Interindividual Pharmacokinetics Variability of the a4b1

Integrin Antagonist, 4-[1-[3-Chloro-4-[N0-(2-methylphenyl)ureido]phenylacetyl]-(4S)-fluoro-(2S)-pyrrolidine-2-yl]methoxybenzoic Acid (D01-4582), in Beagles Is Associatedwith Albumin Genetic Polymorphisms

TAKASHI ITO,1 MASAYUKI TAKAHASHI,1 KENICHI SUDO,1 YUICHI SUGIYAMA2

1Drug Metabolism & Pharmacokinetics Research Laboratories, R&D Division, Daiichi Sankyo Co., Ltd., Tokyo, Japan

2Department of Molecular Pharmacokinetics, Graduate School of Pharmaceutical Sciences, The University of Tokyo,Tokyo, Japan

Received 24 April 2008; revised 12 June 2008; accepted 27 June 2008

Published online 14 August 2008 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21520

Abbreviationmethylphenyl)uryl]methoxybenzapparent volumKd, equilibriumbinding site; fu,

CorrespondenFax: 81-3-5436-E-mail: ito.taka

Journal of Pharm

� 2008 Wiley-Liss

ABSTRACT: The interindividual pharmacokinetic variability of the a4b1 integrinantagonist 4-[1-[3-Chloro-4-[N0-(2-methylphenyl)ureido]phenylacetyl]-(4S)-fluoro-(2S)-pyrrolidine-2-yl]methoxybenzoic acid (D01-4582) was observed in beagles. The involve-ment of albumin genetic polymorphism in this variability was investigated. The albumingenotype was analyzed by sequencing the albumin gene from liver and blood samples.The plasma protein binding of D01-4582 was analyzed using an ultracentrifuge method.The results of the sequencing analysis of albumin cDNA revealed two single nucleotidepolymorphisms, G1075T and A1422T, leading to amino acid changes of Ala335Ser, andGlu450Asp, respectively. These mutations show complete linkage. The prevalence of themutant allele in a population of 47 dogs was 40%. The unbound fractions of D01-4582 inplasma decreased in the order of homozygous, heterozygous, and wild-type beagles. Theequilibrium dissociation constants in the plasma from homozygous beagles wereapproximately sixfold greater than those for wild-type beagles. Prospective pharmaco-kinetic study using genotyped beagles demonstrated that the clearance was related toalbumin genotype. The variability in clearance was well explained by changes in theunbound fractions. These results suggest that the interindividual variability in thecanine pharmacokinetics of D01-4582 is related to an alteration in the binding affinitywhich is associated with albumin genetic polymorphisms. � 2008 Wiley-Liss, Inc. and the

American Pharmacists Association J Pharm Sci 98:1545–1555, 2009

Keywords: albumin; genetic polymorp

hisms; haplotype; pharmacokinetics; proteinbinding; single nucleotide polymorphism (SNP); pharmacogenetics

s used: D01-4582, 4-[1-[3-Chloro-4-[N0-(2-eido]phenylacetyl]-(4S)-fluoro-(2S)-pyrrolidine-2-oic acid; CL, total plasma clearance; Vdss,e of distribution at steady state; T1/2, half-life;dissociation constant; nPt, concentrations ofunbound fraction in plasma.ce to: Takashi Ito (Telephone: 81-3-3492-3131;8567;shi.uh@daiichisankyo.co.jp)

aceutical Sciences, Vol. 98, 1545–1555 (2009)

, Inc. and the American Pharmacists Association

JOURNAL OF

INTRODUCTION

Interindividual variability in pharmacokineticsand pharmacodynamics is a great concern duringdrug discovery and development and must beconsidered in order to reduce undesirable sideeffects and improve therapy. This kind of varia-bility is generally caused by polymorphic drugmetabolizing enzymes, transporters, or pharma-

PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009 1545

1546 ITO ET AL.

cologic target proteins.1–3 During the preclinicalstage, drug candidates are assessed from theviewpoint of variability, as well as other char-acteristics, and candidates whose pharmacologicresponse is considered to be influenced by geneticmilieus are screened out, which sometimes leadsto a no-go decision for the project.

