were major factors for CYP3A5 expression–dependent IC₅₀ values. In conclusion, the present study demonstrates that CYP3A5 genotype and expression level have a significant impact on inhibitory potency for CYP3A-catalyzed drug metabolism, but that the magnitude of its effect is inhibitor-substrate pair specific.

Introduction

CYP3A is an important cytochrome P450 (P450) subfamily, because its members are expressed in many tissues and metabolize more than 50% of marketed drugs. Importantly, CYP3A4 is highly expressed in human intestine and liver, and is often the locus of drug-drug interactions (DDIs) (Lown et al., 1994; Thummel and Wilkinson, 1998). A structurally related isoform, CYP3A5, is polymorphically expressed in the human intestine and liver in different populations and exhibits an overlapping substrate specificity with that of CYP3A4 (Lamba et al., 2002; Huang et al., 2004). Consequently, it has become increasingly necessary to consider the impact of “CYP3A5 expresser” status on the pharmacokinetic and DDI profiles of CYP3A substrates (Isoherranen et al., 2008; Chandel et al., 2009).

With greater access to recombinant CYP3A (rCYP3A) proteins and human liver microsome (HLM) preparations of CYP3A5-genotyped subjects, attempts have been made to compare the catalytic and inhibitory properties of CYP3A4 with those of CYP3A5 (Williams et al., 2002; Huang et al., 2004; Yamaori et al., 2004). It is accepted that the majority of CYP3A substrates favor CYP3A4 when the two P450s are incubated under similar assay conditions and one carefully considers their catalytic efficiency in light of redox partner molar ratios (Jushchyshyn et al., 2005; Christensen et al., 2011; Lee and Goldstein, 2012). To date, only drugs such as vincristine (rCYP3A5/rCYP3A4 V_max/K_m ratio ≥ 8) and tacrolimus (rCYP3A5/rCYP3A4 V_max/K_m ratio ∼2) favor rCYP3A5 (Dai et al., 2006; Dennison et al., 2006; Niwa et al., 2008). In agreement, CYP3A5 expresser status is a major determinant of vincristine M1 formation in vitro and in vivo (Dennison et al., 2007, 2008; Egbelakin et al., 2011). Likewise, the respective in vivo (oral) clearance and the HLM-scaled clearance of tacrolimus are 48 and ∼60% lower in CYP3A5 nonexpressers, compared with expressors (Dai et al., 2006; Barry and Levine, 2010).

From the standpoint of CYP3A inhibition, azole antifungal agents have been well studied (Gibbs et al., 1999; Isoherranen et al., 2008; Niwa et al., 2008). With some substrates, the IC₅₀ ratio (rCYP3A5/rCYP3A4) has been reported to be as high as 19 and implicates CYP3A5 expresser status as an important determinant of K_i in human livers (Patki et al., 2003; Isoherranen et al., 2008; Niwa et al., 2008). Such an observation is supported by the ∼2-fold increase in fluconazole K_i with CYP3A5 expresser (versus CYP3A5 nonexpresser) HLM (Isoherranen et al., 2008). There are reports that itraconazole (ITZ) is rCYP3A4 selective with substrates such as oxybutynin and midazolam (MDZ) (Lukkari et al., 1998; Yamazaki et al., 2010). The latter authors reported a K_i ratio of ∼30 (rCYP3A5/rCYP3A4) as well as a K_i ratio of ∼2 in HLMs (CYP3A5 expressers versus nonexpressers).

Inheritance of at least one copy of the reference CYP3A5 allele (CYP3A5*1) confers a significant level of tissue CYP3A5 content, whereas inheritance of two variant alleles (e.g., CYP3A5*3, CYP3A5*6, or CYP3A5*7) leads to undetectable or very low CYP3A5 expression (Kuehl et al., 2001). Therefore, the data described earlier suggest that CYP3A5 expresser status, although not necessarily impacting the overall pharmacokinetics of a particular CYP3A substrate, may govern the inhibitor concentration ([I])/K_i value and the magnitude of a DDI. Clinically, there are at least four reports linking CYP3A5 genotype to the magnitude of CYP3A inhibition: the impact of ketoconazole (KTZ) on tacrolimus exposure (Chandel et al., 2009), effect of ITZ on the pharmacokinetics of intravenous MDZ (Yu et al., 2004), impact of ITZ on the pharmacological activity of clopidogrel (Suh et al., 2006), and the inhibition of MDZ clearance by fluconazole (Isoherranen et al., 2008). Interestingly, Isoherranen et al. (2008) also evaluated the inhibitory effect of fluconazole on erythromycin N-demethylase activity in HLM and found that its K_i was not affected by CYP3A5 expression. Clinical follow-up revealed no significant association between CYP3A5 genotype and the magnitude of effect of fluconazole on the erythromycin breath test. This raises the possibility that the impact of CYP3A5 expresser status on the magnitude of a CYP3A-mediated DDI may be substrate-dependent (Isoherranen et al., 2008).

In the present study, we aimed to determine the effect of CYP3A5 expression on the potency of inhibition of CYP3A-catalyzed drug metabolism by means of in vitro studies using rCYP3A4 and rCYP3A5 and nine CYP3A5-genotyped HLMs together with four representative CYP3A substrates [MDZ, testosterone (TST), terfenadine, and vincristine] and two inhibitors (ITZ and KTZ). It is concluded that IC₅₀ values for inhibition of CYP3A in HLM are indeed greatly dependent on the expression levels of CYP3A5, and that the quantitative impact of CYP3A5 expression is inhibitor-substrate pair–specific.

Materials and Methods

Chemicals and Reagents

Microsomes containing human rCYP3A4 and rCYP3A5 (Supersomes) were obtained from BD Biosciences (San Jose, CA). rCYP3A4 (1.0 nmol CYP3A4/mg) had been coexpressed with cytochrome b₅ (2.8 nmol/mg) and P450 reductase (2.0 µmol cytochrome c reduced/min/mg). rCYP3A5 (1.0 nmol CYP3A5/mg) had been coexpressed with cytochrome b₅ (2.1 nmol/mg) and P450 reductase (3.2 µmol cytochrome c reduced/min/mg). Terfenadine, vincristine, ITZ, and KTZ were purchased from Sigma-Aldrich (St. Louis, MO). MDZ and TST were purchased from Cerilliant (Round Rock, TX). All other general reagents were of analytical grade or better and were obtained from various commercial sources. Wild-type purified human CYP3A4, mutant purified human CYP3A4 (L211F/D214E and F213W), P450 reductase, and cytochrome b₅ were provided by Drs. James Halpert and Dmitri Davydov (Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California, San Diego, CA).

rCYP3A4- and rCYP3A5-Catalyzed Metabolism of TST, MDZ, Terfenadine, and Vincristine

Assessment of CYP3A4 and CYP3A5 Activity.

