An S‐warfarin and AZD1981 interaction: in vitro and clinical pilot data suggest the N‐deacetylated amino acid metabolite as the primary perpetrator

Abstract

Aim

AZD1981 is an orally bioavailable chemoattractant receptor‐homologous molecule expressed on Th2 cells (CRTh2) receptor antagonist progressed to phase II trials for the treatment of allergic asthma. Previously performed in vitro human hepatocyte incubations identified N‐deacetylated AZD1981 as a primary metabolite. We report on metabolite exposure from a clinical excretion balance, on in vitro studies performed to determine the likelihood of a metabolite‐dependent drug–drug interaction (DDI) and on a clinical warfarin DDI study. The aim was to demonstrate that N‐deacetylated AZD1981 is responsible for the observed interaction.

Methods

The excretion and biotransformation of [¹⁴C]‐AZD1981 were studied in healthy male volunteers, and subsequently in vitro cytochrome P450 (CYP) inhibition and hepatocyte uptake investigations were carried out with metabolites and the parent drug. A clinical DDI study using coadministered twice‐daily 100 mg and 400 mg AZD1981 with 25 mg warfarin was performed.

Results

The excretion balance study showed N‐deacetylated AZD1981 to be the most abundant metabolite present in plasma. In vitro data revealed the metabolite to be a weak CYP2C9 time‐dependent inhibitor, subject to more active hepatic uptake than the parent molecule. Clinically, the S‐warfarin area under the plasma concentration–time curve increased, on average, 1.4‐fold [95% confidence interval (CI) 1.22, 1.50] and 2.4‐fold (95% CI 2.11, 2.64) after 100 mg (n = 13) and 400 mg (n = 11) AZD1981 administration, respectively. In vitro CYP inhibition and hepatocyte uptake data were used to explain the interaction.

Conclusions

N‐deacetylated AZD1981 can be added to the small list of drug metabolites reported as sole contributors to clinical drug–drug interactions, with weak time‐dependent inhibition exacerbated by efficient hepatic uptake being the cause.

What is Already Known about this Subject

• Only a small number of drug metabolites have been implicated as primary perpetrators of clinical drug–drug interactions (DDIs). Regulatory guidance states that metabolites present at greater than 10% of total drug‐related exposure in humans, or 25% of the parent exposure, should trigger further investigation of their inhibitory potency.

What this Study Adds

• From an early clinical [¹⁴C]‐AZD1981 excretion balance study, the major circulating metabolite of AZD1981, N‐deacetylated AZD1981, was a borderline case for DDI concern based on regulatory guidance. In vitro data showing active hepatic uptake of the parent and metabolite, coupled with the greater reversible cytochrome P450 (CYP) 2C9 inhibitory potency of the metabolite and weak time‐dependent CYP2C9 inhibition, raised concerns of a possible S‐warfarin interaction. A clinical study was performed and the observed DDI with S‐warfarin was rationalized from the combination of in vitro and clinical data for the metabolite. This study highlights the need for vigilance in predicting likely drug–drug interactions and shows that detailed analyses are required early in clinical drug development when considering possible interactions involving drug metabolites.

Tables of Links

TARGETS
G protein‐coupled receptors 2	Transporters 3
prostaglandin D2 receptor 2 (PTGDR2)	OATP

LIGANDS
prostaglandin D2 receptors (PGD2)

These Tables list key protein targets and ligands in this article that are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 1, and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 2, 3.

Introduction

AZD1981 (Figure 1) is a potent orally bioavailable chemoattractant receptor‐homologous molecule expressed on Th2 cells [CRTh2; also known as prostaglandin D₂ receptor 2 (PTGDR2)] receptor antagonist. CRTh2 is a G protein‐coupled receptor that is activated by prostaglandin D₂ and several of its metabolites, formed downstream of arachidonic acid. It is expressed on adaptive and innate immune cells associated with the allergic response, and consequently AZD1981 has progressed to phase II clinical trials for therapeutic use in the treatment of allergic asthma. Prior to clinical studies, three metabolites (AZD1981 acyl‐glucuronide, AZD1981 sulphoxide and the AZD1981 N‐deacetylated amino acid) had been identified following a human hepatocyte incubation in vitro 4.

Figure 1.

Structures of AZD1981, N‐deacetylated AZD1981 and sulphaphenazole

When considering the role that individual metabolites play in clinically relevant drug‐metabolizing enzyme or transporter inhibition, it is imperative to take into account both the concentration and inhibitory potency of metabolites 5, 6, 7. A review covering 102 cytochrome P450 (CYP) inhibiting drugs highlighted that, with 80 metabolites characterized at steady state and 78% of those metabolites having exposures greater than 10% compared with the parent drug, there is a need to evaluate metabolites as potential causes of drug–drug interactions (DDIs) 8. A separate analysis of 129 CYP inhibitors (81% being reversible) drew similar conclusions 9. Nevertheless, the overall risk of DDIs caused exclusively by drug metabolites may be considered to be low, with 90% of all drugs (and their metabolites) on the US market not being clinically important CYP inhibitors, and metabolites identified as the primary cause for only 2% of those therapeutic treatments exhibiting DDIs 10. In line with this, it has been proposed that only when metabolite concentrations are close to those of the parent area under the plasma concentration–time curve (AUC) should there be a trigger to study the capability of the metabolite to cause drug‐metabolizing enzyme inhibition in vitro, with two exceptions: when metabolites are more lipophilic than the parent drug and when metabolites are likely to be causative of mechanism‐based inactivations 10. However, both the European Medicines Agency (EMA) and the Food and Drug Administration (FDA) guidance indicate a more conservative approach than this, stating that metabolites which are present at 25% of the parent AUC or 10% of total drug‐related exposure should trigger further characterization with regard to DDI risk 11.