a4b1 integrin, which is expressed on mostleukocytes, plays an important role in the processof the adhesion, migration, and activation ofinflammatory leukocytes at sites of inflammationby interacting with various adhesion molecules.4,5

The inhibition of a4b1 integrin is believed to be apromising approach for the treatment of multiplesclerosis, Crohn’s disease, and rheumatoid arthri-tis.6,7 D01-4582 (Fig. 1) is an a4b1 integrinantagonist synthesized in our institute. Sincethe site of action is in the blood, the desiredpharmacokinetics is anticipated to obtain anefficient pharmacological effect. However, markedrat strain differences in the D01-4582 pharmaco-kinetics were observed in preclinical pharmaco-kinetic studies. The total clearances of D01-4582in Lewis rats, Eisai hyperbilirubinemic rats, andSD rats from Charles River were approximatelyone-sixth of those for Wistar rats, Brown Norwayrats, and SD rats from SLC Japan. Theseobservations were considered to imply interindi-vidual pharmacokinetic variability in humans.Therefore, the mechanisms of the strain differ-ences were intensively investigated to evaluatethe probability of human interindividual varia-bility. As a result, differences in plasma proteinbinding were found to be involved. The threestrains with low clearance showed high plasmaprotein binding, whereas the three strains withhigh clearance showed low plasma protein bind-ing. These differences were due to albumin genetic

Figure 1. Chemical structure of D01-4582 (A) andD82-7319, an analytical internal standard (B).

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

polymorphisms, since all three strains with highclearance have point mutations that are asso-ciated with two amino acid changes, V238L andT293I. These studies are the first that report theeffect of albumin polymorphism on the pharma-cokinetics of a given compound.8 Based on theuncovered mechanism, it was considered thatthe interindividual variability in humans wouldbe minimal because the frequency of albuminvariants in humans is low and the exposure to theunbound drug, which is the pharmacologicallyand toxicologically active form, is not affected bythe change in protein binding in most cases.8,9

In preclinical studies, D01-4582 also showedinterindividual pharmacokinetic variability inbeagles. If the causative mechanism of canineinterindividual variability was identical to that ofrat strain differences, it would be unlikely toobserve interindividual variability in humans,based on the considerations mentioned pre-viously. In contrast, if a distinct mechanism isinvolved, further investigation should be under-taken to assess the possibility of whether themechanism is shared with humans and leads topharmacokinetic variability.

In this report, the involvement of caninealbumin genetic polymorphism in the interindi-vidual pharmacokinetic variability of D01-4582pharmacokinetics in beagles was investigated.Genetic polymorphisms of canine albumin werefound, and the prevalence of a mutant allelewas studied. The effect of the mutations onthe binding affinity to D01-4582 was evaluated.The effect of this albumin genetic polymorphismwas confirmed prospectively in a pharmacokinetic-pharmacogenetic study.

MATERIALS AND METHODS

Chemicals

D01-4582 and 4-[1-[4-[N0-(2-methylphenyl)urei-do]phenylacetyl]pyrrolidin-2-yl-methoxy]benzoicacid, D82-7319, which is an internal standard forthe analysis of D01-4582, were synthesized bythe Medicinal Chemistry Research Laboratory,Daiichi Sankyo (Tokyo, Japan) (Fig. 1). Thesecompounds were obtained through collaborationbetween Daiichi Sankyo and Pharmacopeia(Cranbury, NJ). All other reagents were ofanalytical grade.

DOI 10.1002/jps

CANINE PK-PG WITH ALBUMIN POLYMORPHISMS 1547

Animals

In total, 53 beagles were used in the present study(Tab. 1). Eight beagles were purchased fromNosan Corporation (Yokohama, Japan), 13 bea-gles were from LSG Corporation (Tokyo, Japan),22 beagles were from Toyota Tsusho Corporation(Nagoya, Japan), and 10 beagles were fromCeleste Corporation (Tokyo, Japan). Food andwater were provided ad libitum, with the excep-tion that food was withheld overnight before thestudies. The dogs were housed in separate cages inan animal room with a temperature of 23� 28Cand a relative humidity of 55� 15%. Lighting wascontrolled to provide a 12-h light-dark cycle. Allstudies were approved by the Animal EthicsCommittee of Daiichi Sankyo.