All incubations were conducted at 37°C as follows. For the determination of enzyme kinetic parameters, MDZ (0.5–25 µM) was incubated in a reaction mixture (final volume of 0.2 ml) that consisted of the following: rCYP3A4 or rCYP3A5 Supersomes (1 pmol P450/ml), potassium phosphate buffer (100 mM, pH 7.4), EDTA (1 mM), and NADPH (1 mM) for 5 minutes. The apparent product formation rate with respect to incubation time and protein concentration had previously been determined to be linear. Reactions were terminated by quenching with 1 volume of cold acetonitrile containing internal standard, 1′-hydroxy midazolam d₄ (1′OH-MDZ-d₄). After centrifugation at 4000g for 10 minutes, aliquots of the supernatants were subject to liquid chromatography–tandem mass spectrometry (LC/MS/MS) analysis. Analysis of 1′OH-MDZ in the incubate extracts was performed using an Accela ultra high–performance liquid chromatography (UHPLC) chromatographic system interfaced with an Orbitrap Discovery Mass Spectrometer (Thermo Fisher, Waltham, MA). Samples (10 µl) were injected via a CTC PAL autosampler (Leap Technologies, Carrboro, NC) into a BEH C18 1.7-µm 2.1 × 100–mm column (Waters Corporation, Milford, MA). The mobile phases were composed of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at a flow rate of 0.6 ml/min. A gradient linearly increased the composition of (B) from 10 to 40% over 6 minutes, followed by an increase to 95% over the next 2.5 minutes, maintenance at 95% for 1 minute, and then a return to initial conditions. Detection of 1′OH-MDZ was achieved in the positive ion electrospray (ESI) mode, with a source voltage of 5.0 kV, a capillary voltage of 15.5 V, and a capillary temperature of 235°C. Tube voltage was set at 50 V. Data acquisition was achieved by collecting full-scan accurate mass (resolution of 30,000) mass spectra between m/z (mass-to-charge ratio) 100 and 1000. Ion chromatograms of the exact masses of the MH⁺ ions of 1′-OH-MDZ (m/z 342.08103) and 1′-OH-MDZ-d₄ (m/z 346.10572) were extracted with a mass window of 10 ppm. Peaks in the ion chromatograms were integrated using XCalibur Quan Browser software (Thermo Fisher). The retention times of 1′-OH-MDZ and 1′-OH-MDZ-d₄ were 4.8 minutes, whereas 4-hydroxy MDZ eluted at 4.5 minutes.

For TST (5–500 μM), enzyme kinetic parameters were determined after incubation with rCYP3A4 or rCYP3A5 Supersomes (10 pmol P450/ml) in potassium phosphate buffer (100 mM, pH 7.4) containing EDTA (1 mM) and NADPH (1 mM) (final volume of 0.2 ml). Incubation time was 10 minutes. Again, the apparent product formation rate with respect to incubation time and protein concentration had previously been determined to be linear. Reactions were terminated by transfer of the incubation contents into a Solvinert hydrophobic polytetrafluoroethylene filter plate (Millipore, Billerica, MA) preloaded with 3 volumes of cold methanol containing internal standard, 6β-hydroxy testosterone d₃ (6βOH-TST-d₃). The filter plate was stacked with a 2-ml injection plate, centrifuged, and the filtrates were subjected to LC/MS/MS analysis. Analysis of 6βOH-TST in incubate extracts was performed using an SCL-10A vp pump (Shimadzu, Columbia, MD) connected to an API4000 Qtrap (AB Sciex, Vaughan, ON, Canada). Samples (10 µl) were injected via a CTC PAL autosampler (Leap Technologies) onto a Zorbax (Agilent Technologies, Palo Alto, CA) SB-C18 5-µm, 2.1 × 150–mm column. The mobile phases were composed of 0.1% formic acid in water (A) and 0.1% formic acid in methanol (B) at a flow rate of 0.6 ml/min. A gradient linearly increased the composition of (B) from 35 to 80% over 3.5 minutes, followed by an increase to 100% over the next 0.5 minutes, maintenance at 100% for 1 minute, and then a return to initial conditions. Detection of 6βOH-TST was achieved in the ESI multiple reaction monitoring mode by monitoring m/z transitions of 305.2 → 269.2 (6βOH-TST; retention time = 3.04 minutes) and 308.2 → 272.2 (6βOH-TST-d₃; retention time = 3.03 minutes). The declustering potential was set at 66 V, the collision energy was set at 21 V, and the turbo-V source temperature was set at 400°C. Analyst software version 1.4.2 was used for data analysis (Applied Biosystems Inc., Foster City, CA).

For the determination of enzyme kinetic parameters, terfenadine was incubated for 1 minute in a mixture (final volume of 0.2 ml) consisting of the following: rCYP3A4 or rCYP3A5 Supersomes (5 pmol P450/ml), potassium phosphate buffer (100 mM, pH 7.4), EDTA (1 mM), and NADPH (1 mM). The product formation rate with respect to the incubation time and protein concentration had previously been determined to be linear. Reactions were terminated by quenching with 1 volume of cold acetonitrile containing internal standard, 1′OH-MDZ-d₄. After centrifugation at 4000g for 10 minutes, aliquots of the supernatants were subject to LC/MS/MS analysis. Analysis of terfenadine alcohol in the incubate extracts was performed using an Accela UHPLC chromatographic system interfaced with an Orbitrap Discovery Mass Spectrometer (Thermo Fisher). Samples (10 µl) were injected via a CTC PAL autosampler (Leap Technologies) onto a BEH C18 1.7-µm 2.1 × 100–mm column (Waters Corporation). The mobile phases were composed of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at a flow rate of 0.6 ml/min. A gradient linearly increased the composition of (B) from 2 to 60% over 8 minutes and up to 95% over the next 0.5 minutes, maintained it at 95% for 0.5 minutes, and then returned it to initial conditions. Detection of terfenadine alcohol was achieved in the ESI mode, with a source voltage of 5.0 kV, a capillary voltage of 15.5 V, and a capillary temperature of 235°C. Tube lens voltage was set at 50 V. Data acquisition was achieved by collection of full-scan spectrum (mass resolution 30,000) over a range of 100–1000 amu. Extracted ion chromatograms of terfenadine alcohol (MH⁺ 488.31665) and internal standard 1′-OH-MDZ-d4 (MH⁺ 346.10572) were integrated. The retention times of terfenadine alcohol and 1′-OH-MDZ-d4 were 6.3 and 5.3 minutes, respectively.

