1521-009X/44/8/1201–1212$25.00 http://dx.doi.org/10.1124/dmd.115.069203DRUG METABOLISM AND DISPOSITION Drug Metab Dispos 44:1201–1212, August 2016Copyright ª 2016 by The American Society for Pharmacology and Experimental Therapeutics

Metabolic Disposition of Osimertinib in Rats, Dogs, and Humans:Insights into a Drug Designed to Bind Covalently to a Cysteine

Residue of Epidermal Growth Factor Receptor s

Paul A. Dickinson, Mireille V. Cantarini, Jo Collier, Paul Frewer, Scott Martin, Kathryn Pickup,and Peter Ballard

AstraZeneca, Macclesfield (M.V.C., P.F., S.M., K.P., P.B.), Seda Pharmaceutical Development Services, Cheshire (P.A.D), andQuotient Clinical Ltd., Ruddington, United Kingdom (J.C.)

Received December 23, 2015; accepted May 24, 2016

ABSTRACT

Preclinical and clinical studies were conducted to determine themetabolismand pharmacokinetics of osimertinib and keymetabolitesAZ5104 and AZ7550. Osimertinib was designed to covalently bind toepidermal growth factor receptors, allowing it to achieve nanomolarcellular potency (Finlay et al., 2014). Covalent binding was observed inincubations of radiolabeled osimertinib with human and rat hepato-cytes, human and rat plasma, and human serumalbumin. Osimertinib,AZ5104, and AZ7550 were predominantly metabolized by CYP3A.Seven metabolites were detected in human hepatocytes, also ob-served in rat or dog hepatocytes at similar or higher levels. After oraladministration of radiolabeled osimertinib to rats, drug-related mate-rial was widely distributed, with the highest radioactivity concentra-tionsmeasured at 6 hours postdose in most tissues; radioactivity was

detectable in 42% of tissues 60 days postdose. Concentrations of[14C]-radioactivity in blood were lower than in most tissues. After theadministration of a single oral dose of 20 mg of radiolabeledosimertinib to healthy male volunteers, ∼19% of the dose wasrecovered by 3 days postdose. At 84 days postdose, mean totalradioactivity recovery was 14.2% and 67.8% of the dose in urine andfeces. The most abundant metabolite identified in feces was AZ5104(∼6%of dose). Osimertinib accounted for∼1%of total radioactivity inthe plasma of non–small cell lung cancer patients after 22 daysof 80-mg osimertinib once-daily treatment; the most abundant circula-tory metabolites were AZ7550 and AZ5104 (<10% of total osimertinib-relatedmaterial). Osimertinib is extensively distributed andmetabolizedin humans and is eliminated primarily via the fecal route.

Introduction

Epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors(EGFR-TKIs) (i.e., gefitinib, erlotinib, and afatinib) are recommended asfirst-line therapy for patients with advanced non–small cell lung cancer(NSCLC) that harbors an EGFR sensitizing mutation (Keedy et al., 2011;Reck et al., 2014). Tumors of most patients treated first-line with acurrently approved EGFR-TKI develop resistance, however, leading todisease progression (Kobayashi et al., 2005; Pao et al., 2005; Sequist et al.,2011). The EGFR T790M mutation is found in approximately 60% ofcases that have developed resistance to EGFR-TKIs (Arcila et al., 2011;Oxnard et al., 2011; Yu et al., 2013; Kuiper et al., 2014; Li et al., 2014).Osimertinib (TAGRISSO; AstraZeneca Pharmaceuticals,Macclesfield,

UK) is a novel oral, potent, irreversible EGFR-TKI that is structurally andpharmacologically distinct from all other licensed EGFR-TKIs (AstraZe-neca Pharmaceuticals LP, osimertinib tablet: highlights of prescribinginformation available at http://www.accessdata.fda.gov/drugsatfda_docs/label/2015/208065s000lbl.pdf). Osimertinib is selective for EGFR-TKI-sensitizing mutations and the T790M resistance mutation, with lower

activity against wild-type EGFR compared with currently approvedEGFR-TKIs (Cross et al., 2014). Osimertinib was recently approved bythe US Food and Drug Administration for the treatment of patients withmetastatic EGFRT790Mmutation-positive NSCLCwho have progressedon or after EGFR-TKI therapy. Osimertinib clinical activity has beendemonstrated in the treatment of patients with EGFRmutant NSCLC afterprogression on a previous EGFR-TKI owing to the EGFR T790Mmutation in the ongoing AURA (NCT01802632) and AURA2(NCT02094261) studies (Jänne et al., 2015a,b; Mitsudomi et al., 2015;Yang et al., 2015). Preclinical data suggest that osimertinib is principallymetabolized by cytochrome P450 (CYP3A4) and produces at least twocirculating active metabolites: AZ5104 and AZ7550 (Cross et al., 2014;Planchard et al., 2014).We conducted a series of studies in laboratory animals and clinical

studies in healthy volunteers and patients with NSCLC to characterizethe absorption, distribution, metabolism, and excretion characteristics ofosimertinib. The results of these studies are reported herein.

Materials and Methods

Study Conduct

The clinical studies described here were conducted in accordance with theInternational Conference on Harmonization—Good Clinical Practice, the ethical

dx.doi.org/10.1124/dmd.115.069203.s This article has supplemental material available at dmd.aspetjournals.org.

ABBREVIATIONS: ADME, absorption, distribution, metabolism, and excretion; AMS, accelerator mass spectrometry; AUC, area under theconcentration-time curve; CLint, intrinsic clearance; EGFR, epidermal growth factor receptor; FMO, flavin-containing monooxygenase; GSH,glutathione; HPLC, high-performance liquid chromatography; LSC, liquid scintillation counting; MS, mass spectrometry; MS/MS, tandem massspectrometry; NSCLC, non–small cell lung cancer; P450, cytochrome P450; QWBA, quantitative whole-body autoradiography; TKI, tyrosine kinaseinhibitor; tmax, time to maximum plasma concentration.

1201

http://dmd.aspetjournals.org/content/suppl/2016/05/25/dmd.115.069203.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

principles that have their origin in the Declaration of Helsinki, applicableregulatory requirements, and the AstraZeneca policy on Bioethics.

Where applicable, the preclinical studies described here were performed inaccordance with the Organization for Economic Co-operation and DevelopmentPrinciples of Good Laboratory Practice.

Radiolabeled Osimertinib and Reference Compounds

Radiolabeled osimertinib [(N-[2-[2-(dimethylamino)ethyl-methyl-amino]-4-methoxy-5-[[4-(1-methylindol-3-yl)pyrimidin-2-yl]amino]phenyl]prop-2-enamide)]was synthesized by the Isotope Chemistry Section of AstraZeneca R&D(Macclesfield, UK). The structures of [14C]- and [3H]-osimertinib are illustratedin Supplementary Fig. 1. [3H]-osimertinib was used to determine the covalentbinding of osimertinib and its metabolites to human hepatic proteins (specificactivity 1802 MBq/mg; purity . 98%) and rat hepatic proteins (specific activity1817 MBq/mg; purity . 98%). [14C]-osimertinib was used to determine thecovalent binding of osimertinib and its metabolites to human and rat plasma andhuman serum albumin (specific activity 4.44 MBq/mg ; purity . 98%).[14C]-osimertinib was also used to determine the tissue distribution of osimertinibin the rat (specific activity 4.44 MBq/mg; purity . 98%) and to determine therates and routes of elimination of osimertinib in healthy male volunteers (specificactivity 0.0303 MBq/mg; purity 99%).

Nonradiolabeled osimertinib and reference compounds (AZ5104, AZ7550,and M1) were synthesized by AstraZeneca R&D (Macclesfield, UK). Thestructure and purity of these compounds were assessed by high-performanceliquid chromatography (HPLC) with UV light spectroscopy detection, massspectrometry (MS), and nuclear magnetic resonance spectroscopy. Syntheticroutes for these compounds have been previously described (Finlay et al., 2014).

Plasma Stability

Frozen plasma and protein solutions from the following species were used forstability assessment: mouse (CD-1, male, n . 3), rat (Han Wistar, male, n . 3),dog (Beagle, male, n . 3), human (Asian, mixed sex, n . 20), human serumalbumin (45 mg/ml), human a1-acid glycoprotein (0.70 mg/ml). Stock solutionsof osimertinib were dissolved in dimethyl sulfoxide and then diluted into plasmaor protein solutions to achieve concentrations from 0.1 to 100 mM containing 1%dimethyl sulfoxide. Samples were vortexed and incubated at 37�C for up to6 hours. Aliquots (50 ml) were mixed with water (50 ml), and the reaction wasterminated by the addition of acetonitrile (400 ml), vortexed, and centrifuged, andthe supernatant was analyzed by HPLC-MS/MS.

Covalent Binding in Human and Rat Hepatocytes

[3H]-osimertinib was incubated in triplicate at a nominal concentration of10 mM with cryopreserved rat and human hepatocytes (0.9 million cells/ml) at37�C for 4 hours under a 5% CO2 atmosphere. Positive control incubations with[3H]-labeled zomepiracwere conducted in parallel in both species. Rat and humanhepatocytes (combined male and female donors, n = 10) were obtained fromCelsis IVT (Chicago, IL).