While there are publications detailing how metabolites of some CYP inhibitors have a significant impact in addition to the parent drug 7, 12, examples of metabolites being the main contributor to clinically relevant DDIs reported in the literature appear limited to bupropion, amiodarone, sertraline and gemfibrozil. Threohydrobupropion and erythrohydrobupropion are present at similar or higher circulating concentrations in comparison to bupropion and are more potent inhibitors of CYP2D6 13. Consequently, both metabolites have been suggested to be the cause of the clinical DDI between bupropion and the CYP2D6 substrate desipramine 8, 13, 14, 15. Interactions arising from amiodarone and sertraline administration can only be rationalized when their respective N‐dealkylated metabolites (each being present at higher exposures than the parent drugs) are considered as CYP inhibitors 8. Gemfibrozil is metabolized to an acyl glucuronide metabolite, whose subsequent metabolism causes mechanism‐based inhibition of CYP2C8 16, 17, 18, resulting in serious interactions with cerivastatin 19 and repaglinide 20. It is clear that the issue of DDIs requires close and detailed attention above the common, albeit increasingly sophisticated, prediction methods focused on parent drugs as perpetrators of single drug‐pair interactions 21.

The work detailed in the current manuscript covers an early clinical excretion balance study that gave information on the major circulating metabolites and those found in excreta, the relationship of these data to the earlier in vitro human hepatocyte metabolite identification work and subsequent investigations into the possibility of a clinically relevant metabolite‐dependent DDI risk. Finally, a clinical AZD1981–warfarin interaction study is described and all the data are discussed with reference to the available regulatory guidance documents.

Methods

Drugs and chemicals

Unless otherwise specified, all reagents were purchased from Sigma‐Aldrich (St Louis, MO, USA). Human CYP2C9/CYP reductase coexpressed in Escherichia coli (Cypex, Dundee, UK) were used to assess the reversible inhibition potential of evaluated compounds. A 7‐hydroxy warfarin (7‐OH warfarin, Toronto Research Chemicals, North York, ON, Canada) standard curve was used for metabolite quantification. Human liver microsomes for the time‐dependent inhibition study were obtained from the BD Gentest 150 donor UltraPool (Lot no. 38 289, Franklin Lakes, NJ, USA) at a concentration of 20 mg ml^–1 protein, and NADPH for these incubations was purchased from Roche Diagnostics GmbH (Mannheim, Germany).

For in vitro hepatic uptake experiments, Liverpool™ 10‐donor mixed gender (five female/five male) pooled cryopreserved human hepatocytes (lot IRK, BioreclamationIVT, Baltimore, MD, USA) were used. Hepatocyte incubations were performed in Williams medium E containing L‐glutamine (2 mM) buffered to pH 7.4 with 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES) (25 mM), and for the separation of cells from incubation media after incubation with AZD1981 or N‐deacetylated AZD1981, microtubes (Beckman 0.5 ml, Fisher Scientific GTF AB, Västra Frölunda, Sweden) were used. AstraZeneca rigorously ensures that human biological samples are handled in a responsible and ethical manner, requiring donor informed consent designed to protect the rights and expectations of donors and families. No human biological sample can be acquired unless it is from an approved source which has appropriate ethical approval procedures in place. Both BioreclamationIVT and BD Gentest are approved suppliers for AstraZeneca and, as such, retain the appropriate ethical approval documentation.

A phase I study to assess the metabolism, excretion and pharmacokinetics of [¹⁴C]‐AZD1981 in healthy male volunteers (Study D9830C00006)

Four healthy male subjects received a single oral dose of [¹⁴C]‐AZD1981 (250 mg, 33 kBq, 150 ml) administered as an oral solution (150 ml) through a drinking straw. The container was then rinsed with water (90 ml) and the subject was asked to drink the rinsing through the same drinking straw. Blood, urine and faeces samples (for analysis of total radioactivity, metabolite profiling and identification) were collected predose at specified times up to 11 days after dosing.

Uptake of AZD1981 and AZD1981 metabolites into isolated human hepatocytes

The method and the data modelling were performed as described previously 22. Microtubes containing caesium chloride (15 μl, 4 g 100 ml^–1) and mineral oil/silicon oil mixture (140 μl, with a specific gravity of 1.015 g ml^–1) were prepared and centrifuged (4000 g; 2 min). Cryopreserved hepatocytes were thawed and resuspended in incubation medium, such that cell viability was greater than 80%. Incubations were performed at a cell suspension concentration of 1 million cells ml^–1 and a temperature of 37°C. Samples (100 μl) were removed from incubations at various time points, and after pipetting into the microtubes, the cells were separated from the incubation medium by high‐speed centrifugation (21 000 × g, Eppendorf MiniSpin centrifuge, Eppendorf Nordic, Horsholm, Denmark). Rapid freezing on solid dry ice allowed the tubes to be cut precisely to yield the cell pellets below the oil layer. AZD1981 and the N‐deacetylated AZD1981 metabolite were extracted from the cell pellet using acetonitrile and after dilution with water, the extracts were analysed by liquid chromatography–tandem mass spectrometry (LC–MS‐MS) for quantification of both. Samples obtained from control incubations run in parallel at 4°C were processed in the same manner. Elucidation of the processes of hepatocyte drug uptake and metabolism was achieved using a mathematical model comprising compartments representing incubation medium, outer cell membrane and hepatocytes, as detailed previously 22. Using the nonlinear least‐squares solver of the commercial software package Matlab 2014b (MathWorks Inc., MA, Natick, Massachusetts, USA), estimates of the intrinsic active uptake (CL_int,up), bidirectional passive diffusion (CL_int,diff) and metabolism (CL_int,met) were obtained by simultaneous fit to 37°C and 4°C data. At 4°C, the CL_int,up and CL_int,met processes were considered inactivated 23. Active uptake may affect the DDI potential by raising the unbound liver concentration above that of the blood. In vitro, the unbound intracellular‐to‐extracellular concentration ratio at steady state (K_p,uu) is defined by the ratio of the sum of the input and output terms for the cell, here calculated according to Equation (1). While the described methodology cannot separately distinguish translocation mediated by efflux transporters, K_p,uu, reflecting the net transport in and out of the cell, is independent of the exact parameterization of drug fluxes 22.

