1521-009X/44/5/732–740$25.00 http://dx.doi.org/10.1124/dmd.115.067850DRUG METABOLISM AND DISPOSITION Drug Metab Dispos 44:732–740, May 2016Copyright ª 2016 by The American Society for Pharmacology and Experimental Therapeutics

Database Extraction of Metabolite Information of Drug Candidates:Analysis of 27 AstraZeneca Compounds with Human Absorption,

Distribution, Metabolism, and Excretion Data s

Jessica Iegre, Martin A. Hayes, Richard A. Thompson, Lars Weidolf, and Emre M. Isin

Cardiovascular and Metabolic Diseases, Innovative Medicines and Early Development Biotech Unit (J.I., M.A.H., L.W., E.M.I.)and Respiratory, Inflammation and Autoimmunity, Innovative Medicines and Early Development Biotech Unit (R.A.T.),

AstraZeneca, Mölndal, Sweden

Received October 28, 2015; accepted February 10, 2016

ABSTRACT

As part of the drug discovery and development process, it isimportant to understand the humanmetabolism of a candidate drugprior to clinical studies. Preclinical in vitro and in vivo experimentsacross species are conducted to build knowledge concerning humancirculating metabolites in preparation for clinical studies; therefore,the quality of these experiments is critical. Within AstraZeneca, allmetabolite identification (Met-ID) information is stored in a globaldatabase using ACDLabs software. In this study, the Met-ID in-formation derived from in vitro and in vivo studies for 27 AstraZenecadrug candidates that underwent human absorption, distribution,metabolism, and excretion studies was extracted from the data-base. The retrospective analysis showed that 81% of humancirculating metabolites were previously observed in preclinical invitro and/or in vivo experiments. A detailed analysis was carried out

to understand which human circulating metabolites were notcaptured in the preclinical experiments. Metabolites observed inhuman hepatocytes and rat plasma but not seen in circulation inhumans (extraneous metabolites) were also investigated. Themajority of human specific circulating metabolites derive frommultistep biotransformation reactions that may not be observed inin vitro studies within the limited time frame in which cryopreservedhepatocytes are active. Factors leading to the formation of extra-neous metabolites in preclinical studies seemed to be related tospecies differences with respect to transporter activity, secondarymetabolism, and enzyme kinetics. This retrospective analysis as-sesses the predictive value of Met-ID experiments and improves ourability to discriminate between metabolites expected to circulate inhumans and irrelevant metabolites seen in preclinical studies.

Introduction

A typical drug discovery and development campaign includesmetabolite identification (Met-ID) studies performed on a largenumber of compounds with the aim of optimizing metabolic clearance,designing away from reactive metabolite formation, creating aware-ness of active metabolites, and, most importantly, building confidencewith respect to which human circulating metabolites to expect inclinical studies (Isin et al., 2012; Stepan et al., 2013). This knowledgewill support the planning and execution of appropriate safety studies(e.g., selection of toxicology animal species prior to clinical trials).Without a streamlined way of capturing this structural information,metabolite knowledge generated on the many hundreds of projectcompounds studied is easily lost, as is the ability for collectivelearning. To remedy this issue, we have established a companywidemetabolite database with the aim of capturing information onmetabolites generated at all stages of the drug discovery and devel-opment process from both in vitro and in vivo experiments carried outat all AstraZeneca sites.Before taking drug candidates to the clinic to assess safety,

tolerability, pharmacokinetic properties, and efficacy, it is essential to

build confidence in the safety of the new chemical entity using a host ofin silico as well as in vitro and in vivo preclinical models. In addition tothe safety of the parent drug candidate, it is equally important toevaluate the safety of the expected human metabolites (Frederick andObach, 2010; Nedderman et al., 2011). Prior to exposing humanvolunteers or patients to a drug candidate, Met-ID studies in vitro and inanimal models constitute the basis as surrogates for safety testing ofexpected human metabolites (Luffer-Atlas, 2012). These approacheshave the limitations inherent to the in vitro models (e.g., hepatocytesneglecting extrahepatic metabolism, viability of cells), as well aslimitations owing to differences in biotransformation and drugdisposition between the various species used in the safety studies,and they thus may be of questionable predictive value. From abiotransformation perspective, the output from these and otherinvestigational studies is a starting point in building an understandingof the metabolic routes in humans and ultimately an understanding ofthe full disposition of the drug candidate of interest. It is also animportant part of the preparations for the first clinical studies in whichthe drug candidate will be administered to humans for the first time(“first-in-human studies”). Although urine and plasma samples shouldboth be collected for metabolite investigations in these first clinicalstudies, the focus is on metabolites in the systemic circulation forcomparison with those metabolites circulating in the animal speciesused in the early preclinical safety studies. Plasma samples from these

dx.doi.org/10.1124/dmd.115.067850.s This article has supplemental material available at dmd.aspetjournals.org.