For covalent binding assessment, aliquots were removed at the beginning(0 minutes) and the end (4 hours) of the incubation, quenched with acetone, andvortex mixed. Samples were kept at 8�C for 1 hour before protein harvesting ontofilter paper using a Brandel Cell Harvester (Alpha Biotech Ltd, Glasgow, UK) andwashed with methanol. Immobilized protein was solubilized in 5% SDS at 55�C for20 hours. Protein content was determined using the Thermo Scientific Pierce BCAProtein Assay Kit (Thermo Scientific, Somerset, NJ), and the concentration ofradioactivity was determined by liquid scintillation counting (LSC).

For estimation of the turnover of parent osimertinib, separate aliquots wereremoved at 0 and 4 hours, quenched with acetonitrile containing 0.8% formicacid, and kept at 220�C for more than 20 minutes before centrifugation at4000 rpm at 4�C. The supernatants were diluted 1:1 with ultrapure water andanalyzed by liquid chromatography with tandem mass spectrometry (MS/MS).

Covalent Binding in Rat and Human Plasma and Human Serum Albumin

[14C]-osimertinib was incubated in singlicate at nominal concentrations of 10mMin rat (male, Harlan, UK) and human (males, n = 3, Medicines Evaluation Unit,Manchester, UK) plasma and in human serum albumin (40 mg/ml) in phosphate-buffered saline at 37�C. Phosphate buffered saline was used as a negative control.

For covalent binding assessment, aliquots were removed at various time pointsup to 6 hours. Protein was precipitated with acetone and washed extensively withmethanol until the radioactivity present in the methanol supernatant was less thanapproximately three times the background radioactivity (as assessed by LSC).Protein was solubilized in 5% SDS followed by incubation overnight at 55�C.Protein content was determined using the Thermo Scientific Pierce BCA ProteinAssay Kit, and the concentration of radioactivity was determined by LSC.

Rat Quantitative Whole-Body Autoradiography

The tissue distribution of radioactivity in male partially pigmented (Lister-Hooded) and male and female albino (Wistar Han) rats was determined after a singleoral dose [4 mg/kg (8 mmol/kg)] of [14C]-osimertinib (1.85 MBq/mg) usingquantitative whole-body autoradiography (QWBA). After osimertinib dosing,animals were euthanized from 0.5 hours postdose and at intervals up to 60 dayspostdose (Table 1). Sagittal sections (nominal thickness 30 mM) were obtainedthrough the carcass to include the following tissues: exorbital lachrymal gland (males)or ovaries (females), intra-orbital lachrymal gland, Harderian gland, adrenal gland,thyroid, brain, and spinal cord. Sections were freeze-dried and placed in contact withimaging plates (Raytek Scientific Ltd., Sheffield, UK). [14C]-Blood standards ofappropriate activity (also sectioned at a nominal thickness of 30 mm) were placed incontact with all imaging plates. The sampling time after administration of[14C]-osimertinib was 0.5, 1, 6, and 24 hours, and 2, 7, 21, and 60 days in partiallypigmented male rats and 1, 6, and 24 hours in albino male and female rats.Concentrations of radioactivity in tissues were quantified from whole-bodyautoradiograms by Covance Laboratories Ltd. (Harrogate, UK) using a validatedimage analysis system. The lower limits of quantification were 1186.4 dpm/g(0.023 nmol/g) in the male partially pigmented animals, 860.9 dpm/g (0.017 nmol/g)in male albino rats and 670.8 dpm/g (0.013 nmol/g) in female albino rats. The upperlimits of quantification were 7,437,550 dpm/g (145 nmol/g) in all three groups.

Metabolism in Recombinant Expressed Cytochrome Isozymes

The metabolic turnover of osimertinib was assessed by incubation ofosimertinib or positive control substrates with microsomes prepared from insectcell lines that heterologously expressed the individual recombinant humancytochrome isoforms CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9,CYP2C19, CYP2D6, CYP2E1, CYP3A4, and CYP3A5 (BD Gentest; CorningLife Sciences, Corning, NY). Metabolic stability incubations were performed at37�C. The incubations contained 0.1 M phosphate buffer (pH 7.4), microsomes(0.5 mg/ml), and NADPH (5 mM final concentration) and were started by theaddition of osimertinib or positive control substrates (final incubation concentra-tion, 1 mM) after a 10-minute preincubation period. Aliquots were removed intoacetonitrile (containing internal standard) at various time points up to 30 minutes,diluted with water, and centrifuged. The concentration in the supernatant wasdetermined by HPLC-MS/MS.

The metabolism of AZ5104 and AZ7550 was assessed by incubation ofAZ5104/AZ7550 or positive control substrates with bactosomes prepared fromEscherichia coli bacteria. Heterologously expressed individual cytochrome iso-forms (100 pmol/ml final concentration) were preincubated in the presence ofAZ5104/AZ7550 (2 mM) or positive control substrates for 15 minutes at 37�C.Incubations were conducted in 0.1 M potassium phosphate buffer (pH 7.4) with afinal solvent concentration, 1%. Incubations were initiated by addition of NADPH(1 mM final concentration), and termination was by removal of samples intoacetonitrile (containing internal standard). Samples were taken at up to 30 minutesand analyzed by ultraperformance liquid chromatography with MS detection.

A substrate depletion approach was used to determine the intrinsic clearance(CLint) for each isoform. The contribution of each isoform (relative to the tenisoforms described) was calculated from the mean CLint using intersystemextrapolation factors and relative abundance factors. Additionally, in a separateexperiment, the extent of formation of AZ5104 and AZ7550 by each cytochromeisoform (relative to the formation by CYP3A4) was assessed after incubation ofosimertinib (1 mM) in recombinant expressed cytochrome isozymes.

Metabolism and Metabolite Formation in Mouse, Rat, Dog, and HumanHepatocytes

An in vitro study to investigate the metabolism of osimertinib after incubationwith mouse (CD1), rat (Han Wistar, males), dog (beagle, males), and humanhepatocytes was undertaken. Mouse hepatocytes were obtained in-house at

1202 Dickinson et al.

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

AstraZeneca (London, UK); rat, dog, and human hepatocytes were from CelsisIVT. Osimertinib was incubated (5 mM) with mouse, rat, dog, and humanhepatocytes (1 million cells/ml) at 37�C for 1 hour. Incubations were stoppedusing two volumes of acetonitrile. Samples were spun at 4000g for 15 minutes,and the supernatant was stored at –20�C until required for analysis. Semi-quantitative assessment of osimertinib metabolite profiles were made by analysisof UV traces at 320–330 nm to determine the proportion of the total area that wasrepresented by each component. The limit of quantification for metabolites byLC-UV was 1% relative to osimertinib.

An in vitro study was undertaken to investigate which human cytochromeisoforms form the metabolites AZ5104 and AZ7550 in human hepatocytes; thestudy also investigated whether flavin-containing monooxygenases (FMOs)produce the metabolite M1. Metabolic stability incubations at 37�C includedhuman hepatocytes at 1 million cells/ml and a selective inhibitor (or dimethylsulfoxide for control incubations) for a 10-minute preincubation period.Incubations were started by the addition of osimertinib or a control substrate(1mM final concentration) for 120 minutes, with sampling at suitable time points.Samples were diluted with water, centrifuged at 3000 rpm for 15 minutes, and theconcentration in the supernatant determined by HPLC-MS/MS. The mean rate offormation (pmol/min per million cells) for each metabolite was calculated.Additionally, the substrate depletion method was used to determine CLint.

Absorption, Distribution, Metabolism, and Excretion of Osimertinib inHealthy Volunteers

A phase 1 open-label, single-center pharmacokinetic and mass balance study(ClinicalTrials.gov identifier NCT02096679) was undertaken in healthy malevolunteers. The objective was to determine the rates and routes of excretion of[14C]-osimertinib by assessment of concentrations of total [14C] radioactivity andof osimertinib and its metabolites in whole blood, plasma, and urine as well aspercentage recovery of the radioactive dose in urine and feces.

Healthy male volunteers aged 30–65 years were recruited at a single site(Quotient Clinical Ltd., Nottingham, UK). Inclusion criteria included regularbowel movements (i.e., average production of at least one stool per day). The useof prescribed or nonprescribed concomitant medications was not permitted in the4 weeks (or longer, depending on the medication’s half-life) before the firstadministration of osimertinib (except for paracetamol and nonsteroidal decon-gestants at the discretion of the investigator). In addition, the use of medicationswith CYP450 enzyme-inducing properties was not permitted in the 4 weeksbefore the first administration of osimertinib. Subjects were also excluded if theyhad any intake of any product containing grapefruit or Seville oranges within7 days of the first administration of osimertinib.

