Comparative Study of the Effects of Antituberculosis Drugs and Antiretroviral Drugs on Cytochrome P450 3A4 and P-Glycoprotein

Abstract

Predicting drug-drug interactions (DDIs) related to cytochrome P450 (CYP), such as CYP3A4 and one of the major drug transporters, P-glycoprotein (P-gp), is crucial in the development of future chemotherapeutic regimens to treat tuberculosis (TB) and TB/AIDS coinfection cases. We evaluated the effects of 30 anti-TB drugs, novel candidates, macrolides, and representative antiretroviral drugs on human CYP3A4 activity using a commercially available screening kit for CYP3A4 inhibitors and a human hepatocyte, HepaRG. Moreover, in order to estimate the interactions of these drugs with human P-gp, screening for substrates was performed. For some substrates, P-gp inhibition tests were carried out using P-gp-expressing MDCK cells. As a result, almost all the compounds showed the expected effects on human CYP3A4 both in the in vitro screening and in HepaRG cells. Importantly, the unproven mechanisms of DDIs caused by WHO group 5 drugs, thioamides, and p-aminosalicylic acid were elucidated. Intriguingly, clofazimine (CFZ) exhibited weak inductive effects on CYP3A4 at >0.25 μM in HepaRG cells, while an inhibitory effect was observed at 1.69 μM in the in vitro screening, suggesting that CFZ autoinduces CYP3A4 in the human liver. Our method, based on one of the pharmacokinetics parameters in humans, provides more practical information associated with not only DDIs but also with drug metabolism.

INTRODUCTION

In the development of combination regimens for tuberculosis (TB) and coinfection with TB and human immunodeficiency virus (HIV), the prediction of drug-drug interactions (DDIs) and drug-food interactions relevant to drug-metabolizing enzymes, such as cytochrome P450 (CYP), and membrane transporters, especially P-glycoprotein (P-gp), helps avoid the risk of adverse reactions caused by DDIs and maintain inherent medicinal effects. According to the World Health Organization (WHO) and recent reports, 12 compounds have been evaluated in clinical trials for the treatment of TB, especially for multidrug-resistant (MDR)- and extensively drug-resistant (XDR)-TB (1, 2). Of those, bedaquiline was approved by the U.S. Food and Drug Administration (FDA) last year (3). In the foreseeable future, the other compounds are likely to be introduced into the clinical setting as orphan drugs to treat refractory drug-resistant TB. In general, MDR-/XDR-TB has been treated with a second-line anti-TB drug combination, which is less potent, more toxic, and requires a longer duration of treatment. However, the combination therapy increases the frequency of adverse reactions and DDIs, and the effects of some second-line drugs both on CYPs and P-gp remain unclear. For instance, the effects of p-aminosalicylic acid (PAS) and clofazimine (CFZ), both of which have been used for more than half a century, on CYP3A4 and P-gp are still unclear and remain controversial (4, 5). In addition, the effect on CYP3A4 of thiacetazone (TAC), which is preferentially metabolized by flavin-containing monooxygenase, has not been officially published (6). Moreover, the mechanisms of all the unexpected DDIs between a novel candidate linezolid (LZD) and the well-known CYP3A4 modulators rifampin (RIF) and clarithromycin (CLR) remain unknown (7–10). As for antiretroviral (ARV) drugs, the interactions between nevirapine (NVP) and drug-metabolizing enzymes, namely, CYP3A4 and CYP2B6, are not completely understood (11, 12). For these reasons, there is an urgent need to clarify the hidden mechanisms of DDIs involved in these compounds. On the other hand, according to postmarketing surveillance, the fixed doses of some antimycobacterial drugs have been adjusted due to frequent side effects, inadequate efficacy, or the extent of adverse reactions (13–16). Hence, it is more practical that DDIs in in vitro or animal studies be reassessed after drugs have been released on the market.

In the present study, according to the guidelines on the investigation of DDIs published by the FDA and the European Medicines Agency (EMA), the effects of anti-TB drugs, novel candidates, anti-Mycobacterium avium-M. intracellulare complex (MAC) agents, and representative ARV drugs on CYP3A4 at clinically achievable concentrations were evaluated in vitro by using human enzymes and transporters, followed by ex vivo studies using a human hepatoma cell line, HepaRG (17, 18). Furthermore, to estimate the interactions of those drugs with P-gp, substrates were screened using a commercially available kit. For some substrates, P-gp inhibition tests were carried out using P-gp-expressing MDCK cells, an epithelial cell line of canine kidney origin.

MATERIALS AND METHODS

Drugs and chemicals.