$${\mathrm{K}}_{\mathrm{p},\mathrm{uu}}=\frac{{CL}_{int,up}+{CL}_{int,diff}}{{CL}_{int,met}+{CL}_{int,diff}}$$ (1)

Inhibition of recombinant CYP2C9‐dependent S‐warfarin 7‐hydroxylation and determination of reversible inhibition IC₅₀ values for AZD1981 and metabolites of AZD1981 (acyl glucuronide, sulphoxide and N‐deacetylated AZD1981)

Eight concentrations (0.03–300 μM) of each compound were incubated in triplicate in potassium phosphate buffer (pH 7.4) at 37°C with recombinant CYP2C9 (70 pmoles CYP ml^–1). The experiment was initiated by the addition of NADPH (1 mM incubation concentration) and continued for 30 min before termination with acetonitrile (160 μl to each 80 μl incubation). After centrifugation (20 min, 3220 g, 4°C), the supernatants were analysed by LC–MS‐MS, from which the formation of 7‐OH warfarin was quantified using a standard curve. The half‐maximal inhibitory concentration (IC₅₀) values were estimated using Phoenix Pharsight (Pharsight Corporation, Mountain View, CA, USA) according to Equation (2):

$${\mathrm{E}=\mathrm{E}}_0\times \left(1‐\frac{\text{Imax}\times {\mathrm{C}}^{\gamma }}{{\mathrm{C}}^{\gamma }{{+\mathrm{IC}}_{50}}^{\gamma }}\right)$$ (2)

where E and E₀ are the rates of S‐warfarin metabolism in the presence and absence of inhibitor respectively, Imax is the maximum extent of inhibition, γ is the Hill slope factor and C is concentration of test inhibitor.

Time‐dependent inhibition of human liver microsomal CYP2C9‐dependent diclofenac 4‐hydroxylation by AZD1981 and N‐deacetylated AZD1981

The assay was performed essentially as described previously 24. Six concentrations of AZD1981 and N‐deacetylated AZD1981 (0, 0.6 μM, 1.9 μM, 5.6 μM, 16.7 μM and 50 μM) were preincubated in duplicate in potassium phosphate buffer (pH 7.4) for 0, 1 min, 5 min, 10 min, 20 min and 30 min with pooled human liver microsomes (1 mg protein ml^–1) at 37°C. Aliquots (10 μL) of the pre‐incubation were transferred to the secondary incubations (final volume of 100 μL) containing 30 μM diclofenac and incubations allowed to proceed for 15 min at 37°C. The secondary incubations were then terminated, centrifuged and supernatants analysed for 4‐hydroxy diclofenac by LC–MS‐MS. The data was treated as described previously 24 to generate inactivation rate constants for each concentration of AZD1981 or N‐deacetylated AZD1981 used. Values for K_I and k_inact were estimated using Phoenix Pharsight (Pharsight Corporation,Mountain View, CA), according to Equation (3). The value for K_I was adjusted to account for drug binding to the liver microsomes (performed by equilibrium dialysis, as described previously 25). The value used in the DDI calculation was the K_I observed × fu_inc, where fu_inc represents the unbound fraction of AZD1981 or N‐deacetylated AZD1981 in the incubation with human liver microsomes at a concentration of 1 mg protein ml^–1. It should be noted that recombinant CYP was used to study the reversible CYP2C9‐dependent S‐warfarin inhibition (detailed above), in order to minimize nonspecific drug binding, but human liver microsomes were used in the investigation of time‐dependent CYP inhibition. This is because human liver microsomes are a multidrug‐metabolizing enzyme system and it is, at least theoretically, possible that a drug metabolite can be formed by one drug‐metabolizing enzyme but further metabolized to reactive species by a separate enzyme 24. Furthermore, because of the low enzyme concentration in the second step of the time‐dependent inhibition assay, nonspecific drug binding is minimal 24.

$${\mathrm{k}}_{\mathrm{obs}}=\frac{{\mathrm{k}}_{\text{inact}}\times \mathrm{I}}{{\mathrm{K}}_{\mathrm{I}}+\mathrm{I}}$$ (3)

In Equation (3), I is the preincubation inhibitor concentration, k_obs is the inactivation rate constant for a given I, and k_inact and K_I are the maximal rate of CYP inactivation and the concentration of inactivator taken to elicit half‐maximal inactivation, respectively.