ABBREVIATIONS: ADME, absorption, distribution, metabolism, and excretion; HLM, human liver microsome; LM, liver microsome; Met-ID,metabolite identification; OAT, organic anion transporter.

732

http://dmd.aspetjournals.org/content/suppl/2016/04/13/dmd.115.067850.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

preclinical safety and the first human studies are analyzed forcomparison of metabolite profiles and their relative exposure to eachspecies (Gao and Obach, 2011; Ma and Chowdhury, 2011). Earlyidentification of significantly different metabolite profiles betweenspecies may give time to proactively mitigate and preferably avoidresource-consuming problem-solving studies. Such studies may in-volve, but are definitely not limited to, synthesis of the metabolite(s)and dosing to animals to reach adequate exposure, or even identifyingnew animal species that will form the human metabolite to adequateexposure levels and repeating the in vivo toxicology program (Smithand Obach, 2006, 2009; Guengerich, 2009; Leclercq et al., 2009).In this report, we describe an in-house database in which in vitro and

in vivo animal and human biotransformation data on compounds fromearly as well as late-stage drug development projects have beencaptured and stored over a period of close to 1 decade. This databasecurrently contains .12,500 biotransformation schemes for .5000compounds. We present here our first investigations on how these datacan form a knowledge base from which we can build an understandingof the quality of in vitro and preclinical systems for the prediction ofcirculating metabolites in humans.

Materials and Methods

Met-ID Studies. Met-ID information was generated from the analysis ofAstraZeneca proprietary compounds in in vitro incubations in liver micro-somes (LMs) and hepatocytes from humans, rats, dogs, mice, and rabbits.Similarly, Met-ID information from reactive metabolite trapping experimentsin human liver microsomes (HLMs) with glutathione, potassium cyanide,and methoxylamine and from Aroclor 1254–induced rat liver S9 fraction

experiments were generated. Met-ID data from in vivo samples includingplasma, urine, bile, and feces from dosed rats, dogs, mice, rabbits, cynomolgusmonkeys, and marmosets and plasma, urine and feces from humans werelikewise generated.

Fig. 1. Biotransformation data capture and data flow showing publishing of Met-ID data into IBIS via the global metabolite database. AZ, AstraZeneca; DB, database; MSE,mass spectrometry at elevated collision energy; MSMS, tandem mass spectrometry.

Fig. 2. A graphic illustration of the filters applied to the search to obtain the subsetanalyzed. The first three filters were applied using the text-search option, whereasthe manual reorganization part was done together with a substructure search toidentify the matrices in which the same metabolites were observed.

Database Extraction of Metabolite Information of Drug Candidates 733

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

The analysis of in vitro and in vivoMet-ID experiments is typically performedby liquid chromatography coupled to tandem mass spectrometry on a Watersultra high-performance liquid chromatograph system with a SYNAPT quadru-pole time-of-flight mass spectrometer (Waters, Wilmslow, UK) or on Thermo-Finnigan OrbitTrap instrumentation (Thermo Scientific, Waltham, MA) (Ekdahlet al., 2013). Tandem mass spectra of the parent compound and metabolites areacquired and analyzed using a combination of vendor software (e.g., MetaboLynx;Waters) and interpretation is carried out by biotransformation experts. Structuresof both the parent and metabolites are drawn as a biotransformation scheme inACDLabs Spectrus DB Enterprise software (Advanced Chemistry Development,Toronto, ON, Canada) with reaction arrows linking metabolites to the parentmolecule. Metabolite numbering, quantification (by mass spectrometry responseor radioactivity) of each metabolite, and biotransformation type are alsocaptured at this stage. The scheme is then uploaded to the database, along witha standardized form capturing globally agreed experimental metadata fields(e.g., responsible scientist, test system, project code, date, etc.).

PDF reports of the new database record are also generated and shared withdrug discovery teams via upload to the IBIS corporate drug discovery database(Fig. 1).

ACD Database. The ability of the software to draw and capture Markushstructures (Markush, 1924; Cosgrove et al., 2012) to cover vagueness instructural assignments is a useful feature that facilitates reporting of metabolitesin cases in which the site of metabolism cannot be narrowed down to a singleatom (Yerin and Peirson, 2012). The common platform nature of the databaseallowed biotransformation scientists globally across the AstraZeneca organiza-tion to capture metabolite structural information in the form of biotransformationschemes and experimental metadata.

The ACD database currently contains .12,500 schemes and approximately50,000 metabolic reactions on proprietary AstraZeneca compounds, of whichonly a small percentage are in the public domain.