Eligible subjects were admitted to the study center on day21. Subjects fastedovernight for 10 hours before administration of the investigational product on day

1. Each subject received a single 20-mg dose of [14C]-osimertinib as an oralsolution (free base equivalent) containing a nominal dose of 0.037 MBq (1 mCi)activity. Volunteers were resident at the study center for 21 days afteradministration of [14C]-osimertinib. During residency, samples of blood, urine,and feces were collected and safety assessments were undertaken. Volunteersreturned for further 24-hour residency periods to allow additional safetyassessments and collection of blood, urine, and feces samples (days 28 and 29,days 35 and 36, days 42 and 43, and days 84 and 85). Before dosing subjects,dosimetry calculation were performed by Public Health England (Didcot, UK)using the rat mass balance andQWBA study results as data inputs. This dosimetryassessment indicated that the committed effective dose was 1.0 � 1029 Sv/Bq; aradioactive dose of up to 5 MBq (132 mCi) would have been the maximum forWorld Health Organization category II (World Health Organization, 1977) inmen; however, this was estimated to be an insufficient radioactive dose todetermine metabolites present in plasma at 10% of parent by conventionalradiodetection techniques. Consequently, accelerator mass spectrometry (AMS)detection was chosen as the quantification tool. As this technique is much moresensitive, a much lower radioactive dose was feasible. This further reduced therisk to subjects and also the need to excessively dilute samples.

Pharmacokinetic Analysis. Appropriate pharmacokinetic parameters forplasma osimertinib, AZ7550, and AZ5104 and whole blood and plasma[14C] radioactivity were assessed. The pharmacokinetic analyses were performedat Quintiles (Overland Park, KS). Pharmacokinetic parameters were derived usingstandard noncompartmental methods with Phoenix WinNonLin ProfessionalVersion 6.3 (Pharsight Corporation, Mountain View, CA). Actual sampling timeswere used in the parameter calculations. Area under the concentration-time curve(AUC) was calculated using the linear up/log down method. Where appropriate,the AUC from time zero to the time of the last quantifiable concentration aftersingle dosing was extrapolated to infinity using lz to obtain AUC.

Mass Balance. Radioactive [14C] recovery in feces was calculated conven-tionally by the summation of the amount excreted in each collection intervalduring the residential period (up to 504 hours postdose). Recovery of fecalradioactivity during the outpatient portion of the study (504–2016 hours postdose)was calculated from the amount recovered in the four scheduled 24-hourcollection intervals using the area under the excretion rate–time curve method.Two subjects had their urine mixed during the 24–48 hours postdose collectionperiod, and for another two subjects, their urine was mixed during the 480 to504 hours postdose collection period; thus radioactivity recovery in urine for theentire duration of the study was calculated using the area under the excretion rate–time curve method excluding the urine samples that had been mixed.

Metabolite Identification. After analysis of plasma, fecal, and urine samplesby AMS for total radioactive content, appropriate samples were pooled acrossvolunteers and time points to give one plasma, one urine, and one fecal sample for

TABLE 1

Tissue-to-blood ratios of radioactivity in selected tissues after oral administration of [14C]-osimertinib (4 mg/kg) to male partially pigmented ratsand male and female albino rats

Male Partially Pigmented Rats Male Albino Rats Female Albino Rats

0.5 h 1 h 6 h 24 h 2 days 7 days 21 days 60 days 1 h 6 h 24 h 1 h 6 h 24 h

Bile ducts 63.4 56.5 28.7 2.63 1.57 0.39 NC NC 66.8 27.4 2.99 29.1 19.0 3.25Blood 1.00 1.00 1.00 1.00 1.00 1.00 1.00 NC 1.00 1.00 1.00 1.00 1.00 1.00Brain 1.06 2.17 1.85 0.37 0.19 0.22 0.23 NC 1.41 1.80 0.24 2.09 2.25 0.49Liver 32.6 24.5 13.3 3.23 2.18 1.01 0.41 NC 29.6 17.3 4.47 35.8 18.6 4.60Lung 14.0 21.3 20.4 3.11 1.24 0.85 0.81 NC 25.2 33.9 3.21 37.3 43.1 6.10Muscle 0.90 1.68 2.01 0.46 0.38 0.41 0.27 NC 1.95 2.33 0.41 2.51 2.15 0.55Ovary ND ND ND ND ND ND ND ND ND ND ND 5.38 14.6 1.94Pituitary 11.6 10.5 23.8 5.35 2.31 1.22 1.05 NC 14.9 32.9 14.3 12.0 31.9 13.6Renal cortex 11.7 15.7 22.2 11.4 8.40 4.07 1.06 NC 18.5 26.4 13.9 18.3 15.2 4.69Renal medulla 3.82 9.30 11.5 2.89 1.89 1.12 0.60 NC 10.6 14.5 2.87 17.5 20.00 3.00Skin (non-pigmented) 0.640 1.40 2.18 0.578 0.310 0.199 0.147 NC ND ND ND ND ND NDSkin (pigmented) 1.01 2.10 2.87 3.15 2.02 5.34 0.170 NC ND ND ND ND ND NDSkin ND ND ND ND ND ND ND ND 1.30 2.22 0.49 0.77 1.82 1.18Spleen 11.1 19.4 22.7 5.38 3.09 2.77 3.40 NC 23.8 26.6 4.07 29.2 32.7 10.8Testis 0.44 0.77 2.47 1.08 0.77 0.76 0.68 NC 0.58 1.71 0.70 — — —

Thyroid 13.9 12.1 10.8 2.63 1.71 1.84 1.27 NC 18.7 20.1 2.94 20.3 15.1 3.16Uvea and RPE 7.62 15.0 71.9 97.2 173 151 180 NC 3.34 3.72 1.06 2.94 1.39 0.79

NC, not calculable; ND, not done; RPE, retinal pigment epithelium; —, not applicable.

Metabolic Disposition of Osimertinib in Rats, Dogs, and Humans 1203

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

metabolite profile analyses at Xceleron Inc. (Germantown, MD) and one each formetabolite identification at AstraZeneca. Based on the levels of radioactivityobserved, 24–168 hours pooled samples were created, which balanced charac-terizing as much of the AUC as possible with retaining reasonable sensitivity.Osimertinib and its metabolites were extracted from plasma using acetonitrile andfrom fecal homogenate using acetonitrile, followed by acetonitrile water and thenwater; urine was analyzed without extraction. The metabolite profiles in pooledurine and feces were determined using HPLC-UV combined with offlineradioactivity monitoring by AMS for detection/quantification, MS for structuralelucidation, and UV for retention time matching between HPLC-UV fractionationfor AMS and HPLC-UV-MS systems.

Sample Extraction. The pooled plasma sample was extracted by addingacetonitrile (750 ml) to plasma (300 ml); the mixture was vortexed (1 minute) andcentrifuged (3000g, 5 minutes at room temperature); and the supernatant wasreduced to 50 ml under a stream of nitrogen. The extract was then reconstituted inacetonitrile/water (25/75 v/v, 300ml) and the radioactivity measured by AMS andcompared against an unextracted sample.

The cross-subject and time-pooled fecal homogenate sample (300 mg) wasmixed with acetonitrile (750 ml), vortexed (1 minute), and centrifuged (3000g,5 minutes, room temperature), and supernatant was transferred to a clean vial. Theremaining pellet was then mixed with acetonitrile/water (50/50 v/v, 750 ml),vortexed, and centrifuged as previously with the supernatant combined with theacetonitrile fraction. This combined fraction was diluted with water to achieve afinal composition of 20/80 v/v acetonitrile/water, vortexed and analyzed byHPLC-AMS and HPLC-MS/MS. The column recovery from this fraction was 93%. Theremaining pellet was also extracted with water (750 ml) where this fraction wasused only to assess the amount of radioactivity recovered from the sample.

The pooled urine sample was centrifuged (1924g, 10 minutes, 4�C) beforedirect HPLC-AMS and HPLC-MS/MS analysis and LSC for column recovery(100% recovered).

Circulating Metabolites in Patients at Steady State

The objectives of this study were to detect, characterize, and semiquantifymetabolites of non-radiolabeled osimertinib in human plasma. Steady-state (day22) plasma samples were obtained after daily oral dosing of osimertinib at 80 mgto fasted patients with NSCLC in the AURA phase 1 study (Planchard et al.,2014). Plasma samples were selected and pooled across six individual subjectsusing the method of Hamilton and colleagues (Hamilton et al., 1981) to create asample representative of the exposure to osimertinib within the pooling window.Osimertinib and related metabolites were detected, semiquantified (as apercentage of total detected drug-related material circulating in plasma), and thestructure elucidated by HPLC-UV-MS.

Bioanalysis

The analysis of radioactivity was performed by Xceleron Inc. (Germantown,MD). Urine and freeze-dried feces samples were analyzed for carbon contentusing a PerkinElmer 24000 Series 2 C, H, N analyzer (PerkinElmer, Boston,MA).The carbon content of each sample was calculated using urea calibrationstandards. The [14C]:[12C] ratio was measured in urine and freeze-dried fecesby AMS, and the concentration of radioactivity in each sample was calculated.