A phase I, open‐label, randomized, fixed‐sequence crossover study to evaluate the pharmacokinetic interaction between AZD1981 and warfarin (Study D9830C00017)

Subjects were randomized to one of two treatment groups (either 100 mg or 400 mg of AZD1981). Warfarin was administered as a single dose (25 mg) in a fixed sequence on two occasions, 14 days apart (n = 24 subjects in total). On the first dosing occasion, day 1, a single dose of warfarin was given. On the second dosing occasion, day 15, warfarin was administered concomitantly with AZD1981 after 7 days of pretreatment with AZD1981. AZD1981 was administered as oral tablets twice daily, 100 mg (n = 13) or 400 mg (n = 11, 4 × 100 mg) for 14 days from day 8 to day 21. The initial 7‐day treatment with AZD1981, on days 8 to 14, before the second dose of warfarin on day 15, was to ensure that steady state had been reached with respect to exposure of AZD1981 and potential inhibition of CYP2C9. The remaining 7 days of AZD1981 treatment were to ensure that those conditions were maintained until the major part of the warfarin dose had been eliminated. The plasma concentration of AZD1981 and (R)‐ and (S)‐enantiomers of warfarin were quantified in all 24 subjects. The concentration of metabolites of AZD1981 were quantified in only four subjects, two receiving 100 mg AZD1981 and two receiving 400 mg AZD1981 [labelled subjects 1 and 2 (100 mg), and 3 and 4 (400 mg), respectively in Table 1]. The change in AUC for both (R)‐ and (S)‐warfarin in these subjects, and also the concentration of AZD1981 were found to be similar to the group mean. The average concentration of the N‐deacetylated metabolite was calculated using the AUC obtained following the warfarin dose, dividing by the dosing interval, assuming AZD1981 to be at steady state.

Table 1.

Steady‐state plasma exposures of AZD1981 and metabolites in four individuals following oral dosing of 100 mg and 400 mg AZD1981

	Subject	AZD1981 dose (mg)	Cmax (nmol l –1 )	AUC (nmol l –1 .h)	AUC a (%)
AZD1981	1	100	5700	19 400	86
	2	100	3650	11 000	86
N‐deacetylated	1	100	385	2990	13
	2	100	297	1820	14
Acyl glucuronide	1	100	48	115	<1
	2	100	25	36	<1
Sulphoxide	1	100	33	36	<1
	2	100	21	5	<1
AZD1981	3	400	9030	18 900	75
	4	400	13 800	51 700	83
N‐deacetylated	3	400	783	4985	20
	4	400	1160	7328	12
Acyl glucuronide	3	400	255	780	3
	4	400	352	1605	3
Sulphoxide	3	400	105	399	2
	4	400	303	1370	2

AUC, area under the plasma concentration–time curve; Cmax, peak plasma concentration.

^aPercentage of the sum of AZD1981 plus metabolites (N‐deacetylated, acyl glucuronide and sulphoxide) AUC. Other metabolites detected in the plasma were minor

The lower limit of quantification of AZD1981 and its N‐deacetylated metabolite was 20 nmol l^–1 and for S‐warfarin was 65 nmol ml^–1.

Study conduct and ethics

Study D9830C00017 was conducted at a single clinical centre in Quintiles AB, Phase I Unit in Uppsala, Sweden. Study D9830C00006 was conducted at a single centre in Quintiles Drug Research Unit at Guy's Hospital, Quintiles, London, UK. Both were conducted in accordance with good clinical practice. Protocols, amendments and informed consent forms were reviewed and approved by the appropriate institutional review boards or independent ethics committees for participating sites prior to study initiation.

Estimation of drug–drug interaction

The DDI calculation was made according to Equation (4) 26:

$$\text{AUCi}/\mathrm{AUC}=\frac{1}{\left[\mathrm{A}\times \mathrm{B}\right]\times \text{fmCYP}+\left(1–\text{fmCYP}\right)}$$ (4)

where A and B are defined by Equations (5) and (6):

$$\mathrm{A}=\frac{{\mathrm{k}}_{\mathrm{deg}}}{{\mathrm{k}}_{\text{inact}}\times {\left[\mathrm{I}\right]}_{\mathrm{H},\mathrm{u}}/\left({\mathrm{K}}_{\mathrm{I}}+{\left[\mathrm{I}\right]}_{\mathrm{H},\mathrm{u}}\right){+\mathrm{k}}_{\mathrm{deg}}}$$ (5)

$$\mathrm{B}=\frac{1}{1+\left({\left[\mathrm{I}\right]}_{\mathrm{H},\mathrm{u}}/{\mathrm{K}}_{\mathrm{i}}\right)}$$ (6)

and AUCi/AUC is the ratio of exposures of the victim drug (S‐warfarin) in the presence and absence of inhibiting coadministered drug; fmCYP is the fraction of the victim drug's total clearance due to the inhibited enzyme, CYP2C9 (set to a value of 0.9 for S‐warfarin 27, 28: [I]_H,u is the unbound hepatic concentration of inhibiting drug; k_deg is the rate constant describing the rate of CYP2C9 degradation in vivo (defined as 0.0001 min⁻¹ 29); k_inact and K_I are as described in Equation (3); and K_i is the reversible inhibition constant. K_i was calculated from IC₅₀/2, assuming competitive inhibition. As the substrate (S‐warfarin) was incubated at a concentration close to the substrate concentration at half the maximum velocity (Km), this estimate of K_i will closely approximate to the true K_i 30.