For the determination of enzyme kinetic parameters, vincristine was incubated in a reaction mixture that contained rCYP3A4 (50 pmol P450/ml) or rCYP3A5 (10 pmol P450/ml) for 10 minutes, 5 minutes, and 3 minutes, respectively. The incubation mixture (final volume of 0.2 ml) also contained potassium phosphate buffer (100 mM, pH 7.4), MgCl₂ (1 mM), and NADPH (0.5 mM). The apparent reaction rate was linear with respect to incubation time and protein concentration. Reactions were terminated by the addition of 1 volume of ice-cold acetonitrile. Concentrations of vincristine metabolite (M1) were determined using UHPLC-mass spectrometry analysis (Dennison et al., 2006). Analysis of vincristine and M1 in the incubate extracts was performed using an Accela UHPLC pump connected to an LTQ Orbitrap Velos (Thermo Fisher). Samples (10 µl) were injected via a CTC PAL autosampler (Leap Technologies) onto a BEH C18 1.7-µm, 2.1 × 100–mm UPLC column (Waters Corporation). The mobile phases were composed of 0.1% formic acid in water (A) and methanol (B). A gradient linearly increased the composition of (B) from 10 to 40% over 6 minutes, followed by an increase to 95% over the next 2.5 minutes. Vincristine eluted at 6.2 minutes, and metabolite M1 at 5.7 minutes. The column eluted into the electrospray ion source, heated at 300°C and operated in the positive mode. Source voltage was set at 5.4 kV. The capillary temperature was set at 300°C. Sheath and auxiliary gases were operated at 50 and 15 arbitrary units, respectively. Quantitation of vincristine and M1 was accomplished by integration of peak areas in the respective accurate mass ion extracted chromatograms. Masses used were 413.20776 for vincristine and 406.19991 for M1, with windows of 5 ppm. These masses represent the doubly charged states of both molecules and are more abundant than the singly charged species under the conditions described.

Determination of Kinetic Parameters.

All kinetic data were analyzed by nonlinear regression analysis (KaleidaGraph 4.1; Synergy Software, Reading, PA), which involved fitting of the data to equations describing Michaelis-Menten (eq. 1) or sigmoidal (eq. 2) kinetics. In each case, the goodness of the fit was assessed based on visual inspection, assessment of χ², and the standard error of the parameter estimates.

(1)

(2)

In eqs. 1 and 2, v is the velocity of the metabolic reaction, [S] is the substrate concentration, and n is the Hill coefficient.

Inhibition of rCYP3A4 and rCYP3A5 by ITZ and KTZ.

KTZ (0.001–5 µM) and ITZ (0.002–10 µM) were studied as inhibitors of rCYP3A4- and rCYP3A5-catalyzed metabolism of MDZ, TST, terfenadine, and vincristine. In all cases, the IC₅₀ was generated at a [S] equivalent to the K_m (S₅₀) and at a [S] well below K_m (S₅₀) ([S]/K_m ratio ∼0.05). Both azoles were dissolved in dimethylsulfoxide and the appropriate volume of the stock solution was added to the incubation mixture (final concentration of dimethylsulfoxide did not exceed 0.5%). Incubation and LC/MS/MS analysis was carried out as described for the baseline kinetic studies. Percentage of control activity was determined as the ratio of the amount of 1′OH-MDZ, 6βOH-TST, terfenadine alcohol, and vincristine M1 formed in the presence to that in the absence (solvent alone) of KTZ or ITZ. IC₅₀ values were obtained by means of nonlinear least-squares analysis (eq. 3) using XLfit (IDBS, Guildford, UK).

(3)

In eq. 3, IC₅₀ is the x value for the point in the curve that is midway between the max (maximum activity remaining; minimal percentage inhibition) and min (minimal activity remaining; maximal percentage inhibition). The exponent y is the slope of the curve at its midpoint. (See Supplemental Material for Determination of Ligand-Induced Binding Spectra and Reconstitution of Purified Wild-Type and Mutant CYP3A4 with P450 Reductase and Cytochrome b₅.)

Inhibition of MDZ 1′-Hydroxylase and TST 6β-Hydroxylase Activity in HLM of CYP3A5-Genotyped Livers

Microsomal Preparations.

The collection and use of human tissue for research was approved by the University of Washington Human Subjects Review Board. Human liver samples were obtained from Human Tissue Bank maintained by the School of Pharmacy, University of Washington (Seattle, WA). Of the 60 available livers, three from a subset of CYP3A5*3/*3 donors (one male, two female subjects) and three from a subset of CYP3A5*1/*3 donors (three male donors) were selected randomly from each respective group for microsome preparation. Microsomes from three additional donors genotyped CYP3A5*1/*1 (one male, two female) were purchased from commercial sources to obtain microsomes from three donors of each genotype (CYP3A5*1/*1, CYP3A5*1/*3, and CYP3A5*3/*3). The procedures used for microsome isolation have been published previously (Paine et al., 1997; Lin et al., 2002; Dai et al., 2004). The CYP3A4 and CYP3A5 protein expression levels in each liver were quantified by Western blotting, and the data are summarized in Table 1. (See Supplemental Material for Assessment of CYP3A Inhibition and Determination of IC₅₀.)

TABLE 1.

Differential expression of CYP3A4 and CYP3A5 protein in various HLM preparations

Data are presented as the mean ± S.D. (n = 3 determinations).

		Expression Level
CYP3A5 Genotype	HLM Preparation	CYP3A4	CYP3A5	Percentage of CYP3A5a
		pmol/mg	pmol/mg	%
*1/*1	HL739	11.0	6.00	35.3
	HL867	110	17.0	13.4
	HL47	62.0	9.30	13.0
Mean ± S.D.		61.0 ± 49.5	10.8 ± 5.6*	20.6 ± 12.7
*1/*3	HL150	5.00	22.0	81.5
	HL125	72.0	58.0	44.6
	HL167	19.0	33.0	63.5
Mean ± S.D.		32.0 ± 35.3	37.7 ± 18.4*	63.2 ± 18.4*
*3/*3	HL152	25.0	1.00	3.80
	HL143	38.0	1.00	2.60
	HL131	28.0	1.00	3.40
Mean ± S.D.		30.3 ± 6.8	1.00 ± 0.00	3.29 ± 0.66

HL, human liver.