Plasma samples were analyzed by Covance Laboratories Ltd. (Harrogate,North Yorkshire, UK) for osimertinib, AZ5104, and AZ7550, simultaneouslyafter protein precipitation using acetonitrile. The lower limit of quantification was0.05, 0.0515, and 0.0515 nM for osimertinib, AZ5104, andAZ7550, respectively.Accuracy for osimertinib ranged from 101% to 103.3%, and precision (relativestandard deviation) ranged from 3.4 to 7.2. Similar method performance wasobserved for the two metabolites.

Results

Protein Binding Experiments

The stability of osimertinib was assessed from 0.1 to 100 mM inplasma incubations of up to 6 hours at 37�C to determine whetherequilibrium dialysis was a suitable technique. However, the extractionefficiency of osimertinib was particularly low in human plasma andhuman serum albumin (,1% and ,15%, respectively, Table 2)

compared with dog and a1-acid glycoprotein (both.80%) or in mouseand rat (15%–43%). As a consequence of this low extraction efficiencyin human plasma, it was considered that equilibrium dialysis was not asuitable technique, and consequently plasma protein binding was notdetermined in any species.Ultrafiltration was considered an alternative technique requiring a

shorter incubation; however, experiments (data not shown) indicatedthat the nonspecific binding to the ultrafiltration collection cups was toohigh to use this technique.As a consequence, the plasma protein binding of osimertinib has not

been determined. In the absence of a measured protein binding forAZD9291 and AZ7550, a computational approach was used to estimatebinding values in human plasma. An AstraZeneca proprietary machinelearning algorithm (Wood et al., 2011) for predicting the extent ofprotein binding in human plasma estimates the log KB/F (human) ofosimertinib to be 1.95% or 99% bound. Similarly, for AZ5104 andAZ7550, estimates of human plasma protein binding were 98% for both.Whereas this method is not validated, the learning algorithm has beentrained on a large and diverse set of in-house and commercialcompounds (currently n . 50,000, and covers a wide dynamic LogDrange of 21.5 to 4.5).

Covalent Binding Studies

After incubation of [3H]-osimertinib with human hepatocytes, theamount of material bound to hepatic proteins at 0 and 4 hours was 16.7and 693 pmol eq/mg protein, respectively. After incubation of[3H]-osimertinib with rat hepatocytes, the amount of material bound tohepatic proteins at 0 and 4 hours was 10.9 and 228 pmol eq/mg ofprotein, respectively. The turnover of [3H]-osimertinib during humanand rat hepatocyte incubations was estimated to be 15% and 71%,respectively, resulting in a fractional covalent binding of 0.29 and 0.02,respectively.Similar levels of covalent binding were measured in rat and human

plasma and in human serum albumin. For all samples, the covalentbinding measured at the initial time point was around 1 pmol eq/mgprotein or less. The extent of covalent binding increased linearly overtime in all matrices to 159, 196, and 158 pmol eq/mg of protein in ratplasma, human plasma, and human serum albumin, respectively, after a6-hour incubation (after accounting for binding at the initial time point).There was no evidence that osimertinib covalent binding was reversible.

Rat QWBA

After oral administration of [14C]-osimertinib to male partiallypigmented animals, drug-related material was rapidly absorbed andwidely distributed at the early sampling time points, with the highestconcentrations of radioactivity measured at 6 hours postdose in mosttissues. Radioactivity was detectable in 42% of tissues 60 days afterdosing with the terminal half-life of radioactivity greater than 100 hoursin most tissues (data not shown). The pattern of distribution in male

TABLE 2

Mean (n = 3) osimertinib remaining after incubation at 37�C for 6 hours in mouse,rat, dog, and human plasma, human serum albumin (HSA), solution and human a1-

acid glycoprotein solution (AGP)

Osimertinib Concentration (mM)Osimertinib Remaining after 6 h (%)

Mouse Rat Dog Human HSA AGP

0.1 15 36 90 ,1 11 1011 25 43 88 ,0.1 11 10810 43 34 83 ,0.1 9.7 105100 30 32 98 0.09 14 104

1204 Dickinson et al.

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

albino animals was similar to that observed in male partially pigmentedrats at comparable time points, with the exception of melanin-containingtissues. Specifically, levels of radioactivity in the uvea plus retinalpigment epithelium of albino animals were many-fold lower than thosein partially pigmented animals.Peak radioactivity concentrations in both male and female albino rats

occurred primarily at 6 hours postdose. The distribution in femalealbino rats was very similar to that in male albino rats, although theconcentration of radioactivity in the circulating blood was generallyhigher in female than in male rats at corresponding time points.Concentrations of radioactivity in the blood of animals from each dosegroup were lower (at early time points) than in most tissues; this wasindicated by a substantial number of tissue:blood concentration ratiosthat were above unity (Table 1). At later time points, the tissue-to-bloodratio was below 1, which may reflect differential irreversible bindingpreferential in blood/plasma compared with tissues. The radioactivehalf-life in rat blood was approximately 200 hours, which is consistentwith the half-life of erythrocytes. The brain contained a tissue:bloodradioactivity ratio greater than 1 up to 6 hours postdose, resulting in abrain-to-blood AUC ratio (0–504 hours) of approximately 0.3 after thissingle oral dose. Radioactivity was observed in the bile ducts up to and

including 7 days after dose administration and was also observed in therenal pyramid and urine. Evidence of a possible minor sex-relateddifference in the distribution of osimertinib was seem, with tissues offemale animals generally containing slightly higher levels than tissues inmale animals.

Metabolism in Recombinant Expressed Cytochrome Isozymes

Osimertinib, AZ5104, and AZ7550 were predominantly metabolizedby CYP3A4/5 (54.0%, 95.1%, and 99.9%, respectively (Table 3). Theextent of formation of AZ5104 and AZ7550 after incubation ofosimertinib in recombinant expressed cytochrome isozymes indicatedthat both were primarily formed by CYP3A4 and CYP3A5 (Supple-mental Table 1). Themetabolic products from these incubations have notbeen determined.

Metabolism and Metabolite Formation in Mouse, Rat, Dog, andHuman Hepatocytes

Across all species, 16 different metabolites were identified. Sevenmetabolites were detected in human hepatocytes; these were alsodetected in either rat or dog hepatocytes at similar or higher levels.Themetabolic profile in micewas similar to that in rats. Twometaboliteswere detected above the limit of quantification by UV in humanhepatocytes (Table 4). Ametabolic scheme summarizing themetabolitesobserved is shown in Fig. 1.The rate of turnover of osimertinib in human hepatocytes was

insufficient (,2 ml per min/M cells) for accurate assessment of theimpact of P450 or FMO inhibition on osimertinib turnover in vitro,although it was possible to assess the impact of inhibitors on theformation of metabolites AZ5104 and AZ7550. After incubationof osimertinib with human hepatocytes, the mean rate of formation ofAZ5104 was 0.044 pmol/min per million cells. The mean rate offormation did not decrease when inhibitors for CYP2C9, CYP2C19,CYP2C8, and FMO were included in the incubation. Inclusion of theCYP3A4/5 inhibitor ketoconazole decreased the mean rate of formationof AZ5104 to 0.028 pmol/min per million cells and inclusion of the panP450 inhibitor 1-aminobenzotriazole decreased the mean rate offormation to 0.013 pmol/min per million cells. Similar results wereobserved for the formation of AZ7550 (Supplemental Table 2).

TABLE 3

Percentage of metabolisma of osimertinib, AZ5104, and AZ7750 throughrecombinant expressed cytochrome isoforms

Cytochrome IsoformCompound

Osimertinib AZ5104 AZ7750

1A2 12.0 0.373 ND2A6 15.5 ND ND2B6 ND ND ND2C8 ND 3.19 0.1292C9 15.5 0.160 ND2C19 ND 0.958 ND2D6 ND 0.285 ND2E1 3.0 ND ND3A4 44.4 67.0 99.93A5 9.6 28.1 ND

aRelative contribution for the 10 displayed cytochrome isoforms only.ND, not detected.

TABLE 4

Semiquantification by ultraviolet (UV) spectroscopy of osimertinib metabolites in mouse, rat, dog, andhuman hepatocytes

Peak ID Proposed StructureProportion (%) of UV Chromatogram (320–330 nm)a

Mouse Rat Dog Human

M1 Oxidation (+O) ,1 ,1 1–10 ,1M2 Dealkylation (-C4H9N) ND ,1 ,1 ,1AZ7550 Demethylation (-CH2) 1–10 1–10 1–10 1–10M4 Oxidation (+O) 1–10 1–10 1–10 1–10M5 Oxidation (+O2) ,1 ,1 ,1 NDAZ5104 Demethylation (-CH2) ,1 ,1 ,1 ,1M7 Oxidation (+O) ND ND ,1 NDM8 Cysteine-glycine adduct ND ,1 ND ,1M9 Demethylation (-CH2) + oxidation (+O) 1–10 ,1 ,1 NDM10 Glutathione adduct 1–10 .10 1-10 ,1M11 Acetylation or deamination + glutathione 1–10 1–10 ND NDM12 Dealkylation (-C4H9N) + glutathione 1–10 ND ND NDM13 Oxidation (+O) + glutathione 1–10 1–10 ND NDM14 Oxidation (+O) + sulfation ND ND ,1 NDM15 Glutathione adduct ,1 ,1 ND NDM16 Oxidation (+O) + glucuronidation ,1 ,1 ND ND

a% represents percentage of parent (osimertinib) UV responseND, not detectable by mass spectrometry.