Protein Binding Study

Blood was taken from the beagles (n¼ 35) with0.1% (w/v) EDTA as an anti-coagulant. Plasmawas prepared by centrifuging the blood at 1500gfor 15 min at 48C. After a 3-min warming at 378C,10 mL of the D01-4582 standard solution wasadded to 490 mL of plasma and incubated for30 min at 378C (n¼ 3). The final concentrationsof D01-4582 were 92.6 mM. At the end of theincubation period, 210 mL of each sample wascentrifuged in a 0.23 PC tube (Hitachi Koki Co.,Ltd., Tokyo, Japan) at 200000g for 16 h at 48C.The supernatant (50 mL) was collected andsubjected to LC-MS/MS analysis (Quattro Ultima,Waters, Milford, MA). For the determinationof the binding parameters, plasma preparedindividually from genotyped beagles was used(n¼ 3/genotype). The final incubation volume was

Table 1. Distribution of Dogs Used in the Present Study

Determinationof PartitionCoefficient

ProteinBindingStudy

PK/PGStudy

NotGenotyped

U U

U

U

U

U U

Sum¼ 6 35 9 6

DOI 10.1002/jps JO

220 mL (duplicated) and the final D01-4582concentrations were 5, 10, 20, 30, 50, 75, 100,and 150 mM. The albumin concentration wasdetermined for the three genotype dogs using thebromcresol green method10 by monitoring theabsorbance at 630 nm (VERSAmax, MolecularDevices, Sunnyvale, CA). Human serum albumin(Sigma–Aldrich, St. Louis, MO) was used as areference protein.

Determination of Partition Coefficient

The partition coefficient (RB) of D01-4582 betweenwhole blood and plasma was evaluated usingwhole blood prepared from 6 beagles. After pre-incubation of 495 mL of blood at 378C, 5 mL of theD01-4582 standard solution was added to theblood solution (200 ng/mL, n¼ 3). The bloodsamples were incubated for 15 min at 378C, thencentrifuged at 1500g for 15 min, and the plasmaconcentration was determined.

Determination of Canine Albumin cDNA Sequencesfrom Liver Samples

The cDNA sequences of the albumins weredetermined from the liver samples (n¼ 26). Aliver sample (ca. 0.2 g) was removed from each dogand then the dogs were euthanized. For eachsample, RNeasy (QIAGEN, Hilden, Germany)was used to extract the total RNA from the liversamples, followed by the purification of poly Aþ

RNA using an Oligotex-dT30 (Takara Bio, Inc.,Shiga, Japan). For each poly Aþ RNA sample,an LA PCR Kit (AMV) Ver. 1.1 (Takara Bio, Inc.)

Genotyping Study

nLiver

SampleBlood

SampleLiver and

Blood Sample

6U 8

U 4U 12

U 8U 6

U 920 21 6 53

URNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

1548 ITO ET AL.

with an adaptor primer was used to synthesizefirst-strand cDNA by reverse transcription. Theconditions were 308C for 10 min, 508C for 30 min,998C for 5 min, and 58C for 5 min. To analyze thebase sequence of the albumin cDNA, a polymerasechain reaction (PCR) was performed with specificprimer sets by GenAmp (Applied Biosystems,Drive Foster, CA). The following primer set wasdesigned according to the reference sequenceregistered in GenBank (Accession No. AB090854):50-GACGAGAAGCATATAAGAGTG-30 (nt 68–88)and 50-GCTGAGATGTTTGTTATTGTG-30 (nt 1853–1833). The PCR reaction included 0.2 mM dNTPmixture, 1�PCR buffer, 2.5 mM MgSO4, 1 mMforward primer, 1mM reverse primer, 10mL cDNAsample, and 1 U KOD-Plus (Toyobo, Osaka,Japan). Distilled water was added to give a totalvolume of 50 mL. The PCR conditions were:denaturation at 948C for 15 sec (first cycle,2 min), annealing at 468C for 30 sec, and extensionat 688C for 2 min over a total of 30 cycles. Afterthe residual primer and dNTP were digested witha PCR Product Pre-Sequencing Reagent Pack(Toyobo), a GFX PCR DNA Gel Band PurificationKit (GE Healthcare Bio-Sciences Corp, Piscat-away, NJ) was used to purify the PCR products. ABigDye terminator v1.1 cycle Sequencing Kit anda v3.1 cycle Sequencing Kit (Applied Biosystems)were used to sequence the PCR products. TheDNA sequence was determined with a DNAsequencer, Prism 3100 Avant (Applied Biosys-tems). The presence of mutations was determinedby comparing the sequence result with theGenBank reference sequence. The nucleotidesequence was numbered from the first atg thatappears at the 35th nucleotide from the beginningof the registered sequence. The amino acidsequence was determined from glutamic acid atthe 25th amino acid, since albumin is synthesizedas preproalbumin, which has a signal peptide atits N-terminal that consists of 24 amino acids forintracellular trafficking and secretion to blood.