^aCYP3A5 expression represents CYP3A5 protein expression (pmol/mg) as a percentage of a total CYP3A protein expression in HLM (CYP3A4 pmol/mg + CYP3A5 pmol/mg).

^*P < 0.05, significantly different from expression level of CYP3A5 in HLM of a liver genotyped CYP3A5*3/*3.

Simulation of the CYP3A5 Expression-Dependent IC₅₀ Values of KTZ and ITZ for MDZ 1′-Hydroxylation and TST 6β-Hydroxylation

IC₅₀ values of KTZ and ITZ for MDZ 1′-hydroxylation and TST 6β-hydroxylation were simulated to model human liver microsomes with variable CYP3A4 and CYP3A5 expression. The following model was used (eq. 4), which assumes that MDZ and TST can be metabolized by CYP3A4 and CYP3A5:

(4)

where [S] is the applied concentration of substrate and [I] is the concentration of inhibitor. [E]_CYP3A4 and [E]_CYP3A5 are concentrations of CYP3A4 and CYP3A5, respectively, in the microsomes. v and v_i are metabolite formation velocities in the absence and presence of inhibitor, respectively. By incorporating K_m and V_max values of MDZ or TST shown in Table 2, and K_i values estimated from IC₅₀ values (K_i = IC₅₀ based on an [S] that was greater than 10-fold lower than the K_m value) of KTZ or ITZ shown in Table 3, as well as expression levels (concentrations) of CYP3A4 and CYP3A5 to eq. 4, the percentage of control activity (v_i/v × 100) was predicted and the IC₅₀ values were determined by nonlinear least-squares analysis using the MULTI program (Yamaoka et al., 1981). Leveraging the IC₅₀ values at the low [S] circumvents the need to consider different mechanisms of inhibition (competitive versus noncompetitive versus mixed). The relationship between the CYP3A5 expression and the simulated IC₅₀ values of KTZ and ITZ for MDZ 1′-hydroxylation and TST 6β-hydroxylation was determined. CYP3A5 expression represents the CYP3A5-specific content (pmol enzyme/mg protein) as a percentage of a total CYP3A protein content in HLM (CYP3A4 pmol/mg + CYP3A5 pmol/mg).

TABLE 2.

Kinetic parameters for the rCYP3A4- and rCYP3A5-catalyzed metabolism of MDZ, TST, terfenadine, and vincristine

Data are presented as the mean ± S.D. (n = 3 determinations), except for the CYP3A5/CYP3A4 V_max/K_m (S₅₀) ratio.

P450	Substrate	Model	Km (S50)	Vmaxa	nb	Vmax/Km(S50)	Vmax/Km (S50) Ratioc
			µM
CYP3A4	MDZ	Hyperbolic	1.8 ± 0.3	19 ± 1	—	11.0
	TST	Sigmoidal	78 ± 34	86 ± 3	1.2 ± 0.13	1.1
	Terfenadine	Hyperbolic	3.4 ± 0.4	24 ± 2	—	7
	Vincristine	Sigmoidal	75 ± 22	15 ± 4	1.5 ± 0.15	0.2
CYP3A5	MDZ	Hyperbolic	3.0 ± 0.5	35 ± 2	—	11.7	1.1
	TST	Hyperbolic	67 ± 13	56 ± 4	—	0.8	0.7
	Terfenadine	Hyperbolic	1.5 ± 0.2	18 ± 1	—	12.0	1.7
	Vincristine	Sigmoidal	47 ± 6.7	70 ± 7	1.5 ± 0.11	1.5	7.5

^aPresented as pmol/min/pmol P450, except for the formation of vincristine M1 metabolite. No attempt was made to formally quantitate M1, so data are presented as arbitrary units (see Materials and Methods).

^bHill coefficient.

^cRatio of CYP3A5 V_max/K_m (S₅₀) versus CYP3A4 V_max/K_m (S₅₀).

TABLE 3.

KTZ and ITZ as inhibitors of MDZ, TST, terfenadine, and vincristine metabolism

The data are presented as the mean ± S.D. (n = 3 determinations).

Substrate	P450	Substrate Conc.a	IC50b		CYP3A5/CYP3A4 IC50 Ratio
			KTZ	ITZ	KTZ	ITZ
		µM	nM	nM
MDZ	rCYP3A4	1.4 (0.8)	40 ± 10	240 ± 50	5.0c	9.6
		0.1 (0.06)	2.0 ± 0.2	17 ± 3	8.5d	8.8
	rCYP3A5	2.7 (0.9)	200 ± 4	2300 ± 400
		0.2 (0.07)	17 ± 3	150 ± 60
TST	rCYP3A4	78 (1.0)	9.0 ± 1	40 ± 7	4.4	33
		4.0 (0.05)	17 ± 1	45 ± 3	4.7	9.1
	rCYP3A5	67 (1.0)	40 ± 6	1300 ± 400
		3.0 (0.04)	80 ± 7	410 ± 180
Terfenadine	rCYP3A4	2.4 (0.7)	90 ± 5	900 ± 200	2.2	2.3
		0.2 (0.06)	8.2 ± 1	23 ± 4	1.3	2.8
	rCYP3A5	1.2 (0.8)	200 ± 5	2100 ± 500
		0.1 (0.07)	6.5 ± 2	64 ± 14
Vincristine	rCYP3A4	75 (1.0)	150 ± 30	360 ± 50	0.8	1.9
		4.0 (0.05)	111 ± 10	29 ± 4	0.6	1.7
	rCYP3A5	45 (1.0)	120 ± 10	680 ± 230
		2.0 (0.04)	70 ± 10	50 ± 5

Conc., concentration.

^a[S]/K_m ratio in parentheses.

^bIC₅₀ was determined at a [S] close to the K_m (S₅₀) and at a [S] below K_m (S₅₀) (see Table 2).

^cIC₅₀ ratio at higher [S].

^dIC₅₀ ratio at lower [S].

In addition, the relationship between CYP3A5 expression and IC₅₀ values against CYP3A-catalyzed metabolism was simulated using various K_i and V_max/K_m ratios (CYP3A5/CYP3A4). The previous model (eq. 4), which assumes that a virtual drug can be metabolized by CYP3A4 and CYP3A5, was used. Simulation of the effect of K_i ratio on CYP3A5 expression–dependent IC₅₀ values was performed using kinetic parameters for a virtual CYP3A substrate and inhibitor set as follows: [S] = 100 nM, V_max/K_m ratio = 1, and K_i,CYP3A4 = 100 nM. The effect of V_max/K_m ratio on CYP3A5 expression–dependent IC₅₀ values was also simulated using kinetic parameters for a virtual CYP3A substrate and inhibitor set as follows: [S] = 100 nM (< < K_m) and K_i ratio = 10 (K_i,CYP3A4 = 100 nM, K_i,CYP3A5 = 1000 nM).