Metabolic Disposition of Osimertinib in Rats, Dogs, and Humans 1205

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

Absorption, Distribution, Metabolism, and Excretion ofOsimertinib in Healthy Volunteers

Eight healthy male subjects with a mean age of 42 years and a meanbody mass index of 26.9 kg/m2 were included in the study. Overall, fiveof the volunteers were white and three were black.Pharmacokinetics. Pharmacokinetic parameters for plasma

osimertinib, AZ5104, AZ7550, and plasma and whole blood radioac-tivity are summarized in Table 5. After a single oral dose of 20 mg[14C]-osimertinib, the median time to maximum plasma concentration(tmax) for osimertinib and AZ5104 was 6.00 hours, and tmax for AZ7550was 36 hours. In contrast, themedian tmax for radioactivity in plasmawas144 hours and in bloodwas 132 hours. Plasma osimertinib, AZ5104, and

AZ7550 accounted for only 0.8%, 0.08%, and 0.07%, respectively, oftotal plasma radioactivity based on geometric mean AUC ratios. Thegeometric mean AUC ratio for whole-blood total radioactivity to plasmatotal radioactivity was near unity (0.917). The mean plasma terminalhalf-life for osimertinib was 61.2 hours compared with a mean plasmaterminal half-life of 474 hours in plasma and 562 hours in blood. Meanapparent total clearance and apparent volume of distribution forosimertinib were 26.7 liters/h and 2260 liters, respectively, comparedwith 0.213 liter/h and 146 liters, respectively, for total plasmaradioactivity.Mass Balance. The mean cumulative total amount of radioactivity

recovered in urine, feces, and both combined in nanomolar equivalents,

Fig. 1. Metabolic scheme for osimertinib in-cubated in mouse (M), rat (R), dog (D), andhuman (H) hepatocytes.

1206 Dickinson et al.

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

as well as the percentage of total radioactive dose administered, aresummarized in Supplemental Table 3. In addition, the arithmeticmean cumulative dose recovered over time in urine, feces, and in totalis shown in Fig. 2. By the end of the study (84 days postdose),arithmetic mean total cumulative recovery (urine and feces com-bined) as a percentage of dose was 81.9% (range, 77.2%–89.8%). Onaverage, approximately 47% of the dose was recovered by 7 dayspostdose, with 68.9% recovered by the end of the residential period(21 days). Approximately 12% of the dose was recovered in feces by3 days. Mean recovery of total radioactivity was 14.2% of the dose(range, 10.5%–17.1%) in urine and 67.8% of the dose (range 62.9%–

74.4%) in feces.Geometric mean plasma osimertinib and total plasma radioactivity

concentration–time profiles are displayed graphically in Fig. 3. Totalradioactivity concentrations in plasma and whole blood were similarthroughout the time course; the ratio of whole blood to plasma totalradioactivity ranged from 0.720 to 1.30 over 83 days.Based on the molar equivalent, approximately 1.7% of the dose

was excreted in urine as osimertinib, AZ5104, and AZ7550(Supplemental Table 4). Consequently, renal clearance of osimertinibwas lowwithmean (6 standard deviation) renal clearance of 0.23560.116 liters/h (Supplemental Table 4).Metabolites. The extraction procedure from feces was optimized

for both the solvent extraction and reconstitution solvents to ensurethe maximum recovery of radioactivity and minimization of peakbroadening on the HPLC column. The use of sequential acetonitrile,acetonitrile/water, and water washes ensured that greater than 98% of[14C]-osimertinib and metabolites AZ5104 and AZ7550 spiked intocontrol fecal homogenate could be recovered using the extractionprocedure. Indeed, for these components, around 90% was extractedin the acetonitrile fraction. The optimum reconstitution solvent wasfound to be acetonitrile/water (25/75 v/v), where 98% of radioactivityfrom the extracts was recovered and peak shape remained sharp.Witha reconstitution solvent containing only 5% acetonitrile, recoverieswere typically around 50%. In addition, the column recovery forradioactivity from dosed sample extracts was 100%. Consequently,the optimized extraction procedure was able to quantitatively recoverosimertinib and the two metabolite standards (representative of themultiple metabolites identified from in vitro metabolism studies) andtherefore was considered suitable for the extraction of fecal samplesfrom osimertinib dosed human volunteers.Metabolite identification investigations concentrated on samples

pooled for the first 7 days, where approximately 47% of the dose wasexcreted (40% in feces and 7% in urine). Only a further approxi-mately 22% was excreted in samples from day 7 to 21, the end of theresidential period. Consequently, the fecal samples from 0 to 24 hours(,1% excreted) and 7 to 21 days were not included in the pool asthese would have significantly diluted the detectable material andcompromised the ability to determine metabolic transformations. Inaddition, the extraction efficiency for samples from different timepools (days 1–7, 7–14 and 14–21) was determined to be similar ataround 65% in the HPLC-AMS fraction, with only an additional 3%extracted in the water fraction; this was not included in the HPLC-AMS fraction, as it would have reduced sensitivity for metaboliteidentification. Consequently, the extraction efficiency was lowerfrom dosed volunteer samples (;70%) than from fecal homogenatesspiked with osimertinib and metabolites (;100%), which, consider-ing the irreversible binding nature of themolecules, strongly suggeststhat in dosed samples, around 30% of the radioactivity is unextractedbecause they are covalently bound to proteinaceousmaterial, which isa clearance mechanism in its own right. The extraction efficiency wassimilar for each of the time window samples, which suggests that

TABLE5

Sum

maryof

osim

ertin

ib,AZ5104,AZ7550,andplasmaandwhole

bloodradioactivity

pharmacokinetic

parametersin

healthymalevolunteers

afterasing

leoral

20-m

gdose

of[14C]-osim

ertin

ib

Osimertin

ib(n

=8)

AZ51

04(n

=8)

AZ75

50(n

=8)

[14C]Plasm

a(n

=8)

[14C]Blood

(n=8)

AUC,anM

×hGeometricmean(geometric

CV,%)(range)

1590

(36.2)

(803–2490)

143(29.6)

(106

–242)

(n=7)

132(19.9)

(92.5–

170)

(n=7)

190,000(17.7)

(153,000

–242,000)

(n=7)

174,000(12.8)

(143,000

–219,000)

(n=7)

AUC(0–72),a

nM×h

Geometricmean(range)

1010

(632–1490)

72.1

(53.8–

108)

50.9

(40.9–

64.0)

9670

(8060–

12,200)

10,700

(8240–

13,200)

AUC(0–t),anM

*hGeometricmean(range)

1580

(799–2480)

125(93.6–

234)

113(72.3–

153)

178,000(146,000

–227,000)

158,000(135,000

–206,000)

Cmax,nM

aGeometricmean(geometric

CV,%)(range)

29.9

(21.5)

(22.3–

38.5)

1.41

(27.3)

(0.918

–2.09)

0.957(17.2)

(0.804

–1.34)

217(15.8)

(189

–280)

260(19.5)(187

–330)

t max,h

Median(range)

6.00

(6.00–

8.12)

6.00

(6.00–

36.00)

36.00(6.00–

48.00)

144.00

(96.00

–168.00)

132.02

(96.00

–16

8.07

)t la

g,h

Median(range)

0.00

(0.00–

0.00)

0.51

(0.50–

1.00)

0.76

(0.00–

1.50)

0.00

(0.00–

0.00)

0.25

(0.00–

0.50)

t 1/2lz,h

Mean(range)

61.2

(48.2–

77.5)

55.2

(41.3–

77.0)

82.0

(60.2–

100)

474(415

–537)

562(477–685)

CL/F,liters/h

Mean(range)

26.7

(16.1–

49.9)

N/A

N/A

0.213(0.165

–0.262)

(n=7)

0.231(0.183

–0.280)

(n=7)

Vz/F,liters

Mean(range)

2260

(1630–

3600)

N/A

N/A

146(117

–168)

(n=7)

191(128–230)

(n=7)

MRAUCb

Geometricmean(range)

N/A

0.0909

(0.0566–

0.180)

(n=7)

0.0848

(0.0635–

0.162)

(n=7)

N/A

N/A

MRCmaxb

Geometricmean(range)

N/A

0.0471

(0.0346–

0.0937)

0.0319

(0.0232–

0.05502)

N/A

N/A

RAUCc

Geometricmean(range)

0.00826(0.00429

–0.0163)

(n=7)

0.000751

(0.000508–

0.00158)

(n=7)

0.000698

(0.000566–

0.00103)

(n=6)

N/A

0.917(0.809

–1.06)(n

=7)

AUC,areaun

dertheconcentration–tim

ecurve;CL/F,apparenttotalclearance;Cmax,m

axim

umdrug

concentration;

CV,coefficient

ofvariation;

N/A,n

otapplicable;t

1/2lz,term

inalhalf-life;t la

g,lag

time;t m

ax,tim

eto

maxim

umconcentration;

Vz/F,apparent

volumeof

distributio

nduring

term

inal

phase.

aFor

radioactivity

,AUCunits

arenm

ol/eq×handCmax

units

arenm

ol/eq.

bMetabolite

(AZ5104

orAZ7550)to

parent

(osimertin

ib)ratio

ofAUCor

Cmax.

c Ratio

ofAUCplasmaosim

ertin

ib,AZ5104,AZ7550,or

whole-blood

radioactivity

toplasmatotal.