RIF, isoniazid (INH), pyrazinamide (PZA), ethambutol (EMB), PAS, ethionamide (ETH), cycloserine (DCS), streptomycin (STR), amikacin (AMK), capreomycin (CAP), CFZ, LZD, roxithromycin (RXM), ampicillin (AMP), metronidazole (MTZ), ketoconazole (KTC), carbamazepine (CMZ), ofloxacin (OFX), cyclophosphamide (CPA), trypan blue (0.4%), and Eagle's minimum essential medium (EMEM) were purchased from Sigma-Aldrich (MO, USA). Prothionamide (PRO) and azithromycin (AZM) were purchased from LKT Laboratories (MN, USA). TAC and erythromycin (ERY) were purchased from Acros Organics (Geel, Belgium). Rifabutin (RFB), rifapentine (RFP), and efavirenz (EFV) were purchased from Toronto Research Chemicals (Ontario, Canada). NVP and ritonavir (RIT) were purchased from United States Pharmacopeial Convention (MD, USA). Raltegravir (RAL) was purchased from Selleck Chemicals. CLR, kanamycin (KAN), phenytoin (PHN), dexamethasone (DEX), ifosfamide (IFA), phenobarbital sodium (PB), verapamil hydrochloride (VER), quinidine sulfate dihydrate (QD), 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT), hydrochloric acid, sodium hydroxide, sodium dodecyl sulfate, dimethyl sulfoxide (DMSO), trypsin-EDTA solution with phenol red, and Dulbecco's phosphate-buffered saline (D-PBS) were purchased from Wako Pure Chemical Industries (Osaka, Japan). Spironolactone (SL) was purchased from Tokyo Chemical Industry (Tokyo, Japan). Moxifloxacin (MXF) was obtained from Kemprotec Limited (Middlesbrough, United Kingdom). Rifalazil (RLZ) was obtained from Kaneka Corporation (Hyogo, Japan). Levofloxacin (LVX) and DC-159a were obtained from Daiichi Sankyo Corporation (Tokyo, Japan). Gatifloxacin (GAT) was obtained from Kyorin Pharmaceutical (Tokyo, Japan). Ciprofloxacin hydrochloride monohydrate (CIP) was obtained from Bayer HealthCare (Osaka, Japan). Amoxicillin (AMX), meropenem (MEM), and potassium clavulanate (CLA) were supplied by Meiji Seika Pharma (Osaka, Japan). Enviomycin (EVM) was obtained from Asahi Kasei Pharma Corporation (Tokyo, Japan).

Screening for CYP3A4 inhibitors.

The CYP3A4-BFC (7-benzyloxy-trifluoromethylcoumarin) high-throughput inhibitor screening kit was purchased from BD Biosciences (MA, USA). This experiment was performed to measure the inhibitory effects of each agent on the metabolism of a nonfluorescent CYP3A4 substrate, 7-benzyloxy-trifluoromethylcoumarin (BFC), to a fluorescent metabolite, 7-hydroxyl-trifluoromethylcoumarin (HFC), according to the manufacturer's instructions, except for the plate-reading conditions, which were set at an excitation of 405 nm and an emission of 535 nm. The 50% inhibitory concentration (IC₅₀) values of each drug were calculated by linear interpolation. The standard curve of HFC was prepared over the range of 2,000 to 0.9 pmol. KTC and VER were used as positive-control inhibitors.

Cytotoxicity test.

Cytotoxicity testing was implemented by the modified MTT assay (19, 20). DMSO and sodium hydroxide in high concentrations were used as positive-control agents. Briefly, after the end of a differentiation period, HepaRG cells (7.2 × 10⁴ cells/well for a 96-well plate) were treated daily with and without test compound in triplicate and were incubated in a humidified 5% CO₂ atmosphere at 37°C for 48 h.

CYP3A4 induction and inhibition test using HepaRG cells.

A HepaRG culture kit for 50 plates, working growth medium 710, working differentiation medium 720, and induction medium 640 were purchased from BioPredic International (Rennes, France). Trypan blue (0.4%) was diluted with D-PBS at a final concentration of 0.05% before use. The procedures of thawing, seeding, maintenance, differentiation, and freezing of HepaRG cells were performed according to the instructions for this kit. After the end of the differentiation period, the cells were treated daily with and without test compound at three concentrations in triplicate and were incubated in a humidified 5% CO₂ atmosphere at 37°C for 48 h. In principle, the highest concentration of each drug was set at >2× the maximum drug concentration achieved in human blood (C_max) at the generally recommended dosage, because the drug concentration in the portal vein during the process of absorption is generally greater than the C_max value after oral administration (18). The C_max values were obtained from pharmaceutical package inserts, the Handbook of Anti-Tuberculosis Agents (21), and previously published reports (65–67). In brief, the cells were incubated in the presence of 50 μM midazolam in EMEM for 1 h at 37°C. Each supernatant was transferred into a new glass tube, followed by 100 μl of methanol and 100 μl of 0.05 μg/ml diazepam dissolved in acetonitrile. The mixed suspension was evaporated to dryness at 60°C for 80 min. Thereafter, a 200-μl aliquot of methanol was added, sonicated for 2 s, and vortexed for 5 s. Finally, the solution was filtered through a 0.45-μm syringe filter and injected in a high-performance liquid chromatography mass spectrometry (LC/MS) system.