The DDI calculations for AZD1981 and the N‐deacetylated metabolite were made using the individual average concentrations at steady state for the four subjects for whom concentrations were available for both entities: 1.8 μM and 1.0 μM, respectively, for AZD1981 and 0.16 μM and 0.25 μM, respectively, for the N‐deacetylated metabolite, following the 100 mg dose of AZD1981; 4.4 μM and 1.7 μM, respectively, for AZD1981 and 0.60 μM and 0.39 μM, respectively, for the N‐deacetylated metabolite, following the 400 mg dose (Table 2).

Plasma protein binding was assayed in vitro, for both AZD1981 and the N‐deacetylated metabolite, by equilibrium dialysis at 37°C for 18 h (according to the previously published method 31). Plasma protein binding was found to be 97.14% for AZD1981, with whole blood/plasma partitioning (B/P) of 0.7 (three male donors pooled) and an average unbound concentration in systemic blood was therefore calculated to be 0.18 μM and 0.07 μM, respectively, for the two subjects receiving the 400 mg dose. Free hepatic concentrations of 1.4 μM and 0.6 μM, respectively, were then calculated from Equation (7), where K_p,uu is the unbound cell‐to‐media concentration ratio estimated from the in vitro hepatocyte uptake data 22. Similarly, unbound average hepatic concentrations of 0.33 μM and 0.36 μM, respectively, were calculated for the two individuals receiving 100 mg AZD1981.

For the N‐deacetylated metabolite, plasma protein binding assessed in vitro was 99.04% (five pooled donors) and, with a B/P value of 0.7, an average unbound concentration in systemic blood was calculated to be 5 nM and 8 nM, respectively, for the two subjects receiving the 400 mg dose; the free hepatic concentrations were therefore calculated to be 64 nM and 100 nM, respectively, using a K_p,uu value of 12 (Equation (7)). Unbound average hepatic concentrations of 26 nM and 41 nM, respectively, were calculated for the two individuals receiving 100 mg AZD1981.

$$\text{Free}\text{hepatic}{\text{concentration}=\mathrm{K}}_{\mathrm{p},\mathrm{uu}}\times \text{free}\text{blood}\text{concentration}$$ (7)

Results

An open, single‐dose, single‐centre, phase I study to assess the metabolism, excretion and pharmacokinetics of [¹⁴C]‐AZD1981 in healthy male volunteers

Four healthy male subjects were enrolled in and completed the study; three subjects were Caucasian and one was Black, and their mean body mass index was 26 kg m^{– 2}.

The geometric mean total recovery of the [¹⁴C]‐AZD1981 dose was 95.1% (range 94.4–95.6%), with corresponding geometric mean recovery values in the urine and faeces of 52.7% and 41.9%, respectively. The major amount of radioactive material in the urine was recovered within 6 h. The excretion was slower in the faeces, with the major amount excreted by this route within 72 h (Figure 2). AZD1981 was observed as the major urinary component, accounting for 14.5% of the total dose. The major metabolites in the urine were identified as acylglucuronide (11.1%) and sulphoxide (9.6%). A number of minor metabolites, each accounting for less than 2% of the dose, were also identified in the urine. The largest radioactive peak in the faeces, corresponding to 20.4% of the administered dose, was identified as AZD1981. The other components present in the faeces each accounted for less than 5% of the excreted dose. The major part (60%) of the drug‐related material (radioactivity) plasma exposure was accounted for by unchanged AZD1981. The most abundant metabolite present in the plasma was the N‐deacetylated metabolite, accounting for 9.2% of the total radioactive drug‐related material in the 24‐h sample. B/P concentration ratios were calculated to be 0.7 for AZD1981 and the N‐deacetylated amino acid metabolite.

Figure 2.

[¹⁴C]‐AZD1981 human excretion balance study. Geometric mean cumulative recovery (percentage of dose) of radioactivity in the urine (red diamonds, n = 3), faeces (blue triangles, n = 4) and total (green circles n = 3, urine + faeces) vs. time. For one subject, urine was not collected quantitatively

Uptake of AZD1981 and the AZD1981 N‐deacetylated amino acid into isolated human hepatocytes

To assess the mechanisms governing hepatocyte uptake of AZD1981 and its N‐deacetylated metabolite quantitatively, in vitro experiments were performed at 37°C and 4°C using human hepatocytes in suspension incubations. For both compounds, the amount of drug recovered in the cellular fraction was significantly reduced at low temperature (Figure 3). Application of a mechanistic mathematical model enabled the estimation of kinetic parameters from obtained cellular concentration profiles. According to Equation (1), best‐fit intrinsic clearances associated with the CL_int,up, CL_int,diff and CL_int,met of AZD1981 were 77 μl min^–1 per million cells [95% confidence interval (CI) 67, 87), 11 μl min^–1 per million cells (95% CI 7, 15) and 0.24 μl min^–1 per million cells (95% CI 0.08, 0.39), respectively. Best‐fit intrinsic clearances associated with CL_int,up, CL_int,diff and CL_int,met for the N‐deacetylated metabolite were 36 μl min^–1 per million cells (95% CI 30, 41), 2.8 μl min^–1 per million cells (95% CI 1.4, 4.1) and 0.77 μl min^–1 per million cells (95% CI 0.42, 1.11), respectively.

Figure 3.