Statistical Analysis

Data are presented as the mean of values obtained in at least three experiments with the S.D. Statistical analyses were performed with the unpaired Student’s t test, and a probability of less than 0.05 (P < 0.05) was considered to be statistically significant.

The average fold error (afe) between the simulated and observed IC₅₀ values was calculated using eq. 5 as described previously for evaluating accuracy of DDI predictions (Brown et al., 2005):

(5)

The differences between the observed and predicted IC₅₀ values in the individual livers for each inhibitor-substrate pair were tested using a two-tailed paired t test, and a P value <0.05 was considered significant.

Results

Kinetic Parameters Describing the rCYP3A4- and rCYP3A5-Catalyzed Metabolism of MDZ, TST, Terfenadine, and Vincristine.

rCYP3A4 and rCYP3A5 (coexpressed with P450 reductase and cytochrome b₅) were incubated with four different model substrates (TST, MDZ, vincristine, and terfenadine) to determine the kinetic parameters. In all cases, the rate of metabolite formation was determined under linear conditions and over a suitable substrate concentration range. Under the assay conditions chosen, rCYP3A4- and rCYP3A5-catalyzed formation of 1′OH-MDZ and terfenadine alcohol conformed to Michaelis-Menten (hyperbolic) kinetics (Table 2). On the other hand, vincristine M1 formation with both P450s was best described by sigmoidal kinetics. Only TST 6β-hydroxylation exhibited differential kinetics with rCYP3A4 (sigmoidal) and rCYP3A5 (hyperbolic).

As shown in Table 2, the kinetic parameters for rCYP3A4-catalyzed metabolism of MDZ, TST, terfenadine, and vincristine were calculated as 1.8 ± 0.3, 78 ± 34, 3.4 ± 0.4, and 75 ± 22 μM (K_m or S₅₀ values) and 19 ± 1, 86 ± 3, 24 ± 2, and 15 ± 4 pmol/min/pmol (V_max values), respectively. The kinetic parameters for rCYP3A5-catalyzed metabolism of MTZ, TST, terfenadine, and vincristine were 3.0 ± 0.5, 67 ± 13, 1.5 ± 0.2, and 47 ± 7 μM (K_m or S₅₀ values) and 35 ± 2, 56 ± 4, 18 ± 1, and 70 ± 7 pmol/min/pmol (V_max values), respectively. The corresponding V_max/K_m (S₅₀) ratios (rCYP3A5/rCYP3A4) for TST, MDZ, terfenadine, and vincristine were 0.7, 1.1, 1.7, and 7.5, respectively. This permitted an assessment of CYP3A inhibition by ITZ and KTZ with a set of model substrates that spanned a 10-fold range of V_max/K_m (S₅₀) ratios (rCYP3A5/rCYP3A4).

Inhibition of rCYP3A4 and rCYP3A5 by ITZ and KTZ.

The inhibitory effect of KTZ and ITZ on rCYP3A4- and rCYP3A5-catalyzed MDZ 1′-hydroxylation, TST 6β-hydroxylation, terfenadine alcohol formation, and vincristine M1 formation was analyzed (Supplemental Figs. 1–4; Supplemental Tables 1 and 2). In each case, IC₅₀ values were obtained at a higher (∼K_m) and lower (<K_m) [S]. As expected, the IC₅₀ decrease (≥11-fold) at the lower [S] was consistent with competitive inhibition. However, for rCYP3A4 with TST (ITZ and KTZ) and vincristine (KTZ), as well as rCYP3A5 with TST (KTZ and ITZ) and vincristine (KTZ), the IC₅₀ change was minimal. Such a profile is more consistent with noncompetitive inhibition (Table 3).

As shown in Table 3, the IC₅₀ ratios (rCYP3A5/rCYP3A4) of KTZ and ITZ at the higher [S] were 5.0 and 9.6 (MDZ), 4.4 and 33 (TST), 2.2 and 2.3 (terfenadine), and 0.8 and 1.9 (vincristine), respectively, suggesting substrate and/or inhibitor dependency for the difference in the IC₅₀ values for rCYP3A4- and rCYP3A5-catalyzed drug metabolism. With the exception of the inhibition of TST 6β-hydroxylation by ITZ, similar differences in IC₅₀ ratios were observed at the lower [S].

In the present study, rCYP3A4 exhibited sigmoidal kinetics with TST as a substrate, whereas rCYP3A5 did not (Table 2). Therefore, we hypothesized that the large rCYP3A5/rCYP3A4 IC₅₀ ratio (33) of ITZ at the higher concentration of TST (∼S₅₀) was largely driven by the sigmoidal kinetics exhibited by rCYP3A4 (Table 3). To test the hypothesis, ITZ was evaluated as an inhibitor of wild-type and mutant (L211F/D214E) CYP3A4. When reconstituted with P450 reductase and cytochrome b₅, the former exhibits sigmoidal kinetics with TST (Hill coefficient n = 1.3). In contrast, TST 6β-hydroxylation catalyzed by the mutant form of CYP3A4 conforms to hyperbolic kinetics (Hill coefficient n = 1). Consequently, at least two amino acid residues (Leu211 and Asp214) are thought to govern the sigmoidal kinetics of CYP3A4 (Harlow and Halpert, 1998). As shown in Supplemental Table 3 and Supplemental Fig. 6, there was no major difference in the IC₅₀ value of ITZ (wild-type versus mutant CYP3A4) when TST was incubated at two concentrations (25 and 100 µM). The higher concentration of TST approximated the S₅₀ (∼100 µM) of both CYP3A4 forms (Harlow and Halpert, 1998).