The average percentage of 1-hydroxymidazolam production in the wells treated with the test compound was divided by that in wells treated with 0.1% DMSO. Each value was expressed as the mean and standard deviation (n = 6). In order to assess the concentration dependency, statistical analysis was conducted by one-way layout regression analysis using Microsoft Office Excel 2010 software. Statistical significance was set at P values of <0.05, <0.01, and <0.001. In order to assess the reactivity of CYP3A4 to test compounds, the criteria were defined as shown in Table 1.

TABLE 1.

Criteria for assessing the reactivity of CYP3A4 to agents in HepaRG cells

Reactivity category	% reactivity
Very strong induction	≥400
Strong induction	300–400
Moderate induction	200–300
Weak induction	150–200
Little or no effect	80–150
Weak inhibition	60–80
Moderate inhibition	40–60
Strong inhibition	20–40
Very strong inhibition	≤20

LC/MS conditions.

A Shimadzu LCMS-2010EV equipped with an electrospray ionization source consisting of a solvent manager, an autosampler, and a column oven was used and set in the positive-ion electron impact mode. The operating parameters were as follows: interface curved desolvation line temperature, 250°C; interface heat block temperature, 200°C; nebulizing gas flow, 1.5 liter/min; and detector voltage, 1.5 kV. A Capcell Pak C₁₈ MG II S-3 column was used for the stationary phase (50 by 2.0 mm, 2.7-μm particle size). The cooler and column temperatures were maintained at 4°C and 40°C, respectively. The mobile phase was composed of 0.3% formic acid (phase A) and methanol with 0.3% formic acid (phase B). The concentrations in mobile phase B were linearly increased from 45% to 100% over 7 min, held at 100% for another 2 min, and then returned to the initial conditions. The flow rate was 0.2 ml/min, and the run time was 18 min.

MS was performed in the multiple-reaction monitoring mode at ions m/z 341.77 for 1-hydroxymidazolam and m/z 284.74 for diazepam (used as an internal standard). Stock solutions of 1-hydroxymidazolam (500 μM) and diazepam (0.5 mg/ml) were dissolved in methanol and acetonitrile, respectively. Working solutions were prepared by diluting with the same solvents. Standard curves were prepared by spiking EMEM with the working solutions. The serial dilutions for the standard curves were prepared with methanol, and the aliquot was transferred into glass tubes, followed by 500 μl of EMEM containing 0.1% DMSO. The subsequent procedure was the same as for preparing the samples. The retention times for 1-hydroxymidazolam and diazepam were 2.12 min and 7.23 min, respectively. Standard curves for 1-hydroxymidazolam were constructed over the range from 20 to 0.05 μM. This method was validated in our research institute. Each correlation coefficient was >0.9991. The lower limit of quantification was 5 nM.

Screening for the substrates of human P-gp.

The experiment was carried out according to the method previously reported (22). ABC transporter ATPase assay reagents kit was purchased from Nacalai Tesque (Kyoto, Japan). Human P-gp and control membranes were purchased from BD Biosciences (MA, USA). VER and QD were used as positive-control substrates. The half-saturation concentration (Michaelis content [K_m]) and maximum ATP hydrolysis rate (V_max) were calculated from a Hanes-Woolf plot.

P-gp inhibition test using P-gp-expressing MDCK cells.

We commissioned this experiment to Cyprotex (Cheshire, United Kingdom). MDR1-MDCK cells were obtained from the National Institutes of Health (Rockville, MD, USA). Following culture, monolayers were prepared by rinsing both basolateral and apical surfaces twice with Hanks' balanced salt solution (pH 7.4) at 37°C. The cells were then incubated with buffer (pH 7.4) in both apical and basolateral compartments for 40 min at 37°C and 5% CO₂, with a relative humidity of 95% to stabilize the physiological parameters. For the apical-to-basolateral study, buffer (pH 7.4) was removed from the apical compartment and replaced with loperamide dosing solutions before being placed in the companion plates. The solutions were prepared by diluting loperamide in DMSO with buffer to give a final loperamide concentration of 5 μM (final DMSO concentration adjusted to 1%). The fluorescent integrity marker lucifer yellow was also included in the dosing solution. The experiment was performed in the presence and absence of the test compound (applied to both the apical and basolateral compartments). For the basolateral-to-apical study, loperamide (final concentration, 5 μM) was placed in the basolateral compartment. Incubations were carried out in an atmosphere of 5% CO₂ with a relative humidity of 95% at 37°C for 60 min. After the incubation period, the companion plate was removed and the apical and basolateral samples were diluted for analysis by LC tandem MS (LC-MS/MS). KTC and cyclosporine were used as positive-control inhibitors. The concentrations at which the test compounds were assessed were 0 to 10 μM for CFZ and 0 to 500 μM for LZD. Loperamide was quantified by LC-MS/MS under Cyprotex generic analytical conditions. The integrity of the monolayers throughout the experiment was checked by monitoring lucifer yellow permeation using fluorimetric analysis.