In vitro hepatic uptake: amount of AZD1981 (A) and N‐deacetylated AZD1981 (B) recovered in the cellular fraction from 1 μM incubations performed in duplicate at 37°C (open green circles) and 4°C (open blue diamonds) with suspended human hepatocytes at 1 million cells per ml. Solid lines depict the best‐fit simulations obtained using the mechanistic mathematical model (see section on methods)

Inhibition of recombinant CYP2C9‐dependent S‐warfarin 7‐hydroxylation by AZD1981 and metabolites of AZD1981: determination of reversible inhibition IC₅₀ values

The kinetic parameters for S‐warfarin metabolism to 7‐hydroxy S‐warfarin were determined prior to the study. Km and maximal velocity (Vmax) values were 14 μM and 0.4 nmol min^–1 nmol^–1 CYP2C9, respectively (data not shown) and consequently the concentration of S‐warfarin used in the inhibition experiments was 10 μM, so that, assuming competitive inhibition, the IC₅₀ values estimated could be assumed to approximate to twice the true Ki values 30. AZD1981 and the AZD1981 N‐deacetylated amino acid inhibited S‐warfarin metabolism with IC₅₀ values of 89 μM (95% CI 78, 103) and 2.4 μM (95% CI 2, 3), respectively (Figure 4). AZD1981 acyl‐glucuronide and AZD1981 sulphoxide both displayed weak CYP2C9 inhibition (53 and 160 μM, respectively; data not shown).

Figure 4.

Reversible inhibition of CYP2C9‐mediated metabolism of S‐warfarin to 7‐OH warfarin by AZD1981 (green circles) and N‐deacetylated AZD1981 (blue diamonds). Data collected in triplicate. Solid lines depict best‐fit profile according to Equation (2)

Time‐dependent inhibition of human liver microsomal CYP2C9‐dependent diclofenac 4‐hydroxylation by AZD1981 and the AZD1981 N‐deacetylated amino acid

Diclofenac is known to be metabolized to 4‐hydroxy diclofenac by CYP2C9 and was assayed as described previously 24. Preincubations of the AZD1981 N‐deacetylated amino acid demonstrated time‐dependent CYP2C9 inhibition, well described by linear regression of log_e(% control activity remaining)–time data to define individual CYP‐inactivation rate constants (k_obs values) (Figure 5A) that allowed k_inact and K_I values to be determined according to the hyperbolic function described in Equation (3) (Figure 5B). K_I and k_inact values of 9.9 μM (95% CI 4.2, 14.9) and 0.023 min⁻¹ (95% CI 0.019, 0.028) were estimated. The true K_I value was estimated to be 7.2 μM through adjustment for incubation binding (fraction unbound in a 1 mg human liver microsomal protein per ml incubation equal to 0.73, determined as detailed in the methods section). No time‐dependent CYP2C9 inhibition was observed when AZD1981 was preincubated with human liver microsomes.

Figure 5.

Time‐dependent inhibition of cytochrome P450 (CYP) 2C9‐dependent diclofenac 4′‐hydroxylation activity by N‐deacetylated AZD1981 (A) CYP2C9 activity remaining after preincubation with N‐deacetylated AZD1981 at 0 μM (open green circles), 0.167 μM (open blue diamonds), 1.85 μM (open red triangles), 5.56 μM (closed green circles), 16.7 μM (closed blue diamonds) and 50 μM (closed orange triangles), each in duplicate. The slope of each linear regression line represents the inactivation rate constant (k_obs). (B) k_obs vs. N‐deacetylated AZD1981 concentration at preincubation concentration

Results of a phase I pharmacokinetic interaction study with repeated oral doses of AZD1981 (100 mg twice daily and 400 mg twice daily) for 2 weeks and single doses of warfarin (25 mg)

There were 28 subjects who received randomized study medication, of whom 24 completed the study. All subjects were male Caucasians, with a mean age of 26 years and a body mass index of 24.5 kg m^{– 2}. Twice‐daily oral administration of 100 mg and 400 mg AZD1981 dose‐dependently and enantioselectively interacted with the pharmacokinetics of a 25 mg single dose of racemic warfarin (Figure 6). The peak plasma concentration (Cmax) and AUC of R‐warfarin were similar when warfarin was administered alone or together with AZD1981. The Cmax for S‐warfarin was also only minimally affected. However, the AUC of S‐warfarin increased with the concomitant administration of AZD1981, on average 1.4‐ (95% CI 1.22, 1.50) and 2.4‐ (95% CI 2.11, 2.64) fold after 100 mg (n = 13) and 400 mg (n = 11) twice daily, respectively, compared with control (n = 24).

Figure 6.

Mean plasma concentration–time profiles for AZD1981 (green circles), R‐warfarin (orange downward triangles) and S‐warfarin (blue upward triangles). Filled triangles indicate R/S‐warfarin concentration when dosed alone, and open triangles indicate R/S‐warfarin concentration when dosed with (A) 100 mg AZD1981 (n = 13) or (B) 400 mg AZD1981 (n = 11). The full pharmacokinetics of AZD1981 were only collected over the first 12 h after warfarin administration. Standard deviation is indicated by error bars

Table 1 gives the steady‐state plasma exposures of AZD1981 and metabolites in the four subjects for which AZD1981 metabolites were measured (n = 2 for both the 100 mg and 400 mg doses). The major circulating metabolite of AZD1981 was the N‐deacylated amino acid, representing 12–20% of the total quantified drug‐related material in the plasma at steady state (Table 1). Figure 7 shows the full plasma concentration–time profiles for AZD1981 and N‐deacetylated metabolite in the four investigated subjects. The observed S‐warfarin AUC shifts in these individuals were representative for each dose group (Table 2: AUC changes of 1.43, 1.39 and 2.03, 2.49 at 100 mg and 400 mg, respectively).