Metabolic Disposition of Osimertinib in Rats, Dogs, and Humans 1207

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

there was no significant difference in the extent of irreversible binding offecal material between different time-pooled fecal samples and that themetabolic profile may be similar across these time periods. Conse-quently, further work concentrated on analyzing the 24- to 168-hourfecal sample, where 40% of the dose was excreted as this maximized themetabolite concentrations and minimized dilution effects of includinglater time points samples.In this 24- to 168-hour pooled and concentrated fecal sample, less

than 70% of the radioactivity was extracted (representing around only100 dpm of radioactivity), and the remainder was associated with thefecal pellet, most likely covalently bound to proteinaceous material.Therefore, as 40% of the radioactive dose was in this sample andaround 35% unextractable owing to covalent binding, this left 26%(40 � 65%) of the dosed radioactivity in the extract and available formetabolite profiling and identification studies. In this 24- to 168-hourfecal extract, eight osimertinib-related products were identified (Figs.4 and 5; Table 6), with AZ5104 being most abundant (5.6% of dose).In total, 14% of the administered radioactive dose could be accountedfor as specific metabolites in this pooled fecal sample extractcontaining 26% of the extractable dosed radioactivity. Therefore,we have been able to identify 55% (14 of 26%) of the extractablemetabolites in this sample with the remaining extracted radioactivityassociated with minor components each ,1% of the total dose in theAMS fractions (Fig. 5). In the 0- to 168-hour urine sample, fiveosimertinib-related products were identified (Fig. 4; Table 6).Themajor product was M25 (1.9% of dose, structure unknown), and theremaining products each accounted for,1% of the total administereddose. The total metabolites identified in the pooled urine sample(4.4%) represent approximately 63% of the radioactivity in thesample, with the remaining minor components unidentified in theradiochromatogram.Therefore, in both fecal and urinary extracts from samples taken up to

7 days postdose, where approximately 47% of the dose was excreted,metabolites could be identified to account for a total of 18% of theradioactive dose administered, which represents 38% of the totalradioactivity and 55% of the extractable radioactivity in these samples.The remaining radioactivity was associated with either minor identifiedmetabolites each representing ,1% of dosed material (;30%) orcovalently bound to proteinaceous material (;30%).The plasma metabolite profile from the human absorption, distribu-

tion, metabolism, and excretion (ADME) study was not investigated

because the extraction efficiency in the day 1 to 7 pooled sample was toolow (8%) to adequately determine metabolites at a concentration of 10%of osimertinib. In total, this plasma pool contained only 9.5 dpm/g, with0.65 dpm/g extracted. Instead, a plasma metabolic profile was de-termined and metabolites identified from a pooled patient steady-statesample, where metabolite concentrations are higher because accumula-tion has occurred (Planchard et al., 2016). However, the low plasmaextraction efficiency from in vivo and in vitro samples, coupled with thedemonstrable covalent binding to plasma proteins, suggest that most ofthe radioactivity in plasma from humans dosed with osimertinib is,indeed, irreversibly bound to plasma proteins, most likely serumalbumin.

Fig. 2. Arithmetic mean cumulative total radioactivity recoveries in urine, feces,and both combined in healthy male volunteers after a single oral dose of[14C]-osimertinib, 20 mg.

Fig. 3. Arithmetic mean concentration-time profiles of osimertinib and radioactivityin plasma in healthy male volunteers after a single oral dose of [14C]-osimertinib,20 mg (semilogarithmic scale).

Fig. 4. Proposed metabolic scheme for osimertinib excreted in feces and urine up to7 days (168 hours) in healthy male volunteers after a single oral dose of [14C]-osimertinib, 20 mg.

1208 Dickinson et al.

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

Circulating Metabolites in Patients at Steady State

Seven circulatory metabolites were characterized from patient plasmaafter 22 days of continuous dosing with osimertinib 80 mg, where thetwo most abundant metabolites were AZ5104 and AZ7550, observed at8% and 6% of total (Figs. 6 and 7; Table 6). No metabolite wasdetermined to be .10% of the total circulating osimertinib-relatedmaterial. Proposed excretory and circulatory metabolite schemes aredisplayed in Figs. 4 and 6, respectively.

Discussion

This article describes preclinical and clinical studies undertaken todetermine the metabolism and pharmacokinetics of osimertinib and itsmetabolites. These results help us understand the ADME for osimertinib

for the treatment of patients with advanced EGFR-mutant NSCLC whohave had progression after prior therapy with an EGFR-TKI.The covalent binding of osimertinib to EGFR via the cysteine

residue 797 was a deliberate compound design goal (Finlay et al.,2014). Osimertinib reactivity against the target protein was opti-mized versus less selective binding to cysteine residues on otherproteins while still maintaining nanomolar cellular potency againstthe target protein (Finlay et al., 2014). Studies were undertaken todetermine the covalent binding of osimertinib and/or its metabolitesto various biologic matrices. Covalent binding was observed inincubations containing [3H]-osimertinib and cryopreserved humanand rat hepatocytes and in incubations containing [14C]-osimertiniband rat and human plasma and human serum albumin (with similarlevels of covalent binding measured in human plasma, rat plasma,and human serum albumin). In addition, the radioactive extractionefficiency was much lower from human plasma (8%) than from feces(;70%), which suggests that covalent binding of osimertinib-relatedmaterial to plasma proteins is far more extensive than to excretedproteins.A clinical ADME study has also been reported for afatinib (Stopfer

et al., 2012), another covalent binding EGFR-TKI. Similar to the datareported here, [14C]-afatinib was excreted predominantly in feces(85.4% of dose), with urinary elimination a minor route. In contrast toosimertinib and unusually for a TKI, afatinib was excreted predomi-nantly as the intact parent molecule. The minor metabolism of afatinibobserved in excreta was mainly through conjugation of the Michaelacceptor to give glutathione, cysteine-glycine, or cysteine adducts, inaddition to N-oxidation of the dimethyl amino functionality. In plasma,afatinib accounted for most of the circulating drug-related material(72.9% over 24 hours), although this proportion decreased with time asthe half-life of plasma radioactivity (118 hours) was substantially longerthan that of afatinib (33.9 hours) owing to afatinib componentscovalently binding to albumin. Unlike osimertinib, afatinib cova-lent binding to human serum albumin has been reported to bereversible (http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Public_assessment_report/human/002280/WC500152394.pdf).

Fig. 5. Reconstructed AMS chromatogram of feces extract after a single oraladministration of [14C]-osimertinib to healthy male volunteers (20 mg, 36.1 KBq),indicating the metabolites identified.

TABLE 6

Quantitative estimate of radioactivity of metabolites in excreta in healthy male volunteers after a single 20-mg oral dose of [14C]-osimertinib and semiquantification byultraviolet (UV) light spectroscopy of circulating metabolites in area under the curve (AUC)(0-24 h) human plasma pool extract in patients with non-small cell lung cancer after

oral administration of multiple doses of osimertinib (80 mg/day, day 22)

Component Description Modification

Percentage of AdministeredDose in Healthy Volunteers

Percentage Relative to Total Drug-RelatedUV Response in Patientsb

Feces (Pooled 24–168 h) Urine (Pooled 0–168 h) Plasma Pool

Parent Osimertinib NA 1.2 0.71 76M1 Oxidation +O ND 0.13 ,2c

M2 Dealkylation –C4H9N ND ND 4AZ7550 Demethylation - CH2 2.3 0.44 6c

AZ5104 Demethylation - CH2 5.6 0.96 8c

M8 Cysteine-glycine adduct +C5H10O3N2S ND ND 2M17 Oxidation +O 0.33 ND ,2M18 Oxidation and glucuronidation +C6H8O7 ND ND ,2M19 Unknown Unknown 0.80 ND NDM20 Oxidation +O 0.76 ND NDM21 Cysteinyl adduct +C3 H6 O2 NS 1.5 ND NDM22 Unknown Unknown 0.48 ND NDM23 Dealkylation + oxidation -C4 H9 N, +O 0.47 ND NDM24 Unknown Unknown NDa 0.25 NDM25 Unknown Unknown ND 1.9 NDTotal 14 4.4 NA

NA, not applicable, ND, not detected.aPeak in accelerator mass spectrometry (AMS) profile had similar retention to M20 in feces AMS profile but no mass ion to confirm identity.bLower limit of quantitation = 2% of osimertinib UV (310–350 nm) response.cMetabolites confirmed with authentic standard.