To determine whether ITZ and KTZ bind CYP3A4 and CYP3A5 differentially in the absence of a substrate but bind within the active site of CYP3A4 and CYP3A5, ligand-induced difference spectra (binding spectra) were generated with both P450s (Supplemental Fig. 5). Both ITZ and KTZ generated typical type II binding spectra with rCYP3A4 and rCYP3A5, characteristic of nitrogen coordination to the heme and causing a heme iron high to low spin shift (Kunze et al., 2006; Locuson et al., 2007). However, the maximum change in the magnitude of the ligand-induced spectral change for ITZ was smaller with rCYP3A5 [0.010 arbitrary units (AU)] than with rCYP3A4 (0.023 AU). For KTZ, the difference in the maximum spectral change between the two P450s was smaller; the maximum absorbance difference was 0.030 AU with rCYP3A4 and 0.023 AU with rCYP3A5. With regard to ligand affinity, the K_s of ITZ was 118 nM with CYP3A4. The K_s with rCYP3A5 was 3-fold higher (352 nM). For KTZ, the determined K_s value for rCYP3A5 (460 nM) was 5-fold higher than the K_s value with CYP3A4 (86 nM).

Inhibitory Effect of ITZ and KTZ on MDZ and TST Metabolism in HLM of CYP3A5-Genotyped Livers.

The inhibitory effect of KTZ and ITZ on CYP3A-catalyzed MDZ 1′-hydroxylation and TST 6β-hydroxylation was evaluated using CYP3A5-genotyped HLM (Supplemental Figs. 7 and 8; Table 1). As summarized in Table 4, the IC₅₀ values of KTZ and ITZ for MDZ 1′-hydroxylation and TST 6β-hydroxylation varied up to 30-fold between HLMs. Interestingly, the IC₅₀ values of KTZ and ITZ were the lowest in the group of HLMs with CYP3A5*3/*3 genotype, followed by CYP3A5*1/*1 genotype, and the highest in CYP3A5*1/*3 genotype. Overall, the IC₅₀ values of ITZ toward MDZ and TST were significantly higher in the CYP3A5-expressing livers than in the CYP3A5*3/*3 livers (Table 4). The IC₅₀ values of KTZ toward MDZ were also significantly higher in the CYP3A5-expressing livers than in the CYP3A5*3/*3 livers, but no significant differences in the IC₅₀ values of KTZ for TST metabolism were observed among CYP3A5 genotypes. Since the expression level of CYP3A5 protein in HLM with CYP3A5*1/*1 and CYP3A5*1/*3 genotypes was significantly higher than that in the HLM of livers genotyped CYP3A5*3/*3, the IC₅₀ values overall appeared to track with the expression level of CYP3A5 (Tables 1 and 4).

TABLE 4.

Impact of CYP3A5 expression on the inhibition of MDZ and TST metabolism in HLM by KTZ and ITZ

Data are presented as the mean ± S.D. (n = 3 determinations).

CYP3A5 Genotype	HLM Preparation	IC50
		MDZa		TSTb
		+ KTZ	+ ITZ	+ KTZ	+ ITZ
		nM	nM	nM	nM
*1/*1	HL739	35.8 ± 3.8	108 ± 24	31.5 ± 4.8	85.4 ± 10.5
	HL867	34.3 ± 3.2	111 ± 15	19.7 ± 3.8	120 ± 12
	HL47	30.7 ± 4.0	123 ± 21	29.1 ± 2.7	111 ± 11
Mean ± S.D.		33.6 ± 2.6*	114 ± 8*	26.8 ± 6.3	105 ± 18*
*1/*3	HL150	132 ± 13	1133 ± 782	308 ± 52	513 ± 105
	HL125	11.8 ± 1.5	161 ± 27	53.9 ± 3.8	158 ± 14
	HL167	24.3 ± 2.2	176 ± 13	64.2 ± 6.6	195 ± 15
Mean ± S.D.		56.1 ± 66.3	490 ± 557	142 ± 144	289 ± 195*
*3/*3	HL152	6.92 ± 1.12	47.3 ± 8.5	25.8 ± 3.8	20.0 ± 2.3
	HL143	12.9 ± 2.0	83.0 ± 5.9	27.1 ± 5.5	21.6 ± 1.7
	HL131	11.5 ± 1.0	53.3 ± 7.4	10.2 ± 1.1	46.7 ± 4.2
Mean ± S.D.		10.4 ± 3.1	61.2 ± 19.1	21.0 ± 9.4	29.4 ± 15.0

HL, human liver.

^aInitial concentration of MDZ was 1 μM.

^bInitial concentration of TST was 10 μM.

^*P < 0.05, significantly different from the IC₅₀ obtained with HLM of a liver genotyped CYP3A5*3/*3.

By incorporating the kinetic K_m and V_max values (Table 2) and K_i (IC₅₀-derived) values of KTZ and ITZ (Table 3) for rCYP3A4- and rCYP3A5-catalyzed TST and MDZ metabolism, as well as expression levels of CYP3A4 and CYP3A5 protein in HLM (Table 1) into eq. 4, the relationship between CYP3A5 expression and IC₅₀ values of KTZ and ITZ for CYP3A-catalyzed MDZ and TST metabolism in individual HLM was simulated (Fig. 1, solid lines). The simulated IC₅₀ values were not significantly different from those observed (P > 0.05 for all inhibitor-substrate pairs), yet the afe analysis showed that the IC₅₀ values of KTZ and ITZ toward TST hydroxylation (afe 0.77 and 0.999 for KTZ and ITZ, respectively) were more accurately predicted than those against MDZ (afe 0.26 and 0.40 for KTZ and ITZ, respectively). Interestingly, the relatively low prediction accuracy for the KTZ-MDZ IC₅₀ was independent of the genotype, with the predicted IC₅₀ values consistently lower than those observed. In contrast, the discrepancies between predicted and observed values for ITZ-MDZ interaction were mainly driven by the CYP3A5*1/*1 livers. Of the two azoles tested, the IC₅₀ of ITZ was most sensitive to CYP3A5 expression level in HLM (Fig. 1).

Fig. 1.

Relationship between CYP3A5 expression and IC₅₀ values of KTZ and ITZ for the CYP3A-catalyzed metabolism of MDZ and TST in HLM. Relationship between CYP3A5 expression and IC₅₀ values of KTZ (A) and ITZ (B) for MDZ and KTZ (C) and ITZ (D) for TST metabolism. CYP3A5 expression represents CYP3A5 protein expression (pmol/mg) as a percentage of a total CYP3A protein expression in HLM (CYP3A4 pmol/mg + CYP3A5 pmol/mg). Solid and dotted lines represent simulated data and their 2-fold error ranges, respectively (see Materials and Methods); ●, ▲, and ■ represent observed data using HLMs with CYP3A5*1/*1, CYP3A5*1/*3, and CYP3A5*3/*3 genotypes, respectively. Data are presented as the mean ± S.D. (n = 3 determinations).