AstraZeneca Metabolite Database Search. The AstraZeneca metabolitedatabase can be searched in two ways: by text or by structure searching.Experimental metadata (e.g., incubation time, compound and project identifiers,and responsible scientist) are captured as text fields alongside the biotransfor-mation scheme. These are accessible within the database by a simple text-searchfunction.

Exact, similarity, or substructure searches can be performed and are accessed viathe ChemSketch drawing module. Parent or metabolite structures or substructuresare drawn in ChemSketch or simply imported from the database and suitablymodified. The desired structure is selected and searched against the entire database.

For the purposes of this work, Met-ID information on AstraZeneca com-pounds that underwent radiolabeled studies in humans and for which plasmametabolite profiles were available was extracted using the text-search option.The list was refined further by carrying out text searching for compounds havinghuman hepatocytes and rat plasma Met-ID information (Fig. 2). Subsequently,the substructure search option was used to determine in which matrices eachmetabolite was observed.

Results

Database Extraction. As a result of the database extraction effort, alist of 27 candidate drugs that underwent radiolabeled cross-speciesin vitro, preclinical in vivo, and human absorption, distribution,

Fig. 3. Comprehensive metabolic scheme forcompound AZD1. The manually constructedscheme includes all of the metabolites ob-served in all of the matrices on which Met-IDexperiments were performed.

Fig. 4. Definition of the terms used in this article to describe theexperiments performed at different stages of the drug discovery anddevelopment process.

734 Iegre et al.

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

metabolism, and excretion (ADME) studies was compiled. For eachcompound, the database provided a list of records corresponding to thevarious biotransformation reactions reported to occur in differentmatrices (e.g., hepatocytes, LMs, induced rat liver S9 fraction, urine,plasma, feces, bile) of different species (e.g., humans, rats, dogs,monkeys, mice, rabbits). An example of an ACD database recordcontaining Met-ID information obtained from the analysis of male ratplasma is given in Supplemental Fig. 1. Prior to the descriptive andstructural analysis, a unique metabolite scheme, including all of themetabolites observed during these Met-ID studies, was built on theextracted data for each of the 27 compounds. One such summaryscheme is shown in Fig. 3. Moreover, tables listing all metabolitesversus matrices were created for each compound to facilitate the vari-ous analyses; as an example, the table created for AZD1 is given inSupplemental Table 1.Descriptive Analysis. The aim of the first part of the work was to

evaluate the quality of our in vitro and in vivo preclinical experiments interms of human metabolite prediction (Fig. 4). All 27 AstraZenecacompounds underwent Met-ID studies in human and rat hepatocytesand in vivo in rats and humans using radiolabeled compound.Depending on the drug project, some compounds were also testedeither in vitro and/or in vivo in other species (e.g., cynomolgusmonkeys, marmosets, mice, or dogs). However, because these specieswere not uniformly used as a test system for the whole subset ofcandidate drugs reported in this work, these data were included only inthe descriptive analysis.From the Met-ID studies on the 27 AstraZeneca candidate drugs, a

total of 564 metabolite structures were identified, including thosewith Markush labeling. Of the 564 metabolite structures, 234 wereobserved in human in vivo and/or in vitro samples and 144 of thesewere found in systemic circulation (i.e., observed in plasma). Of the144 human circulating metabolites, 117 (81%) were observed beforethe human ADME study in at least one other test system. The actualnumber of previously observed human circulating metabolites varied,however, between candidate drugs (Fig. 5). In 16 of the 27 cases, it wouldhave been possible to predict all human circulating metabolites based onMet-ID experiments prior to beginning first-in-human studies.A descriptive analysis was performed to assess the frequency at

which human excreted metabolites (i.e., observed in urine and/or fecessamples) were observed in preclinical in vitro or in vivo studies prior tobeginning first-in-human studies. The analysis showed that 73% of theexcreted human metabolites (157 of 216) were observed in at least onestudy type prior to the human ADME studies.An evaluation of the predictive value of the different in vitro and in

vivo systems used preclinically was carried out for human circulatingmetabolites. The Venn diagram in Fig. 6 shows that by performingin vitro experiments only (all studied test species and systems), 46%

(33% human hepatocytes only) of human circulating metabolites wereobserved. On the other hand, by performing in vivo studies only (allstudied test species and matrices), 77% of circulating metabolites inhumans were observed.Human Circulating Metabolites Not Observed in Preclinical