Metabolic Disposition of Osimertinib in Rats, Dogs, and Humans 1209

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

Shibata and Chiba have investigated the correlation betweenclearance predicted from hepatocyte incubations and total bodyclearance for three covalent binding TKIs (afatinib, ibrutinib, andneratinib). The authors noted a disconnect between in vivo and in vitroclearance attributed to extrahepatic (covalent) conjugation to gluta-thione not captured in the hepatocyte system (Shibata and Chiba,2015). For osimertinib, the in vitro studies indicated that directconjugation with glutathione would be a minor metabolic pathway inhumans; however, it is clear from the human ADME study that a highproportion of [14C]-osimertinib-related material is excreted bound toproteinaceous material contributing a substantial fraction of overallclearance. As such, this suggests that preclinical hepatocyte andrecombinant cytochrome studies may overestimate the contribution ofcytochrome P450 (P450) metabolism to the clearance of osimertinibin the clinic. Hence, osimertinib, and other irreversible TKIseliminated through nonspecific covalent binding, may be lesssusceptible to P450 mediated drug-drug interactions than predictedfrom hepatocyte experiments.The metabolism of osimertinib in mouse, rat, dog, and human

hepatocytes was primarily to oxidative and dealkylated products withdirect conjugation to a range of glutathione (GSH), cysteine-glycine,glucuronide, and sulfate conjugates. All metabolites that formed inhuman hepatocytes were also seen in incubations with rat or doghepatocytes. Direct conjugation with GSH and cysteine-glycine inhuman hepatocyte incubations indicates potential elimination pathways

other than cytochrome. Experiments on osimertinib analogs suggestthat GSH conjugates are formed chemically and not catalyzed byglutathione-S-transferase (AZ data on file). GSH conjugates representeda small fraction of the identified metabolites in the human ADME study,which may reflect that osimertinib has higher affinity for proteinnucleophiles than GSH.Important information on the metabolic pathways and drug-

metabolizing cytochrome enzymes involved in the metabolic clear-ance of osimertinib was gathered in a series of in vitro reactionphenotyping studies. Whereas covalent binding may contribute to thein vitro turnover, the reduction in the formation rate of AZ5104 andAZ7550 in hepatocyte incubations with a CYP3A inhibitor and theformation of AZ5104 and AZ7550 by CYP3A isoforms suggests thatthe turnover observed in the recombinant cytochrome enzymeexperiments is, at least in part, due to oxidative metabolism. Fromthese studies, and considering the major metabolites in circulationand excreted, it was concluded that CYP3A4 and CYP3A5 were theprincipal cytochrome enzymes responsible for metabolism ofosimertinib, AZ5104, and AZ7550. Whereas other cytochromesmay contribute to the metabolism of osimertinib in terms of potentialdrug-drug interactions, coadministration of osimertinib with a potentinhibitor or inducer of CYP3A4/5 may affect the exposure ofosimertinib, AZ5104, and AZ7550; however, this would be moder-ated by the existence of alternative metabolic and elimination routes,including covalent binding to proteins. From the human ADMEstudy, it would appear that around 30% of the dose is excreted infeces bound to proteinaceous material, which is considered a majoreliminatory pathway.Data from the rat QWBA study provided useful information

regarding the tissue distribution of osimertinib and its metabolitesand informed the microradiolabeled dose and sample collectiondesign of the ADME study undertaken in healthy volunteers.Penetration of osimertinib drug-related radioactivity across theblood-brain barrier occurred and may provide therapeutic advantagein clinical indications where brain metastases are present (Ahn et al.,2015). The potential of osimertinib in the treatment of leptomeningealdisease is being investigated in a phase 1 study in patients withleptomeningeal disease (BLOOM, NCT02228369) (Yang et al.,2015). Radioactivity was observed in the bile ducts up to andincluding 7 days after dose administration, suggesting that drug-related material was secreted in the bile and probably contributed tothe overall fecal elimination. Additionally, radioactivity in the renal

Fig. 6. Proposed metabolic scheme for osimertinib in plasma in patients with NSCLCafter oral administration of multiple doses of osimertinib (80 mg/day, day 22).

Fig. 7. HPLC-UV (310–350 nm) chromatogram of the AUC(0–24 h) human plasmapooled extract after oral dosing of osimertinib (80 mg multiple dose, day 22, patientswith advanced NSCLC), indicating the metabolites identified.

1210 Dickinson et al.

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

pyramid and urine suggests that renal elimination was a probableroute of excretion. This was confirmed in the ADME study, where themean total osimertinib radioactivity in healthy volunteers was 14.2%of the dose in urine and 67.8% of the dose in feces at 84 days postdose.Of the excreted radioactivity, osimertinib constituted approximately1.2% in feces and 0.71% in urine (in samples up to 7 days), indicatingthat most of the elimination was via metabolism (either enzymatic orcovalent binding to proteins).Data from the ADME study in healthy volunteers provided additional

information about the distribution and excretion of osimertinib and itsmetabolites. On average, only approximately 12% of the dose wasrecovered in feces by 23 days postdose; this is much longer thangastrointestinal transit, suggesting that osimertinib was well absorbed(at least 88%) and the radioactivity subsequently found in feces wasnot a consequence of poor drug absorption but rather was related toexcretion after drug absorption. The geometric mean AUC ratio forwhole blood-to-plasma total radioactivity ratio was near unity,suggesting that osimertinib and its metabolites were distributed inwhole blood and plasma equally. The mean terminal half-life forosimertinib, AZ5104, and AZ7550 was 61.2, 55.2, and 82.0 hours,respectively; this was compared with 474 hours for total plasmaradioactivity. The long radioactive half-life is not unexpected basedon the findings of the rat QWBA study, which also showed prolongedretention of osimertinib-related material. The prolonged eliminationof total plasma radioactivity was likely because osimertinib bindscovalently to human plasma components, as demonstrated in thestudies investigating covalent binding of osimertinib and/or itsmetabolites to human hepatic proteins, human plasma, and humanserum albumin. Indeed, covalently bound radioactivity was likely amajor component of plasma, as demonstrated by osimertinib,AZ5104, and AZ7550 accounting for 0.8, 0.08, and 0.07% of totalplasma radioactivity, respectively, based on geometric mean AUCratios. This finding would explain the low percentage of radioactivitythat could be extracted from plasma samples (8%) in contrast to feces(;70%). Taken together, these findings strongly suggest that thedifference in AUC between radioactivity and osimertinib, AZ5104,and AZ7550 is due to osimertinib-related material bound to plasmaproteins, especially albumin. No findings in the clinical or preclinicalstudies to date indicate that covalent binding of osimertinib and/or itsmetabolites to proteins is associated with any toxicologic sequelae,such as idiosyncratic liver or immune-mediated toxicity (Zhou et al.,2005; Park et al., 2005). In humans, the most abundant circulatingmetabolites observed in steady-state plasma extracts from patientsdosed with non-radiolabeled osimertinib at 80 mg were AZ5104(N-demethylation on the indole) and AZ7550 (N-demethylation onthe dimethyl amine), representing 8% and 6% of total drug-relatedmaterial, respectively. No metabolite was more than 10% of totalcirculating osimertinib-related material. This finding is consistentwith the most abundant metabolites identified in human excretasamples and mouse, rat, dog, and human hepatocyte preparations. Theseries of studies described here provide important information on themetabolism and pharmacokinetics of osimertinib and its metabolites,AZ5104 and AZ7550, in laboratory animals, healthy volunteers, andpatients with NSCLC. Further clinical development of osimertinib isongoing.

Acknowledgments

The authors thank the following: Hal Galbraith (Quintiles) for performingthe pharmacokinetic analysis; Angela Jordan, Mike Hickey, and Ryan Braggfor supplying radioisotopes and stable labels of osimertinib and metabolites;Phil Fernyhough (Covance Laboratories Ltd., Harrogate, UK) for the ratautoradiography study; Steven English and Marie Croft (Xceleron Inc.,

Germantown,) for the human mass balance data; Linda Andersson (AstraZeneca,Mölndal, Sweden) for the hepatocyte covalent binding studies; RobinSmith (AstraZeneca, Macclesfield, UK) and Pharmaron for CYP phenotypicdata; Covance laboratories for osimertinib, AZ5104, and AZ7550 clinicalsample bioanalysis; the anonymous reviewers for helpful direction andcomments on developing the manuscript; Donna Tillotson, from iMedComms, an Ashfield Company, who provided writing support funded byAstraZeneca.

Authorship ContributionsParticipated in research design: Dickinson, Cantarini, Collier, Martin,

Ballard.Conducted experiments: Cantarini, Collier, Martin, Pickup.Contributed new reagents or analytical tools: Martin.Performed data analysis: Dickinson, Cantarini, Frewer, Martin, Pickup,

Ballard.Wrote or contributed to writing of the manuscript: Dickinson, Cantarini,

Collier, Pickup, Ballard.