RESULTS

Screening for CYP3A4 inhibitors.

At first, the inhibitory effects of anti-TB drugs, novel candidates, macrolides, and ARV drugs on human CYP3A4 activity were assessed by using the screening kit. As shown in Table 2, rifamycins, namely, RIF, RFB, RFP, and RLZ, showed inhibitory effects on the metabolism of the CYP3A4 substrate BFC, with IC₅₀s of 19.7 μM, 3.96 μM, 22.8 μM, and 2.76 μM, respectively, which indicates that these compounds are potential inhibitors and/or substrates of CYP3A4. Likewise, all the intended macrolides exhibited inhibitory effects on CYP3A4 at several levels (Table 2). Compared to previous studies using human liver microsomes (HLMs), the inhibitory concentration for INH presented here was evidently low (IC₅₀, 0.9 μM) (Table 2) (23, 24). Both PZA and EMB exerted no effects on CYP3A4 at any given concentration (see Table S1 in the supplemental material). Regarding second-line anti-TB drugs, ETH and PRO showed inhibitory effects on CYP3A4, with IC₅₀s of 78.4 μM and 36.6 μM, respectively (Table 2). PAS, TAC, and LZD exhibited inhibitory effects on CYP3A4 at concentrations >100 μM (Table 2). Among the fluoroquinolones (FQs) tested, only CIP had an inhibitory effect on CYP3A4 at a high concentration (272 μM), in accordance with a previous report (see Table S1) (25). No remarkable effects were observed in aminoglycosides, cyclic peptides, β-lactams, β-lactamase inhibitors, MTZ, or DCS (see Table S1). In agreement with a previous report, CFZ was found to be a CYP3A4 inhibitor, with an IC₅₀ of 1.69 μM, which is nearly equal to the C_max value at the commonly used therapeutic dosage in humans (Table 2) (4, 26). In terms of the three classes of ARV drugs, inhibitory effects on CYP3A4 were observed in NVP, EFV, and RIT, but not RAL, belonging to integrase inhibitors (Table 2; see also Table S1 in the supplemental material). Importantly, the inhibitory effects of EFV and RIT on CYP3A4 were observed at clinically relevant concentrations (15.3 μM and <0.07 μM, respectively) (Table 2).

TABLE 2.

Inhibitory effects of each agent on human CYP3A4 activity determined by in vitro screening

Agent	Abbreviation	Cmax (μg/ml), p.o. dosagea	IC50 (μg/ml)b	IC50 (μM)
Rifampin	RIF	7, 450 mg	16.2	19.7
Rifabutin	RFB	0.375 ± 0.267, 300 mg	3.36	3.96
Rifapentine	RFP	15.05 ± 4.62, 300 mg	20.0	22.8
Rifalazil	RLZ	0.0264 ± 0.011, 25 mg	2.60	2.76
Isoniazid	INH	3–8, 300 mg	0.12	0.90
Ethionamide	ETH	6–12.5, 500 mg	13.0	78.4
Prothionamide	PRO	∼1.8× lower than for ETH	6.60	36.6
p-Aminosalicylic acid	PAS	21.4 (fasting) or 32.5 (fed), 6 g	51.2	335
Thiacetazone	TAC	1.59 ± 0.47, 150 mg	36.2	153
Linezolid	LZD	21.2, 600 mg	406	1204
Clofazimine	CFZ	0.7, 100 mg (multiple doses)	0.80	1.69
Clarithromycin	CLR	2.24, 400 mg	10.4	13.9
Erythromycin	ERY	0.82, 200 mg	2.80	3.82
Azithromycin	AZM	0.66, 1,200 mg	32.7	43.6
Roxithromycin	RXM	4.94 ± 1.3, 150 mg	2.90	3.46
Nevirapine	NVP	3.9, 400 mg	28.0	105
Efavirenz	EFV	2.54 ± 0.62, 600 mg	4.82	15.3
Ritonavir	RIT	15.83, 600 mg	<0.05	<0.07

^aC_max values were obtained from pharmaceutical package inserts, the Handbook of Anti-Tuberculosis Agents (21), and previously published reports. p.o., per os.

^bIC₅₀, 50% inhibitory concentration.

Effects of each agent on intrinsic CYP3A4 activity in HepaRG cells.