Studies. For the 27 candidate drugs studied, only 27 circulating humanmetabolites (of 144) were seen for the first time in humans, whereasthe remainder were previously observed in in vitro and/or in vivopreclinical studies (Fig. 6). The biotransformation routes leading to these27 previously unobserved metabolites were scrutinized to understand theunderlying reasons for this lack of predictability. An analysis of themetabolic schemes indicates that the metabolites missed were primarilyderived frommultistep reactions, and evidence for the pathways leading tothese metabolites had already been captured in in vitro experiments by theidentification of intermediate or precursor metabolites. In several cases,these multistep pathways resulted in the formation of a conjugatedmetabolite but examples of oxidative multistep biotransformations werealso identified. The initial enzymaticO-dealkylation of AZD24 is followedby phosphate hydrolysis leading toM24-a (Fig. 7), which was observed inhuman hepatocytes as well as in rat plasma and urine. However, metaboliteM24-b, derived from the sulphonation of the intermediate phenol (M24-a),was observed for the first time in human plasma.Another example of a previously unobserved human circulating

metabolite involves AZD8 (Guo et al., 2015). A hydroxylated metaboliteM8-a (Fig. 8) was observed in human plasma and urine, mouse and ratplasma, and human hepatocytes. The metabolite M8-b derives fromthe reaction of the drug (and/or M8-b) with endogenous CO2 to form acarbamic acid intermediary metabolite, which is subsequently conju-gated with glucuronic acid. This resulting N-carbamoyl glucuronideM8-b was captured in human plasma only and was hence unobservedprior to human studies (Schaefer, 2006).Only one compound (AZD10) had a metabolite found for the first

time in human plasma, in which no precursors were captured in in vitro

Fig. 5. Bar chart comparison of the number of human circulating metabolites (in red) with the ones previously observed in preclinical studies (in blue).

Fig. 6. This analysis includes all of the in vitro data available (hepatocytes and/ormicrosomes and/or S9 fractions from humans, rats, dogs, monkeys, rabbits, mice,and common marmosets) and in vivo data (plasma, urine, feces, and bile fromhumans, rats, dogs, monkeys, mice, and marmosets) for the 27 compounds. Thenumbers indicate the quantity of observed metabolites.

Database Extraction of Metabolite Information of Drug Candidates 735

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

systems. As shown in Fig. 9, the hydrolytic ring opening of thecandidate drug AZD10 gives the previously unobserved metaboliteM10-a.Extraneous Metabolites. The analysis of the 27 AstraZeneca

compounds indicates that 58% of the metabolites observed during thepreclinical Met-ID studies (in vitro and/or in vivo) are not relevant towhat is observed subsequently in human circulation. In other words,during preclinical Met-ID work, 330 of the metabolites formed andcharacterized are not found in circulation in humans.A deeper structural analysis of the metabolites observed in human

hepatocytes and rat plasma, but not identified in human plasma, wasperformed and the results are shown in Table 1. Human hepatocytesand rat plasma were studied since these matrices were considered themost relevant to the retrospective analysis objective of this work. Inparticular, human hepatocytes allowed an in vivo/in vitro intraspeciescomparison to be made and rat plasma enabled the cross-speciescomparison. The rat was chosen over other species because it wasused in radiolabeled Met-ID studies for all of the subset of compoundsanalyzed.Structural Analysis of Extraneous Metabolites Captured in

Human Hepatocytes. Several metabolites observed in human cryo-preserved hepatocytes but not in human plasma appear to be furthermetabolized in humans and hence these extraneous metabolites areintermediates or precursors of what is observed in in vivo studies. Anexample is depicted in Fig. 10whereM7-a is a metabolite observed in invitro systems only (human and rat hepatocytes). The metabolitesformed via further metabolism ofM7-a (N-oxidation toM7-b, oxidationto carboxylic acid M7-c, glucuronidation to M7-d) are not observed inhepatocytes but are captured in human urine (M7-b, M7-c, M7-d) andhuman plasma (M7-b).Interestingly, in several examples, metabolites seen in human

hepatocytes but not in human plasma (or urine) are formed via amidedealkylation of tertiary amides. Examples for this biotransformationpathway are the secondary amide metabolites M9-a, M9-b derived fromAZ9 (Fig. 11A), and M13-a derived from AZD13 (Fig. 11B), whichwere formed in human hepatocytes but were not found in human plasmaor urine.Structural Analysis of Extraneous Metabolites Captured in Rat

Plasma. Metabolites captured in rat plasma but not observed in humanplasma were investigated to understand whether certain biotransfor-mations are species specific.Of the metabolites seen in rat plasma but not in human plasma, a

relatively large number are organic anions. However, although they arenot observed in plasma, some of these metabolites are found in human

urine. The majority of metabolites captured in rat plasma, but notcirculating in humans, were formed via biotransformations occurringon alkyl substituents of heterocyclic aromatic rings. No evidence forthese pathways or further metabolism was found in humans in vitro orin vivo. An example is the C-methyl-hydroxylation of the 2-methyl-pyrazine ring of AZD10, leading to the formation of M10-b (Fig. 12).M10-c is a metabolite derived from further metabolism of M10-b. Theproposed pathways involve the hydrolysis of the amide to carboxylicacid and a decarboxylation of the carboxylic acid to form M10-c.Neither M10-b nor M10-c was seen in human matrices.Some of the extraneous metabolites reported to be present in rat

plasma suggest that biotransformations differ between species. Anexample for a reversible metabolic pathway as seen in rat and humanplasma is shown in Fig. 13. M7-e derived from compound AZD7 isobserved in vivo in rats but not in humans. On the other hand, M7-f isobserved in human plasma and has been reported to form in rathepatocytes but is not observed in rat plasma.