References

Ahn MJ, Tsai CM, Yang JCH, Shepherd FA, Satouchi M, Kim DW, Bazhenova L, Hirashima T,Rukazenkov Y, and Cantarini M, et al. (2015) 3083 AZD9291 activity in patients with EGFR-mutant advanced non-small cell lung cancer (NSCLC) and brain metastases: data from Phase IIstudies. Eur J Cancer 51:S625–S626.

Arcila ME, Oxnard GR, Nafa K, Riely GJ, Solomon SB, Zakowski MF, Kris MG, Pao W, MillerVA, and Ladanyi M (2011) Rebiopsy of lung cancer patients with acquired resistance to EGFRinhibitors and enhanced detection of the T790M mutation using a locked nucleic acid-basedassay. Clin Cancer Res 17:1169–1180.

Cross DA, Ashton SE, Ghiorghiu S, Eberlein C, Nebhan CA, Spitzler PJ, Orme JP, Finlay MR,Ward RA, and Mellor MJ, et al. (2014) AZD9291, an irreversible EGFR TKI, overcomesT790M-mediated resistance to EGFR inhibitors in lung cancer. Cancer Discov 4:1046–1061.

Finlay MR, Anderton M, Ashton S, Ballard P, Bethel PA, Box MR, Bradbury RH, Brown SJ,Butterworth S, and Campbell A, et al. (2014) Discovery of a potent and selective EGFR inhibitor(AZD9291) of both sensitizing and T790M resistance mutations that spares the wild type form ofthe receptor. J Med Chem 57:8249–8267.

Hamilton RA, Garnett WR, and Kline BJ (1981) Determination of mean valproic acid serum levelby assay of a single pooled sample. Clin Pharmacol Ther 29:408–413.

Jänne PA, Ahn M-J, Kim D-W, Kim S-W, Planchard D, Ramalingam SS, Frewer P, Cantarini M,Ghiorghiu S, and Yang JC-H (2015a) Phase I study of AZD9291 in patients with EGFR-TKI-resistant advanced NSCLC: updated progression free survival and duration of response data. AnnOncol 26(suppl 1):i60 (Abstract LBA63A).

Jänne PA, Yang JC, Kim DW, Planchard D, Ohe Y, Ramalingam SS, Ahn MJ, Kim SW, Su WC,and Horn L, et al. (2015b) AZD9291 in EGFR inhibitor-resistant non-small-cell lung cancer. NEngl J Med 372:1689–1699.

Keedy VL, Temin S, Somerfield MR, Beasley MB, Johnson DH, McShane LM, Milton DT,Strawn JR, Wakelee HA, and Giaccone G (2011) American Society of Clinical Oncologyprovisional clinical opinion: epidermal growth factor receptor (EGFR) mutation testing for pa-tients with advanced non-small-cell lung cancer considering first-line EGFR tyrosine kinaseinhibitor therapy. J Clin Oncol 29:2121–2127.

Kobayashi S, Boggon TJ, Dayaram T, Jänne PA, Kocher O, Meyerson M, Johnson BE, Eck MJ,Tenen DG, and Halmos B (2005) EGFR mutation and resistance of non-small-cell lung cancer togefitinib. N Engl J Med 352:786–792.

Kuiper JL, Heideman DA, Thunnissen E, Paul MA, van Wijk AW, Postmus PE, and Smit EF(2014) Incidence of T790M mutation in (sequential) rebiopsies in EGFR-mutated NSCLC-patients. Lung Cancer 85:19–24.

Yang JC-H, Kim D-W, Kim S-W, Cho BC, Lee J-S, Ye X, Yin X, Yang Z, Jiang H, and Ahn M-J(2016) Osimertinib activity in patients (pts) with leptomeningeal (LM) disease from non-smallcell lung cancer (NSCLC): Updated results from BLOOM, a phase I study (Abstract). J ClinOncol 34:Abstract 9002.

Li W, Ren S, Li J, Li A, Fan L, Li X, Zhao C, He Y, Gao G, and Chen X, et al. (2014) T790Mmutation is associated with better efficacy of treatment beyond progression with EGFR-TKI inadvanced NSCLC patients. Lung Cancer 84:295–300.

Mitsudomi T, Tsai C-M, Shepherd FA, Bazhenova L, Lee J-S, Chang G-C, Crino L, Satouchi M,Chu Q, Lawrance R, et al. (2015) AZD9291 in pre-treated T790M positive advanced NSCLC:AURA2 Phase II study (Abstract). World Conference on Lung Cancer, Denver, CO, September6–9, 2015.

Oxnard GR, Arcila ME, Sima CS, Riely GJ, Chmielecki J, Kris MG, Pao W, Ladanyi M,and Miller VA (2011) Acquired resistance to EGFR tyrosine kinase inhibitors in EGFR-mutantlung cancer: distinct natural history of patients with tumors harboring the T790M mutation. ClinCancer Res 17:1616–1622.

Pao W, Miller VA, Politi KA, Riely GJ, Somwar R, Zakowski MF, Kris MG, and Varmus H(2005) Acquired resistance of lung adenocarcinomas to gefitinib or erlotinib is associated with asecond mutation in the EGFR kinase domain. PLoS Med 2:e73.

Park BK, Kitteringham NR, Maggs JL, Pirmohamed M, and Williams DP (2005) The role ofmetabolic activation in drug-induced hepatotoxicity. Annu Rev Pharmacol Toxicol 45:177–202.

Planchard D, Brown KH, Kim DW, Kim SW, Ohe Y, Felip E, Leese P, Cantarini M,Vishwanathan K, and Jänne PA, et al. (2016) Osimertinib Western and Asian clinicalpharmacokinetics in patients and healthy volunteers: implications for formulation, dose,and dosing frequency in pivotal clinical studies. Cancer Chemother Pharmacol 77:767–776.

Planchard D, Dickinson PA, Brown KH, Kim D, Kim S, Ohe Y, Felip E, Leese P, Cantarini M,and Ranson M (2014) Preliminary AZD9291 Western and Asian clinical pharmacokinetics in

Metabolic Disposition of Osimertinib in Rats, Dogs, and Humans 1211

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from

patients and healthy volunteers: implications for formulation, dose and dosing frequency inpivotal clinical studies (Abstract). Ann Oncol 25(Suppl 4):464P.

Reck M, Popat S, Reinmuth N, De Ruysscher D, Kerr KM, and Peters S; ESMO GuidelinesWorking Group (2014) Metastatic non-small-cell lung cancer (NSCLC): ESMO ClinicalPractice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 25 (Suppl 3):iii27–iii39.

Sequist LV, Waltman BA, Dias-Santagata D, Digumarthy S, Turke AB, Fidias P, Bergethon K,Shaw AT, Gettinger S, and Cosper AK, et al. (2011) Genotypic and histological evolution oflung cancers acquiring resistance to EGFR inhibitors. Sci Transl Med 3:75ra26.

Shibata Y and Chiba M (2015) The role of extrahepatic metabolism in the pharmacokinetics ofthe targeted covalent inhibitors afatinib, ibrutinib, and neratinib. Drug Metab Dispos 43:375–384.

Stopfer P, Marzin K, Narjes H, Gansser D, Shahidi M, Uttereuther-Fischer M, and Ebner T (2012)Afatinib pharmacokinetics and metabolism after oral administration to healthy male volunteers.Cancer Chemother Pharmacol 69:1051–1061.

Wood D, Buttar D, Cumming J, Davis A, Norinder U, and Rodgers S (2011) Automated QSARwith a Hierarchy of Global and Local Models. Mol Inform 2011:960–972.

World Health Organization(1977). Use of ionising radiation and radionuclides on human beings formedical research, training and nonmedical purposes. Technical report series no 611. Geneva.

Yang JC-H, Ahn M-J, Ramalingam SS, Sequist LV, Novello S, Su W-C, Hirashima T, Kim D-W,Lawrance R, Cantarini M, et al. (2015) AZD9291 in pre-treated T790M positive advancedNSCLC: AURA study Phase II extension cohort (Abstract). World Conference on Lung Cancer,Denver, CO, September 6–9, 2015.

Yu HA, Arcila ME, Rekhtman N, Sima CS, Zakowski MF, Pao W, Kris MG, Miller VA, LadanyiM, and Riely GJ (2013) Analysis of tumor specimens at the time of acquired resistance to EGFR-TKI therapy in 155 patients with EGFR-mutant lung cancers. Clin Cancer Res 19:2240–2247.

Zhou S, Chan E, Duan W, Huang M, and Chen YZ (2005) Drug bioactivation, covalent binding totarget proteins and toxicity relevance. Drug Metab Rev 37:41–213.

Address correspondence to: Peter Ballard, AstraZeneca, Alderley Park, Maccles-field, Cheshire, SK10 4TG, UK. E-mail: Peter.Ballard@astrazeneca.com

1212 Dickinson et al.

at ASPE

T Journals on D

ecember 26, 2021

dmd.aspetjournals.org

Dow

nloaded from