Figure 7.

Plasma concentration–time profiles for AZD1981 (solid markers) and AZD1981 N‐deacetylated amino acid (open markers) following (A) AZD1981 100 mg (subject 1: blue circles and subject 2: green diamonds) and (B) AZD1981 400 mg (subject 3: green diamonds and subject 4: blue circles)

Table 2.

Observed and predicted AZD1981–warfarin interaction

	Subject 1	Subject 2	Subject 3	Subject 4
	100 mg	100 mg	400 mg	400 mg
Observed AUC increase S‐warfarin	1.43	1.39	2.03	2.49
Observed AUC increase R‐warfarin	0.98	1.09	0.89	1.03
AZD1981
Average C (nM, total)	1771	995	4391	1671
Predicted AUC increase due to competitive inhibition	1.0	1.0	1.0	1.0
Predicted AUC increase due to TDI	1.0	1.0	1.0	1.0
Predicted total AUC increase	1.0	1.0	1.0	1.0
N‐deacetylated AZD1981
Average C (nM, total)	251	164	602	386
Predicted AUC increase due to competitive inhibition	1.0	1.0	1.1	1.1
Predicted AUC increase due to TDI	2.2	1.7	3.0	2.4
Predicted total AUC increase	2.2	1.7	3.2	2.5

AUC, area under the plasma concentration–time curve; C, concentration; TDI, time dependent inhibition

Estimation of clinical drug–drug interaction potential

Although warfarin was administered to 24 subjects, the concentrations of metabolites of AZD1981 were quantified in only four subjects, two receiving 100 mg AZD1981 and two receiving 400 mg AZD1981. The plasma concentration data for these subjects is detailed in Table 2. Using Equations (4), (5) and (6), the change in S‐warfarin plasma exposure due to dosing of 100 mg and 400 mg AZD1981 was predicted to be 1 (no interaction). The predicted change in S‐warfarin plasma exposure due to the estimated hepatic exposure of the AZD1981 N‐deacetylated amino acid after 100 mg dosing of AZD1981 was calculated to be 1.7‐fold and 2.2‐fold for subjects 1 and 2 (with the interaction calculated to come entirely from the time‐dependent CYP2C9 inhibition). For the 400 mg AZD1981 dose, interactions for subjects 3 and 4 were predicted to be 2.5‐fold and 3.2–fold, respectively (Table 2).

Discussion

Understanding the important factors for consideration in making successful DDI predictions from in vitro data has advanced significantly over the course of the last two decades. Early influential guiding publications 32 have matured to accurate quantitative predictions becoming more commonplace, with appropriate attention to detail being imperative 7, 33, 34, 35, 36. Nevertheless, most investigations have been restricted to cases where the parent drug is the perpetrator.

Previous in vitro studies showed AZD1981 to be metabolically stable (statistically significant intrinsic clearance values could not be defined from standard assays) 4. However, three metabolites (sulphoxide, acyl‐glucuronide and the AZD1981 N‐deacetylated amino acid) were identified following incubation with isolated human hepatocytes. Early in the clinical development of AZD1981, a radiolabelled excretion balance study was performed and revealed the N‐deacetylated metabolite to be the most abundant metabolite present in the plasma of healthy male volunteers. The N‐deacetylated metabolite was identified as more lipophilic than the parent drug (with a logD_7.4 of 0.22 compared with −0.12), raising the possibility of more potent CYP inhibition. Although not a dramatic increase, consideration of the secondary amine revealed on the molecule through the de‐acetylation, also gave rise to concern of increased CYP2C9 inhibition risk, as this feature is also present on sulphaphenazole (Figure 1), for which the N‐phenyl group has been shown to have an important interaction with a hydrophobic part of the CYP2C9 protein active site 37. This, coupled with the carboxylic acid function (common to parent drug and metabolites) that is associated with CYP2C9 inhibition and organic anion transporting polypeptide (OATP) affinity leading to increased unbound hepatic drug concentrations relative to the blood 38, 39, directed attention towards further in vitro analyses.

Experiments showed the sulphoxide and acyl glucuronide metabolites to be extremely weak CYP2C9 inhibitors but revealed the AZD1981 N‐deacetylated amino acid to be approximately 40‐fold more potent than AZD1981 as a reversible inhibitor of CYP2C9‐dependent S‐warfarin 7‐hydroxylation. Weak time‐dependent inhibition of CYP2C9 was also detected (the k_inact/K_I ratio of 2.3 μl min⁻¹ nmol⁻¹, being approximately 50‐fold lower in inactivation efficiency than tienilic acid 40) and in vitro human hepatocyte data established both the parent and N‐deacetylated metabolite to be subject to considerable active uptake, with intracellular/incubation medium‐free drug concentration ratios of 8 and 12 for the parent and metabolite, respectively. As these data raised concerns of a possible DDI, a clinical warfarin interaction study was performed, with warfarin administered as a racemic mixture containing equal amounts of the R‐ and S‐enantiomers. S‐warfarin is known to be primarily metabolized by CYP2C9, and R‐warfarin is known to be a substrate for CYP1A2 and CYP3A 41.