Discussion

The AstraZeneca metabolite database provides a central repositoryfor all metabolite structural information on AstraZeneca compounds.This database is a structure- and text-searchable resource that allowsglobal sharing of metabolite data across AstraZeneca sites, therapeuticareas, and projects and ensures that metabolite knowledge accumulatedover many years is permanently accessible. The database contains Met-ID data generated across the entire value chain from early discovery toproduct life cycle management. The majority of the data are fromcompounds synthesized during the discovery phase with human livermicrosomal or hepatocyte Met-ID data; however, a subset of develop-ment compounds contains a more comprehensive package of data,including Met-ID from human ADME studies.The AstraZeneca metabolite database differs from commercially

available drug metabolite databases (BIOVIA Metabolite Database;BIOVIA, San Diego, CA) since it is built on studies of proprietary drugsand drug-like compounds combined with strictly defined experimentalprotocols and analytical methodologies.The descriptive analysis of 27 AstraZeneca candidate drugs showed

that in vitro and in vivo preclinical experiments were able to generate81% of the human circulating metabolites and 73% of human excretedmetabolites captured in the human ADME studies. In vitro experimentsin human hepatocytes alone were able to predict 33% of humancirculating metabolites. Inclusion of data from other in vitro sys-tems in addition to human hepatocytes increases the predictability to

Fig. 7. Multistep biotransformation of AZD24 observed in humanplasma resulting in the formation of the previously unobservedsulphonated metabolite M24-b.

Fig. 8. M8-b shows a previously unobserved human metabolitederiving from multistep metabolism of AZD8.

736 Iegre et al.

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

46%. This number is comparable to the outcome of an analysis byAnderson et al. (2009), who showed that in vitro data predicted themajor human circulating metabolites for 41% of the studied drugcandidates. In another study, human hepatocytes were reported topredict complete circulating metabolite profiles for 65% of the studiedcompounds (Dalvie et al., 2009). This observation is not surprisingconsidering the caveats associated with in vitro systems (see belowand De Graaf et al., 2002; Dalvie et al., 2009), particularly thelimited incubation times as a result of the decaying metabolicactivity, which leads to the inability of predicting secondarymetabolites (Dalvie et al., 2008). Micropatterned hepatocyte co-culture systems are reported to have considerably improved abilityto predict major human metabolites over more conventional in vitrohuman systems such as microsomes and hepatocytes (Wang et al.,2010). That these types of systems can show enhanced metaboliccapability over primary hepatocyte suspensions was demonstratedby the incubation of a set of 27 drug compounds for 7 days in amicropatterned coculture of hepatocytes. It was found that of thecirculating human metabolites exceeding 10% of total circulatingdrug-related material, 75% were found in the in vitro coculturesystem compared with 53% found in human hepatocyte suspen-sions. Various other in vitro approaches to maintain metaboliccapability for extended periods of time have recently been explored,including three-dimensional cell cultures such as spheroids (Godoyet al., 2013), liver bioreactor systems (Darnell et al., 2012), andstem cell–derived hepatocytes (Ulvestad et al., 2013), as well as

different technical approaches such as the relay method (Ballardet al., 2014). Although these approaches show promise that furtherimproved in vitro systems will predict a greater number of humanmetabolites (particularly secondary or tertiary metabolites), it remainsuncertain whether the metabolites would be captured as excreted or incirculation in humans.In vitro experiments as a whole (i.e., human hepatocytes and LMs; rat

hepatocytes, LMs, and liver S9 fractions; and dog, mouse, rabbit,monkey, marmoset LMs and hepatocytes) gave very little informationin addition to what was observed in vivo. Less than 4% of humancirculating metabolites were unique to in vitro experiments. On thebasis of this analysis, nonhuman in vivo systems may be perceived as abetter system than human hepatocytes for predicting human circulatingmetabolites. However, the issue around extraneous metabolites (seebelow) observed in these in vivo systems must be taken into account fora proper comparison.An analysis of human circulating metabolites not captured in

preclinical experiments/studies was carried out and the resultsprovided a plausible explanation for why some metabolites wereseen for the first time in human circulation. Interestingly, only onemetabolite (Fig. 9) was observed for the first time in humans and noreaction type or molecular motif consistently leading to a uniquehuman metabolite was observed. Six of the metabolites classifiedas previously unseen are vague Markush structures. Because of thebroadness of the Markush labeling, identification of the exactbiotransformation that occurred is not possible; therefore, uncertaintyremains regarding their elucidation.In 13 of the 27 cases of previously unobserved human circulating