According to the guidelines published by the FDA and EMA, we next conducted ex vivo studies using the HepaRG cell line. First, we examined the reactivity of the intrinsic CYP3A4 enzyme to various classes of CYP3A4 inducers, namely, CMZ, PHN, DEX, CPA, IFA, and SL (see Table S2 in the supplemental material). Of those, all but the alkylating drugs CPA and IFA significantly enhanced the metabolism of midazolam to 1-hydroxymidazolam via CYP3A4 (see Table S2). Alkylating drugs tended to increase CYP3A4 activity in HepaRG cells by up to 50%, whereas these magnitudes of induction are considered not to be meaningful on the basis of the criteria in this study (Table 1; see also Table S2 in the supplemental material). In contrast, the well-known inhibitors CLR and RIT showed inhibitory effects on CYP3A4 (Table 3; see also Table S2). PZA, AZM, and FQs (except for CIP, which is known to be a CYP3A4 inhibitor) possessed little or no effects on CYP3A4 at any given concentration (see Table S2). CIP exhibited an inhibitory effect on CYP3A4 at high concentrations that were negligible in clinical practice (see Table S2). In accordance with the previous report using human primary hepatocytes, the descending order of the inductive activity of CYP3A4 at both 5 μM and 10 μM was RIF > RFP ≫ RFB (27). Meanwhile, the CYP3A4 inductive effect of RIF at the concentration close to the C_max value was approximately equal to that of RFP in HepaRG cells (Table 3). Intriguingly, despite belonging to the rifamycins, RLZ showed weak inhibition of CYP3A4 activity at >0.25 μM (Table 3). In view of the results in HLMs reflecting direct reaction with enzymes, RIF, RFB, and RFP were proven to cause the autoinduction of CYP3A4, accompanied by self-decomposition in the liver. As for INH, an inhibitory effect on CYP3A4 was observed at 100 μM, which was much higher than those in HLMs (Table 3). Both ETH and PRO showed a slight effect on CYP3A4 in HepaRG cells in contrast to those in HLMs (see Table S2). TAC exerted a weak inhibitory effect at high concentrations that were negligible in clinical practice (Table 3). MTZ exhibited a weak inhibitory effect at >50 μM (Table 3). PAS tended to modestly inhibit CYP3A4 activity in a dose-dependent manner, but no remarkable inhibition was observed at as high as 900 μM (see Table S2). Both DCS and LZD barely showed inhibitory effects at any given concentration (Table 3; see also Table S2 in the supplemental material). Notably, CFZ exhibited a weak inductive effect on CYP3A4 at >0.25 μM in spite of the inhibitory effect at 1.69 μM in HLMs, which is consistent with the C_max value (about 1.84 μM) in leprosy patients receiving 100 mg daily with multiple doses, suggesting that CFZ autoinduces CYP3A4 in the human liver (Table 3) (26). Regarding ARV drugs, EFV had weak to strong inductive effects on CYP3A4 in HepaRG cells (Table 3). RAL exerted no effects on CYP3A4 at concentrations up to 20 μM (see Table S2). Contrary to our expectations, NVP, which is known as a CYP inducer, showed weak to moderate inhibition of intrinsic CYP3A4 activity (Table 3) (11).

TABLE 3.

Inductive or inhibitory effects of each agent on intrinsic CYP3A4 activity in HepaRG cells^a

Agent (P)b	Mean ± SD effect (%) by concn (mmol):																Close to Cmax (µM)c
	0	0.1	0.25	0.5	1	5	10	18	20	25	40	50	60	100	120	200
RIF (<0.001)	100 ± 5.92				313 ± 10.7	384 ± 30.3	430 ± 30.6		461 ± 15.0		473 ± 17.8						10
RFB (<0.001)	100 ± 2.05			173 ± 17.5	195 ± 18.1	169 ± 18.0	166 ± 9.52										0.5
RFP (<0.001)	100 ± 10.2					330 ± 32.0	357 ± 35.6	413 ± 20.0	473 ± 14.3								18
RLZ (<0.001)	100 ± 7.52		61.8 ± 1.95		66.0 ± 7.61	59.3 ± 4.57	57.9 ± 10.8										0.025
INH (<0.001)	100 ± 8.98						95.9 ± 11.1					85.0 ± 8.08		63.3 ± 9.54			10–50
TAC (<0.001)	100 ± 6.70					102 ± 3.12				90.8 ± 5.58		75.3 ± 5.27					5
CFZ (<0.001)	100 ± 5.34		153 ± 6.12		178 ± 7.63	210 ± 11.9											0.25
MTZ (<0.01)	100 ± 3.11									84.9 ± 2.29		68.1 ± 6.87		68.2 ± 1.58			25
LZD	100 ± 9.31												96.4 ± 15.5		95.4 ± 19.0	97.5 ± 7.31	60
NVP (<0.01)	100 ± 10.4						44.9 ± 3.29		38.8 ± 3.10		38.1 ± 3.77						10–20
EFV (<0.001)	100 ± 10.8				154 ± 20.8	338 ± 14.2	386 ± 28.3										10
RIT	100 ± 4.88	4.01 ± 0.68		6.16 ± 0.78	6.37 ± 2.20												10–20

^aThe experiments were carried out in triplicate, and the representative data of two separate experiments are shown.

^bStatistical significance was determined by one-way layout regression analysis (significant P values are shown in parentheses). Each agent name is abbreviated as indicated in Materials and Methods.