In the present clinical DDI study, the major circulating metabolite was confirmed as the AZD1981 N‐deacylated amino acid, with the shape of the metabolite plasma concentration–time profile being different to that of the parent drug (Figure 7), characterized by a later time to reach peak concentration (Tmax) and a longer elimination half‐life. The results of the DDI study indicated that the AUC of S‐warfarin, the more potent enantiomer, increased with concomitant administration of AZD1981, on average 1.4‐fold and 2.3‐fold after 100 mg and 400 mg twice daily, respectively (Figure 6). This interaction is likely to be of clinical importance, in that it would require increased monitoring of the prothrombin time and international normalized ratio, with probable adjustment of the warfarin dose regimen during coadministration of the two drugs.

Using the in vitro hepatocyte uptake data (K_p,uu values of 8 and 12) and the average plasma concentration data across subjects for AZD1981 and the AZD1981 N‐deacetylated amino acid from the warfarin interaction study, calculated free liver concentrations of 1.0 μM and 82 nM were derived for the parent and metabolite, respectively, at the 400 mg AZD1981 dose. Notwithstanding the high hepatic concentrations estimated, no significant DDI could be rationalized for the parent drug as perpetrator because of the weak reversible CYP2C9 inhibition and absence of time‐dependent CYP inhibition. Despite relatively potent reversible S‐warfarin inhibition by the N‐deacetylated amino acid metabolite determined in vitro (IC₅₀ value of 2.4 μM), no significant DDI was predict4ed via this mechanism either (S‐warfarin AUC increase of 1.04 predicted). However, time‐dependent CYP2C9 inhibition was revealed as the likely source of the S‐warfarin interaction (predicted S‐warfarin AUC increase of 1.7–2.2‐fold at the 100 mg dose and 2.5–3.2‐fold at the 400 mg dose), with the unusual situation of the metabolite being subject to efficient hepatic uptake and slow elimination, coupled with the time‐dependent CYP2C9 inhibition as the underlying cause. It seems extremely likely, therefore, that this is a rare example of a metabolite being the principal contributor to a DDI through drug‐metabolizing enzyme inhibition.

In relation to the potential to cause clinically relevant DDIs, the FDA and EMA guidance state that metabolites which are present to at least 10% of total drug‐related exposure, or 25% or more of the parent AUC, should trigger further investigation of their (drug‐metabolizing enzyme or transporter) inhibitory potency. In the warfarin interaction study, the AZD1981 N‐deacetylated amino acid was present as 16–17% of the parent drug plasma exposure and 12–20% of the total drug‐related exposure. Therefore, this metabolite represented a case for concern. However, prior to conducting the present clinical study, the only clinical information available was from the [¹⁴C]‐AZD1981 excretion balance study, in which the metabolite in question accounted for less than 10% of the total radioactive drug‐related material in the plasma and hence could be viewed as a borderline case. This indicates that a conservative approach should be adopted when considering the regulatory guidance, and that detailed analyses around potential DDI predictions for metabolites are warranted. We were alerted by several minor warning flags: the raised lipophilicity compared with the parent drug, a structural feature indicating enhanced CYP2C9 active site affinity, and the likely propensity for OATP‐driven hepatic uptake of carboxylic acid‐containing drugs. The desire to study the time‐dependent inhibition in detail came from screening data showing a weak but significant time‐dependent CYP2C9 inactivation at a single concentration of the N‐deacetylated amino acid metabolite (decisions made according to the paradigm set out previously 40) and the significant increase in reversible CYP2C9 inhibition observed. Ultimately, this time‐dependent CYP2C9 inhibition was found to be the likely cause of the observed clinical DDI with S‐warfarin, serving to underline the subtleties and complexities of predicting DDIs and, again, underlining the need for vigilance and detailed analyses when considering which clinical DDI studies are required.

Prediction of which metabolites may be defined as major in a clinical setting cannot be made with any degree of certainty from simply assessing the relative concentrations from in vitro metabolite identification studies, unless metabolite pharmacokinetics are accounted for 42, 43. However, the present report highlights that, where possible, even minor risk‐associated metabolites should be thoroughly investigated preclinically with in vitro experiments. When such metabolites cannot be identified preclinically owing to low in vitro metabolic turnover of the parent drug, the risks should be explored very early in clinical development, making use of the clinical pharmacokinetic data coupled with in vitro DDI information to investigate interaction risk.

Competing Interests

All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author). K.G., R.P., P.N., M.B., P.S., B.J., M.G. and C.B. worked for AstraZeneca at the time of the research and had no other relationships or activities influencing the submitted work. T.M. and W.K. worked for Quintiles at the time that the research was conducted and were principal investigators for the clinical studies performed. T.M. is supported by the National Institute for Health Research Biomedical Research Centre at Guy's and St Thomas' NHS Foundation Trust and King's College London.

For the clinical sample work, we thank Dan Weston (metabolite identification) and Jon Attwood (bioanalytical support).

Contributions

K.G., R.P., P.N. and B.J. wrote the article. K.G., R.P., P.N., M.B., P.S., C.B., M.G., T.M. and W.K. designed the research. P.N., M.B., P.S., T.M. and W.K. performed the research. K.G., R.P., P.N., M.B., P.S., C.B. and M.G. analysed the data.

Grime, K. , Pehrson, R. , Nordell, P. , Gillen, M. , Kühn, W. , Mant, T. , Brännström, M. , Svanberg, P. , Jones, B. , and Brealey, C. (2017) An S‐warfarin and AZD1981 interaction: in vitro and clinical pilot data suggest the N‐deacetylated amino acid metabolite as the primary perpetrator. Br J Clin Pharmacol, 83: 381–392. doi: 10.1111/bcp.13102.