metabolites, evidence of biotransformation pathways seen in hu-man plasma is observed during in vitro cross-species experiments.Therefore, because of the limited viability of cryopreserved hepato-cytes, the metabolism found in vivo in humans is more extensiveand the final metabolite is thus not captured in the in vitro systems.Therefore, the predictability of human hepatocytes increases to 87%if metabolic pathways instead of specific human circulating metab-olites are considered, which is similar to that reported by Dalvie et al.(2009) with respect to metabolic pathways. When LMs, hepatocytes,microsomes, and S9 fractions of other species (i.e., rat, mouse, dog,rabbit) are also included, the predictability of in vitro systems goes upto 92% with regard to biotransformation pathways. This assessmentemphasizes the importance of the critical evaluation of observedmetabolites by biotransformation scientists in an attempt to predictand later elucidate complete metabolic pathways in humans. Otherexamples (4 of 27) of previously unseen metabolites suggest thatspecies specificity with respect to glucuronidation may play a role as

Fig. 9. Metabolism of a methylpyrazinone moiety leading to a previouslyunobserved human circulating metabolite.

TABLE 1

Number of human circulating metabolites and extraneous metabolites observed in the different matrices

The extraneous metabolites are subdivided into three classes: not observed but evidence of the same pathways captured in human plasma, not observed in plasma but seen as excreted, and truemisleading metabolites. The first two are the extraneous metabolites that are actually precursors or intermediates to the final metabolites captured in human plasma. The last class comprises metabolitesirrelevant to the human circulating metabolites.

MatrixHuman Circulating

Metabolites Observed

Metabolites Not Seen in Human asCirculating and Observed in the Matrices Listed Here (Extraneous)

Not Observed But Evidence of the SamePathways Captured in Human Plasma

Not Observed inPlasma But Excreted

Irrelevant Metabolites

n

Human hepatocytes 47 5 15 13 (5)a

Rat plasma 46 0 8 25 (19)b

aThe number of irrelevant metabolites is 13 but might be smaller (down to 5) if we consider that 3 of these metabolites are broad Markush structures and 5 might be excreted via bile and thereforefeces, which is not always analyzed (3 compounds do not have feces Met-ID data available in the database).

bThe number of irrelevant metabolites observed in rat plasma is 25 but might be smaller (down to 19) if we consider the vagueness around 6 Markush structures.

Database Extraction of Metabolite Information of Drug Candidates 737

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

to why human circulating metabolites are not seen in other species.Since conjugated metabolites are generally considered to be non-toxic, less concern has been expressed by the U.S. Food and DrugAdministration over their presence in the circulation (http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm079266.pdf).The number of human circulating metabolites predicted from human

hepatocytes is similar to the number predicted from rat plasma.The number of extraneous metabolites in the two matrices is alsocomparable (Table 1). However, structural analysis reveals thatextraneous metabolites found in human hepatocytes often tend to bea precursor or an intermediate of a human circulating metabolite(5 metabolites) or are seen as excreted metabolites (15 metabolites).This latter figure could be greater (20 metabolites) if metaboliteswere considered that could be excreted via bile and for whichmetabolite profile in feces were not analyzed. This could be the casefor metabolites deriving from amide dealkylation and for glucuro-nides and other polar metabolites that, once formed in the liver, maybe directly excreted via the bile or rapidly excreted via the urinewithout accumulation in plasma to detectable concentrations(Loi et al., 2013).Overall, these analyses showed that in vitro experiments in human

hepatocytes generate a few misleading metabolites owing to specificreactions taking place in human in vitro systems, and only fivemetabolites appear to be irrelevant to human in vivo metabolites. Oneof these five irrelevant metabolites underwent formylation biotransfor-mation, and this might be the result of an artifact of the in vitro

experiment itself (i.e., high concentration of formic acid). The otherfour metabolites found uniquely in human hepatocytes are metabolitesof an AstraZenecamarketed drug andmight reflect more extensiveMet-ID work done on this drug.When metabolites found in human hepatocytes are interpreted with

the aim of predicting human circulating metabolites, two pieces ofevidence must be considered: 1) the metabolites may be furthermetabolized in vivo in humans, and 2) finding metabolites in hepato-cytes does not indicate whether they will enter the circulation.On the other hand, extraneous metabolites in rat plasma are generally

irrelevant to metabolites seen in human plasma (25 metabolites) orexcreted metabolites (8 metabolites), suggesting that the prediction ofhuman metabolites from in vivo studies may be misleading.Misleading metabolites in rat plasma are mainly organic anions,