Simulation Analysis of the Effect of CYP3A5 Expression on Inhibition of CYP3A-Catalyzed Drug Metabolism.

To consider the effect of the different K_i values of novel CYP3A inhibitors, and V_max/K_m values between CYP3A4 and CYP3A5 of CYP3A substrates, on CYP3A5 expression–dependent IC₅₀ values, the relationship between CYP3A5 expression and IC₅₀ values against CYP3A-catalyzed metabolism was simulated using various K_i and V_max/K_m ratios (CYP3A5/CYP3A4) (Fig. 2). The effect of the K_i ratio on CYP3A5 expression–dependent IC₅₀ values was simulated using kinetic parameters for a virtual CYP3A substrate and inhibitor set as follows: substrate concentration = 100 nM, V_max/K_m ratio = 1 (i.e., CYP3A4 and CYP3A5 have equal intrinsic clearances for the substrate), and K_i,CYP3A4 = 100 nM (Fig. 2A). When the K_i ratio (CYP3A5/CYP3A4) was 1.0, no effect of CYP3A5 expression on the IC₅₀ value for CYP3A-catalyzed drug metabolism was observed. The simulated IC₅₀ value increased with increasing relative CYP3A5 expression with an increase in the K_i ratio and conversely decreased with decreased K_i ratio.

Fig. 2.

Simulation analysis of effect of CYP3A5 expression on inhibition of CYP3A-catalyzed drug metabolism. (A) Impact of K_i ratio (CYP3A5/CYP3A4) on CYP3A5 expression–dependent IC₅₀ values. Kinetic parameters for a virtual CYP3A substrate and inhibitor used in current simulations are set as follows: substrate concentration = 100 nM, V_max/K_m ratio = 1, and K_i,CYP3A4 = 100 nM. (B) Impact of V_max/K_m ratio (CYP3A5/CYP3A4) on CYP3A5 expression–dependent IC₅₀ values. Kinetic parameters for a virtual CYP3A substrate and inhibitor used in current simulations are set as follows: substrate concentration = 100 nM (< < K_m) and K_i ratio = 10 (K_i,CYP3A4 = 100 nM, K_i,CYP3A5 = 1000 nM). CYP3A5 expression represents the percentage of total CYP3A (CYP3A4 + CYP3A5) expression (see Materials and Methods).

Simulation of the effect of CYP3A V_max/K_m ratio on CYP3A5 expression–dependent IC₅₀ values was also performed using kinetic parameters for a virtual CYP3A substrate and inhibitor set as follows: substrate concentration = 100 nM (< < K_m) and K_i ratio = 10 (K_i,CYP3A4 = 100 nM, K_i,CYP3A5 = 1000 nM) (Fig. 2B). When the V_max/K_m ratio was higher, the IC₅₀ value for CYP3A-catalyzed drug metabolism increased even when CYP3A5 expression was low (<20% of total CYP3A). In contrast, when the V_max/K_m ratio was lower, the IC₅₀ value was constant or slightly changed at lower CYP3A5 levels, then increased at higher CYP3A5 levels. This simulation suggested that variability of IC₅₀ values for CYP3A-catalyzed drug metabolism became more sensitive to CYP3A5 expression when CYP3A5 contribution to substrate clearance was greater.

Discussion

The results of this study show that CYP3A5 genotype and protein expression level have a significant impact on the inhibitory potency of ITZ and KTZ, but that the magnitude of the effect is inhibitor-substrate pair–specific. The IC₅₀ ratios (rCYP3A5/rCYP3A4) of KTZ and ITZ for the inhibition of rCYP3A4- and rCYP3A5-catalyzed metabolism of MDZ and TST were much higher (>4) than those for terfenadine and vincristine (<3). This difference in the IC₅₀ values for CYP3A4 and CYP3A5 may result in a complex CYP3A-mediated DDI profile in humans carrying the CYP3A5*1 allele that confers significant expression of the encoded enzyme (Kuehl et al., 2001). It is noteworthy that the range of IC₅₀ ratios cannot be rationalized solely on the basis of higher ITZ (∼3-fold) and KTZ (∼5-fold) CYP3A4 binding affinity (K_s) in the absence of a substrate. This implies that one has to consider the binding of both substrate and inhibitor within the CYP3A active site and the effect of allosterism in substrate-dependent inhibitory potency. In agreement with what would be predicted from rP450 data, the IC₅₀ values of KTZ and ITZ for MDZ 1′-hydroxylation and ITZ for TST 6β-hydroxylation were significantly higher in HLMs with CYP3A5*1/*1 and CYP3A5*1/*3 genotypes than in HLMs with CYP3A5*3/*3 genotype. However, no significant difference in the IC₅₀ values of KTZ for TST metabolism was observed among the CYP3A5 genotype groups (Table 4). Based on the higher CYP3A5 protein expression level in HLMs with CYP3A5*1/*1 and CYP3A5*1/*3 genotypes compared with CYP3A5*3/*3 genotype, the greater IC₅₀ values of KTZ and ITZ for CYP3A-catalyzed MDZ and TST metabolism are likely due to the higher expression level of CYP3A5 and lower inhibitory potency of KTZ and ITZ toward CYP3A5 compared with CYP3A4 (higher IC₅₀ ratios) (Tables 1, 3, and 4). Alternatively, one can anticipate no significant effect of CYP3A5 expression on the IC₅₀ values of KTZ and ITZ for CYP3A-catalyzed metabolism of terfenadine and vincristine, based on the low IC₅₀ ratios of KTZ and ITZ shown in Table 3. Further investigation would be needed to clarify the impact of CYP3A5 expression on the inhibition of CYP3A-catalyzed metabolism of terfenadine and vincristine. Overall, the relative CYP3A5 expression level rather than genotype (heterozygous or homozygous CYP3A5*1) determined the magnitude of IC₅₀ values of ITZ and KTZ in HLMs.