^cC_max values were obtained from pharmaceutical package inserts, the Handbook of Anti-Tuberculosis Agents (21), and previously published reports.

Effects of each agent on RIF-mediated CYP3A4 activity in HepaRG cells.

Almost all compounds showed a similar result to that obtained from the CYP3A4 induction test shown in Table 3 and Table S2 in the supplemental material (see Table 4 and Table S3 in the supplemental material for comparisons). Meanwhile, discrepancies between the induction and inhibition tests were observed in NVP and RLZ. NVP exerted no significant inhibitory effects on RIF-mediated CYP3A4, even at 40 μM (Table 4). Besides, RLZ enhanced RIF-induced CYP3A4 activity in a synergistic fashion at any given concentration (Table 4). These findings led us to attempt to determine the nuclear receptor related to the CYP3A4 inhibitory effects of these compounds. For instance, RIF activates CYP3A4 activity in hepatocytes, mainly via binding to pregnane X receptor (PXR) (28). EFV, PHN, and CMZ have been known to preferentially activate constitutive androstane receptor (CAR), which leads to the induction of CYPs (29). The induction of CYP3A4 by 1 mM PB has been reported to be regulated by CAR but not PXR (30). As anticipated, the inhibitory effects of NVP on CYP3A4 activity induced by EFV, CMZ, and PB, except for rifamycins, were observed at several levels, which indicates that NVP exhibits an inhibitory effect on CAR-mediated CYP3A4 activity (Fig. 1). In contrast, RLZ affected neither RIF-induced nor EFV-induced CYP3A4 activity at any given concentration, suggesting that the inhibitory effect of RLZ on intrinsic CYP3A4 has no relationship with the above mechanism (data not shown). Besides these results, both INH and TAC showed somewhat higher inhibitory potencies against RIF-mediated CYP3A4 activity than against intrinsic CYP3A4 activity (Table 4). Similar to the induction test, CFZ synergistically enhanced RIF-mediated CYP3A4 activity at 0.25 to 1 μM, as did RFL (Table 4).

TABLE 4.

Inductive or inhibitory effects of each agent on RIF-mediated CYP3A4 activity in HepaRG cells^a

Agent (P)b	Mean ± SD effect (%) by concn (mmol):												Close to Cmax (µM)c
	0	0.0125	0.025	0.25	1	5	10	20	25	40	50	100
RLZ	100 ± 7.20	132 ± 14.1	129 ± 13.0	132 ± 11.9									0.025
INH (<0.01)	100 ± 1.93						95.0 ± 6.24				74.3 ± 3.96	54.6 ± 3.10	10–50
TAC (<0.01)	100 ± 4.73					86.1 ± 4.84			79.3 ± 13.2		67.2 ± 7.65		5
CFZ (<0.001)	100 ± 16.3			129 ± 8.06	130 ± 3.46	94.0 ± 3.51							0.25
NVP (<0.001)	100 ± 4.87						104 ± 7.29	101 ± 7.42		87.0 ± 1.88			10–20

^aThe experiments were carried out in triplicate, and the representative data of two separate experiments are shown.

^bStatistical significance was determined by one-way layout regression analysis (significant P values are shown in parentheses). Each agent name is abbreviated as indicated in Materials and Methods.

^cC_max values were obtained from pharmaceutical package inserts, the Handbook of Anti-Tuberculosis Agents (21), and previously published reports.

FIG 1.

Differences in the inhibitory effect of nevirapine (NVP) on CYP3A4 induction mediated by various CYP3A4 inducers. The inhibitory effects of NVP were estimated by the amount of production of 1-hydroxymidazolam. The experiments were carried out in triplicate, and the representative data (means ± standard deviations) of two separate experiments are shown. RIF, rifampin; RFB, rifabutin; EFV, efavirenz; PB, phenobarbital; CMZ, carbamazepine. The numbers in parentheses represent the exposed micromolar concentrations for each CYP3A4 inducer.

Screening for the substrates and inhibitors of human P-gp.

The macrolides CLR and AZM enhanced ATP hydrolysis, with K_m values of 8.53 μM and 11.3 μM, respectively, implying that these compounds interact with human membrane P-gp (Table 5). Among the intended antimycobacterial drugs, only CFZ was found to be a substrate for P-gp, with a K_m value of 0.35 μM in the in vitro screening (Table 5; see also Table S4 in the supplemental material). Meanwhile, CFZ scarcely inhibited the membrane permeation of loperamide, which results from the transport activity of P-gp in human P-gp-expressing MDCK cells (IC₅₀ > 10 μM), as well as from LZD (IC₅₀ > 500 μM) (data not shown). These results indicate that CFZ is a substrate for human P-gp. As for ARV drugs, NVP, EFV, and RIT were found to be substrates for P-gp, with V_max values of 36.2, 14.2, and 20.6 nmol/min, respectively (Table 5).