which are formed in human in vitro systems but are not furthermetabolized in humans compared with rats. Therefore, whether ametabolite will be found in the circulation may depend not only on itsformation per se but also on the species-specific expression of drugtransporters such as the organic anion transporters (OATs). Speciesdifferences with respect to location on the apical or basolateralmembranes of some human OATs and rat Oats (OAT2) are reportedin literature (Burckhardt, 2012). OATs such as Oat5 have beendemonstrated in rat kidney but not in humans, and their role is eitherto uptake xenobiotics from the primary urine into the cells or torelease organic anions into the lumen.A common trend was observed during the analysis of extraneous

metabolites in rat plasma. In three cases, biotransformations occurring

Fig. 11. Examples of metabolites deriving from amideN-dealkylation found in human hepatocytes but not inhuman plasma.

Fig. 10. Example of a metabolite (M7-a) observed in humanhepatocytes but not observed in humans in vivo. In the latter,products of further metabolism of M7-a were captured.

738 Iegre et al.

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

on substituents of heterocyclic aromatic rings led to metabolites thatwere observed in rats only. Different substrate specificity betweenotherwise similar drug-metabolizing enzymes in rats and humans mightprovide explanations for these observations. Therefore, in contrast withwhat was observed for extraneous metabolites found in human hepa-tocytes, extraneous metabolites found in rat plasma are not related tohuman metabolites.

Summary

The value of a corporate metabolite database was highlightedthrough the analysis of all metabolites generated during discoveryand development for a set of 27 candidate drugs. The analysis showedthat human hepatocytes alone were only able to predict 33% of thehuman circulating metabolites. However, the majority of metabolitesmissed were further transformations of the metabolites seen in hepato-cytes. If instead the criterion was the indication of metabolic pathways,the predictability of human hepatocytes to human plasma increased to87%. This indicates that the advances now being made in human hepaticin vitro systems, with improved viability and longer incubation times,should considerably improve our ability to predict which metaboliteshumans will be systemically exposed to.Rat plasma was similar to human hepatocytes in the number of

correctly predicted human circulating metabolites. However, the consid-erably larger number of metabolites irrelevant to human metabolism seenin rat plasma makes this a much less useful tool than human hepatocytes.

Acknowledgments

The authors dedicate this article to Prof.Neal Castagnoli, Jr. for his contributionsto the field of drug metabolism and AstraZeneca drug discovery and developmentprojects over a span of more than 30 years.

Authorship ContributionsParticipated in research design: Iegre, Hayes, Thompson, Weidolf, Isin.Performed data analysis: Iegre, Hayes, Weidolf.Wrote or contributed to the writing of the manuscript: Iegre, Hayes,

Thompson, Weidolf, Isin.

References

Anderson S, Luffer-Atlas D, and Knadler MP (2009) Predicting circulating human metabolites:how good are we? Chem Res Toxicol 22:243–256.

Ballard TE, Orozco CC, and Obach RS (2014) Generation of major human excretory and cir-culating drug metabolites using a hepatocyte relay method. Drug Metab Dispos 42:899–902.

Burckhardt G (2012) Drug transport by organic anion transporters (OATs). Pharmacol Ther 136:106–130.

Cosgrove DA, Green KM, Leach AG, Poirrette A, and Winter J (2012) A system for encodingand searching Markush structures. J Chem Inf Model 52:1936–1947.

Dalvie D, Chen W, Zhang C, Vaz AD, Smolarek TA, Cox LM, Lin J, and Obach RS (2008)Pharmacokinetics, metabolism, and excretion of torcetrapib, a cholesteryl ester transfer proteininhibitor, in humans. Drug Metab Dispos 36:2185–2198.

Dalvie D, Obach RS, Kang P, Prakash C, Loi CM, Hurst S, Nedderman A, Goulet L, Smith E,and Bu HZ, et al. (2009) Assessment of three human in vitro systems in the generation of majorhuman excretory and circulating metabolites. Chem Res Toxicol 22:357–368.

Darnell M, Ulvestad M, Ellis E, Weidolf L, and Andersson TB (2012) In vitro evaluation of majorin vivo drug metabolic pathways using primary human hepatocytes and HepaRG cells insuspension and a dynamic three-dimensional bioreactor system. J Pharmacol Exp Ther 343:134–144.

De Graaf IAM, Van Meijeren CE, Pektas F, and Koster HJ (2002) Comparison of in vitropreparations for semi-quantitative prediction of in vivo drug metabolism. Drug Metab Dispos30:1129–1136.

Ekdahl A, Aurell-Holmberg A, and Castagnoli N, Jr (2013) Identification of the metabolites oflesogaberan using linear trap quadrupole Orbitrap mass spectrometry and hydrophilic in-teraction liquid chromatography. Xenobiotica 43:461–467.