Results from our simulation of the IC₅₀ values at different CYP3A5 expression levels with KTZ and ITZ using kinetic data from CYP3A4 and CYP3A5 Supersomes and protein expression levels of CYP3A4 and CYP3A5 in the HLMs support our conclusion that relative CYP3A5 expression determines the inhibitory potency toward a given substrate (Fig. 1). When the simulated IC₅₀ values of KTZ and ITZ were compared with the experimentally measured IC₅₀ values, the predicted IC₅₀ values were not significantly different from each other, demonstrating that simulation can be used to predict the effect of CYP3A5 expression on the IC₅₀ values for CYP3A-catalyzed drug metabolism (Fig. 1). Interestingly, the prediction of the IC₅₀ values was more accurate for both inhibitors when TST was used as a substrate than with MDZ (afe 0.32 versus 0.87). This difference in prediction accuracy could be due to unaccounted allosteric, substrate-specific effects. Such allosteric effects were recently shown in vitro and in vivo between fluconazole and MDZ (Yang et al., 2012). However, for both MDZ and TST metabolism, the observed IC₅₀ values were generally within the 2-fold range of the predicted values, with the exception of HL150. In this donor, the observed IC₅₀ value was significantly higher than that predicted and that observed in the other donors. One possible explanation for this inconsistency is the potential contribution of CYP3A7 to drug metabolism in HL150 and relatively weak inhibition of CYP3A7 by KTZ and ITZ, yielding a higher IC₅₀ value of KTZ and ITZ for MDZ and TST metabolism (Nishimura et al., 2003; Sim et al., 2005; Hata et al., 2010). Such a hypothesis is supported by the much higher (∼16-fold) IC₅₀ of KTZ for rCYP3A7-catalyzed TST metabolism versus rCYP3A5 (630 versus 40 nM; unpublished data). Another possible explanation is that minor contributions from other P450 enzymes to MDZ and TST metabolism, such as CYP2C9 and CYP2C19, may be factors in the deviation (Choi et al., 2005).

To further evaluate the potential impact of CYP3A5 expression in CYP3A IC₅₀ values, the effect of K_i ratios (CYP3A5/CYP3A4) of CYP3A inhibitors and V_max/K_m ratios (CYP3A5/CYP3A4) of CYP3A substrates on CYP3A5 expression-dependent IC₅₀ values was simulated. As shown in Fig. 2A, a CYP3A5 expression–dependent change in the IC₅₀ values for CYP3A-catalyzed drug metabolism is clearly predicted when the K_i ratio deviates from 1, whereas no effect of CYP3A5 expression on IC₅₀ value is expected when the K_i ratio is unity. Therefore, it is concluded that the K_i ratio is one of the major determinants for CYP3A5 expression–dependent IC₅₀ values. On the other hand, as shown in Fig. 2B, the effect of CYP3A5 expression on the IC₅₀ value for CYP3A-catalyzed drug metabolism is also dependent on the relative catalytic efficiency of CYP3A5 when compared with CYP3A4, and is more sensitive at higher CYP3A5/CYP3A4 V_max/K_m ratios. Accordingly, if the contribution of CYP3A5 to total CYP3A-catalyzed drug metabolism is high (that is, the intrinsic clearance of the drug is higher for CYP3A5 than for CYP3A4), even very low relative CYP3A5 expression has a significant effect on the observed apparent IC₅₀ values. In such a case, CYP3A5 expression would again become an important factor to consider when determining IC₅₀ values for CYP3A-catalyzed drug metabolism in HLMs. However, if the contribution of CYP3A5 to total CYP3A-catalyzed drug metabolism is low, i.e., intrinsic clearance of CYP3A5 ∼0.1 (Fig. 2), the effect of an individual difference in CYP3A5 expression on the IC₅₀ values is likely to be negligible, despite any difference in the K_i value for CYP3A4 and CYP3A5.

The present findings suggest that, when CYP3A-mediated DDI studies are planned, the inhibition potency of the new chemical entity toward CYP3A4 and CYP3A5 should be tested in vitro using the same probe as will be used in the clinical study to determine whether CYP3A5 genotype will affect interindividual variability in the magnitude of the DDI observed and significantly affect study power. The inhibitor selectivity should be considered in relation to the intrinsic clearance ratio of the substrate. The influence of CYP3A5 expression on the magnitude of DDIs could also be taken into consideration when the in vivo probe substrate is selected. On the other hand, when the new chemical entity is considered as the object drug of the DDI, it will be prudent to determine the contribution of CYP3A5 to the overall clearance of the drug (in CYP3A5 expressers) and to test the effect of CYP3A5 expression on the IC₅₀ of KTZ or another strong CYP3A inhibitor toward that specific compound. It is important to note that, due to the allosteric behavior of CYP3A, the effect of CYP3A5 expression on in vivo IC₅₀ values is specific for the inhibitor-substrate pair and needs to be determined individually in vitro for the combination in the study. Unfortunately, very few compounds have been studied in vivo or in vitro for their potency as CYP3A5 inhibitors, and, as such, extrapolation of these data to the overall variability of the magnitude of CYP3A-mediated DDIs can be challenging. This is emphasized by the fact that, at present, very limited data are available for CYP3A5 contribution to the clearance of the known sensitive substrates of CYP3A. More systematic studies are needed to establish the overall significance of CYP3A5 inhibition in clinical DDIs.

In conclusion, the present study demonstrates that the IC₅₀ value for CYP3A-catalyzed drug metabolism measured using HLMs varies due to differences in CYP3A5 expression/genotype, the substrate studied, and the relative inhibitory potency toward CYP3A4 and CYP3A5. These findings imply that the HLM assay has the potential to misread the IC₅₀ value for CYP3A-catalyzed drug metabolism. However, simulation analysis, as described in the present study, may be useful when considering variability of IC₅₀ values for CYP3A-catalyzed drug metabolism in individual HLMs, and may enable relatively precise prediction of DDIs in vivo involving the inhibition of CYP3A.

Abbreviations

afe

average fold error

AU

arbitrary units

DDI

drug-drug interaction

ESI

positive ion electrospray

HLM

human liver microsome

[I]

inhibitor concentration

ITZ

itraconazole

K_i

inhibition constant

K_s

spectral binding constant

KTZ

ketoconazole

LC/MS/MS

liquid chromatography–tandem mass spectrometry

MDZ

midazolam

m/z

mass-to-charge ratio

1′OH-MDZ

1′-hydroxy midazolam

6βOH-TST

6β-hydroxy testosterone

P450

cytochrome P450

rCYP3A4

recombinant CYP3A4

rCYP3A5

recombinant CYP3A5

[S]

substrate concentration

TST

testosterone

UHPLC

ultra high–performance liquid chromatography

Authorship Contributions

Participated in research design: Isoherranen, Thummel, Rodrigues.

Conducted experiments: Peng, Grubb, Chang, Shirasaka.

Performed data analysis: Shirasaka, Chang, Grubb, Isoherranen, Rodrigues.

Wrote or contributed to the writing of the manuscript: Shirasaka, Rodrigues, Thummel, Isoherranen, Chang, Grubb.