TABLE 5.

Substrates for human P-glycoprotein determined by in vitro screening^a

Agent	Km inb:		Vmaxb (nmol/min)
	μg/ml	μM
CFZ	0.17	0.35	8.84
NVP	76.2	286	36.2
EFV	7.73	24.5	14.2
RIT	0.06	0.09	20.6
CLR	6.38	8.53	28.7
AZM	8.46	11.3	9.07

^aExperiments were performed in duplicate at least two times. Data are expressed as average values from two separate experiments. Each agent name is abbreviated as indicated in Materials and Methods.

^bK_m (Michaelis content) and V_max values were calculated from a Hanes-Woolf plot.

DISCUSSION

CYP3A4 is by far the most abundant isoform of CYPs in the human liver and is reckoned to be a key enzyme related to DDIs occurring in the treatment of TB, especially drug-resistant TB, and coinfection with TB/AIDS (31). To date, the primary human hepatocyte has been generally used as a gold standard to determine the effects of test compounds on drug-metabolizing enzyme activities during preclinical development. Their usage, however, is limited for several reasons: high cost, insufficient availability of donor organs, interdonor functional variability, and limited life span (32). Meanwhile, the well-established human hepatoblastoma, HepG2, can also be used for the same application and cytotoxicity testing because of its good proliferation, immortality, and stable viability; however, the intrinsic enzyme activities and the reactivity of CYPs, especially CYP3A4 and CYP2B6, to CYP modulators are entirely inadequate, even in the three-dimensional culture of HepG2 cells (33–35). Similarly, while immortalized human hepatic cell lines, i.e., Fa2N-4 cells and ADV-1 cells that are derived from BC2 cells, have been reported since 2000, the inherent characteristics of those lines restrict their routine application in the drug discovery setting (36, 37). For instance, the former does not respond to a prototypical inducer of CYP2B6, and so far, the functions of the latter have not been validated in detail compared with HepaRG cells. In contrast, the HepaRG cell line has been reported to be useful for evaluating the effects of drugs on CYP3A4 that are superior to those of the cell lines (38). For these reasons, we evaluated the effects of 30 anti-TB drugs, novel candidates, macrolides, and ARV drugs on human CYP3A4 activity using HepaRG cells in a single experimental system. There were no cytotoxic activities at any given concentration, except for EFV at a concentration of 50 μM with and without RIF (data not shown). Throughout the ex vivo studies using HepaRG cells, almost all compounds showed the expected effects on both intrinsic and RIF-mediated CYP3A4 activity. Among the CYP3A4 inhibitors determined by the in vitro screening, INH, PAS, and macrolides showed less potent CYP3A4 inhibitory effects in HepaRG cells than in HLMs. Thioamides showed little or no influence on CYP3A4 activity in HepaRG cells in contrast with the results in HLMs. Importantly, in the in vitro screening using HLMs, drug-metabolizing enzymes were hardly induced within the incubation period (30 min), which indicates that autoinducers are likely to be mistaken for inhibitors. Indeed, the well-established CYP3A4 inducers, such as rifamycins and EFV, were determined to be CYP3A4 inhibitors (Table 2). Nevertheless, this screening assay is useful for finding inhibitors quickly in practice, and it is far superior to ex vivo experiments in terms of the time required, simple procedures, and total cost.

Second, based on our results, we attempted to clarify the unrevealed mechanisms of DDIs in the treatment of TB and AIDS. A previous clinical report regarding DDIs between CFZ and INH (39) indicated that INH caused increases in CFZ plasma and urine concentration by up to 3-fold and 1.9-fold, respectively, except in one patient, suggesting that INH inhibits the CYP3A4-mediated metabolism of CFZ and consequently leads to the accumulation of unchanged CFZ in the human body. Similarly, in an experimental mouse model, a modest increase in the plasma concentration of CFZ when combined with INH has been demonstrated (40). In addition, the DDI between CFZ- and RIF-based regimens for leprosy has been reported (41). Notably, CFZ delayed the absorption of RIF and decreased the C_max value and area under the time-concentration curve (AUC) for RIF, which might result from the additional induction of intestinal and hepatic CYP3A4 by CFZ, as shown in Table 4. This inference might apply to other reports regarding DDIs between CFZ and physiologically active substances, such as steroid hormone and vitamin A, both of which are known to be CYP3A4 substrates (5). These results might support the interpretation of there being unknown mechanisms operating on DDIs between CFZ and anti-TB drugs, i.e., INH and RIF (39, 42, 43). In regard to the novel drugs and candidates undergoing preclinical and/or clinical trials, bedaquiline, PA-824, TBA-354, and sutezolid have been reported to be substrates for CYP3A4 (44, 45). In particular, PA-824 and TBA-354 have been shown to have weak inhibitory effects on CYP3A4 (46). Therefore, it should be kept in mind that the pharmacological influence of CFZ on the metabolism of coadministered drugs is inevitable, and careful clinical monitoring might be warranted throughout the development of CFZ-containing regimens for the treatment of MDR-/XDR-TB and other mycobacterial infections.