Frederick CB and Obach RS (2010) Metabolites in safety testing: “MIST” for the clinicalpharmacologist. Clin Pharmacol Ther 87:345–350.

Gao H and Obach RS (2011) Addressing MIST (metabolites in safety testing): bioanalyticalapproaches to address metabolite exposures in humans and animals. Curr Drug Metab 12:578–586.

Godoy P, Hewitt NJ, Albrecht U, Andersen ME, Ansari N, Bhattacharya S, Bode JG, Bolleyn J,Borner C, and Böttger J, et al. (2013) Recent advances in 2D and 3D in vitro systems usingprimary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and theiruse in investigating mechanisms of hepatotoxicity, cell signaling and ADME. Arch Toxicol 87:1315–1530.

Guengerich FP (2009) Introduction: human metabolites in safety testing (MIST) issue. Chem ResToxicol 22:237–238.

Guo J, Zhou D, Grimm SW, and Bui KH (2015) Pharmacokinetics, metabolism and excretion of[(14)C]-lanicemine (AZD6765), a novel low-trapping N-methyl-d-aspartic acid receptorchannel blocker, in healthy subjects. Xenobiotica 45:244–255.

Isin EM, Elmore CS, Nilsson GN, Thompson RA, and Weidolf L (2012) Use of radiolabeledcompounds in drug metabolism and pharmacokinetic studies. Chem Res Toxicol 25:532–542.

Leclercq L, Cuyckens F, Mannens GSJ, de Vries R, Timmerman P, and Evans DC (2009) Whichhuman metabolites have we MIST? Retrospective analysis, practical aspects, and perspectivesfor metabolite identification and quantification in pharmaceutical development. Chem ResToxicol 22:280–293.

Loi CM, Smith DA, and Dalvie D (2013) Which metabolites circulate? Drug Metab Dispos 41:933–951.

Luffer-Atlas D (2012) The early estimation of circulating drug metabolites in humans. ExpertOpin Drug Metab Toxicol 8:985–997.

Ma S and Chowdhury SK (2011) Analytical strategies for assessment of human metabolites inpreclinical safety testing. Anal Chem 83:5028–5036.

Markush EA (1924) inventor, Pharma Chemical Corp, assignee. Pyrazolone dye and process ofmaking the same. U.S. patent 1506316. 1924 Aug 26.

Nedderman AN, Dear GJ, North S, Obach RS, and Higton D (2011) From definition to imple-mentation: a cross-industry perspective of past, current and future MIST strategies. Xenobiotica41:605–622.

Schaefer WH (2006) Reaction of primary and secondary amines to form carbamic acid glucu-ronides. Curr Drug Metab 7:873–881.

Smith DA and Obach RS (2006) Metabolites and safety: what are the concerns, and how shouldwe address them? Chem Res Toxicol 19:1570–1579.

Fig. 12. Metabolism occurring on a pyrazinone moietyleading to metabolites observed in rat plasma but not inhuman plasma.

Fig. 13. AZD7 underwent metabolism showing the redox re-versibility between rats and humans.

Database Extraction of Metabolite Information of Drug Candidates 739

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from

Smith DA and Obach RS (2009) Metabolites in safety testing (MIST): considerations of mechanismsof toxicity with dose, abundance, and duration of treatment. Chem Res Toxicol 22:267–279.

Stepan AF, Mascitti V, Beaumont K, and Kalgutkar AS (2013) Metabolism-guided drug design.MedChemComm 4:631–652.

Ulvestad M, Nordell P, Asplund A, Rehnström M, Jacobsson S, Holmgren G, Davidson L, BrolénG, Edsbagge J, and Björquist P, et al. (2013) Drug metabolizing enzyme and transporter proteinprofiles of hepatocytes derived from human embryonic and induced pluripotent stem cells.Biochem Pharmacol 86:691–702.

Wang WW, Khetani SR, Krzyzewski S, Duignan DB, and Obach RS (2010) Assessment of amicropatterned hepatocyte coculture system to generate major human excretory and circulatingdrug metabolites. Drug Metab Dispos 38:1900–1905.

Yerin A and Peirson I (2012) Rendering the stages of structure elucidation: ACD/Labs Markushrepresentation (Abstract CINF-58), in Proceedings of the 243rd ACS National Meeting andExposition; 2012 Mar 25–29; San Diego, CA. American Chemical Society, Washington, DC.

Address correspondence to: Emre M. Isin, Cardiovascular and MetabolicDiseases, Innovative Medicines and Early Development Biotech Unit, AstraZeneca,Pepparedsleden 1, SE-431 83 Mölndal, Sweden. E-mail: Emre.Isin@astrazeneca.com

740 Iegre et al.

at ASPE

T Journals on Septem

ber 23, 2020dm

d.aspetjournals.orgD

ownloaded from