Previously, the effects of NVP on CYP3A4 mRNA expression and CYP enzyme activities have been examined using three lots of cryopreserved human hepatocytes (47). While NVP has caused up to 21-fold increases in CYP3A4 mRNA expression, no obvious induction of testosterone 6-β-hydroxylase activity has been detected in two out of three lots. In spite of the differences in the experimental designs, these results accord with our findings. Hence, we inferred that the DDIs between NVP and CYP3A4 modulators, such as azoles, HIV protease inhibitors, and oral contraceptives, would depend on CYP2B6 induction by NVP, because these CYP3A4 modulators have been proven to be substrates for CYP2B6 (29, 48, 49). However, the increased metabolism of ERY mediated by NVP cannot be explained with this hypothesis, because macrolides have no bearing on CYP2B6 activities (data not shown) (50). In the light of previous studies, NVP might exhibit both inductive and inhibitory effects on CYP3A4 activity on a case-by-case basis (47). Further studies are needed to solve this issue.

Nowadays, unexpected DDIs between LZD and CYP3A4 modulators have caused considerable concern regarding the long-term treatment of infections for MDR-/XDR-TB (7, 10, 51). In the present study, LZD showed an inhibitory effect on CYP3A4, with an IC₅₀ of 1.2 mM (406 μg/ml) in HLMs, while no inductive and inhibitory effects on CYP3A4 were observed even at a high concentration (200 μM) in HepaRG cells, regardless of the presence of RIF (Tables 2 and 3; see also Table S3 in the supplemental material). Additionally, there was no inhibitory effect on P-gp even at a concentration of 500 μM in both the in vitro screening and P-gp-expressing MDCK cells (see Table S4 in the supplemental material). These results support a previous report using human primary hepatocytes (52). We also assessed the effect of LZD on CYP2B6 using an in vitro screening kit, but no inhibition was observed at 500 μM (data not shown). Taken together, LZD is a potential substrate with weak affinity to CYP3A4, similar to that of sutezolid. With the aforementioned assumption, the following mechanisms of (i) increased clearance of LZD in both healthy volunteers and patients coadministered RIF, (ii) increased serum AUC from 0 to 12 h (AUC_0–12) and decreased elimination of LZD in patients receiving CLR, (iii) potential DDI between LZD and aztreonam, and (iv) inhibition of LZD metabolism in the presence of a well-known CYP3A4 inhibitor KTC might be elucidated (7, 10, 53, 54). Based on this concept, both thioamides and PAS might be substrates for CYP3A4 enzymes.

Unexpectedly, CFZ exerted no inhibitory effect on P-gp activity in human P-gp-expressing MDCK cells with concentrations as high as 10 μM, like LZD, whereas the opposite data have been reported using human lung cancer cells or erythroleukemia cells (data not shown) (55, 56). This discrepancy is due to the differences in the origins of the cells tested. If so, similar experiments should be performed using cells derived from the human liver and small intestine.

One limitation of this study is that we did not assess the DDIs related to organic anion-transporting polypeptides (OATPs) that play an important role in hepatic uptake of pharmaceutical drugs. Recently, some drugs available for the treatment of TB and AIDS, e.g., rifampin and ritonavir, have been reported to be OATP1B1 inhibitors and/or substrates (17, 57–59). Actually, the mRNA expression levels of the SLCO1B1, SLCO2B1, and SLCO1B3 genes have been demonstrated in differentiated HepaRG cells as well as in primary hepatocytes, implying that OATP1B1, OATP2B1, and OATP1B3 are active in HepaRG cells (60). Hence, assessing the interactions between the drugs and OATPs is of importance for understanding drug transport across the biological membrane in HepaRG cells. Further studies are warranted to explore hidden DDIs between anti-TB drugs and OATPs.

Recently, mouse models using PXB mice and PXR-humanized mice have been proven to be powerful tools for investigating the DDIs relevant to human drug-metabolizing enzymes, such as CYP3A4 and CYP2B6 (61–63). In accordance with these studies, the inductive effect of RIF and inhibitory effect of KTC on CYP3A4 activity were observed in HepaRG cells, implying that the ex vivo studies reflect experiments using mouse models (62, 64) (Table 3; see also Table S3 in the supplemental material). Further investigation for assessing the data correlation among these models is desired.

In conclusion, HepaRG cells are useful for estimating the inductive and/or inhibitory effects of drugs and chemicals on CYP3A4 activity. Whereas the ex vivo study itself is inconclusive, it sheds light on the pathway for developing safe regimens for the treatment of TB, MAC, and AIDS. When officially determining a suitable dose in humans, the effects of pharmaceutical drugs on drug-metabolizing enzymes and membrane transporters that are likely to be the causative factors involved in DDIs should be reevaluated at clinically achievable concentrations.