Determination of Canine Albumin cDNA Sequencesfrom Blood Samples

The albumin genotype at 2 mutation sites,G1075T and A1422T, was determined usinggenomic DNA isolated from whole blood samples(n¼ 21). A QIAamp DNA Blood Midi Kit (QIA-GEN) was used to extract genomic DNA from the2-mL whole blood samples. PCR was performed onthe separate DNA samples with specific primer

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

sets using GenAmp (Applied Biosystems). Thefollowing primer set was designed according tothe reference sequence registered in GenBank(Accession No. AB090854): 50-GAGGCAAAGGAT-GTGTTCCTG-30 and 50-GCATGTAGGAGGATC-ATCGGT-30 for the G1075T site, and 50-ACT-CTCGTGGAGGTCTCAAG-30 and 50-TTTGGT-AACTCTCTCGCTCA-30 for the A1422T site.The PCR reaction mixture included 0.2 mM dNTPmixture, 1�PCR buffer, 1 mM MgSO4, 0.3 mMforward primer, 0.3 mM reverse primer, 250 ng ofgenomic DNA, and 1 U KOD-Plus (Toyobo) for theG1075T site, and 0.2 mM dNTP mixture, 1�PCRbuffer, 1 mM MgSO4, 0.15 mM forward primer,0.15 mM reverse primer, 250 ng of genomic DNA,and 1 U KOD-Plus (Toyobo) for the A1422T site.Distilled water was added to yield a total volumeof 50 mL. The PCR conditions were: denaturationat 948C for 15 sec (first cycle, 2 min), annealing at578C for 30 sec, and extension at 688C for 1.1 minover a total of 40 cycles for PCR of G1075T site,and denaturation at 948C for 15 sec (first cycle,2 min), annealing at 528C for 30 sec, and extensionat 688C for 50 sec over a total of 38 cycles for PCRof A1422T site. The procedures following PCRwere the same as those described in the liversample PCR section. The obtained PCR productwas about 1.5 kbp for the G1075T site, and about800 bp for the A1422T site. For DNA sequencing,the following primers were used: 50-TTCATTCT-GAACTGCAGCC-30 for the G1075T site and 50-GTTCAGGATACTATAGATGACCAAG-30 for theA1422T site. The presence of mutations wasdetermined by comparing the sequence resultwith the reference sequence. It was confirmed inour laboratory that this method gave the sameresults as those obtained from the liver samples(n¼ 6).

Pharmacokinetic/Pharmacogenetic Study

Dogs whose albumin gene had been genotypedwere enrolled in the study. The genotype that wasidentical to the reference sequence was desig-nated as wild-type. The dogs were assigned to 1 of3 groups: wild-type, heterozygote, and homozy-gote, based on the albumin genotyping results.Each group consisted of three dogs. D01-4582 wasdissolved in saline with 3 eq. mol NaOH and wasadministered intravenously (1 mg/0.5 mL/kg) tothe dogs through a forelimb vein. Blood sampleswere collected directly from the forelimb vein withheparinized syringes at 0.083, 0.5, 1, 2, 4, 6, 8 h.

DOI 10.1002/jps

CANINE PK-PG WITH ALBUMIN POLYMORPHISMS 1549

Plasma was prepared by centrifuging the blood at1500g for 15 min at 48C. The protein binding ratioof D01-4582 in each dog had been assessed in theprevious protein binding study.

Analytical Procedures

Sample Preparation for LC-MS/MS Analysis

Samples with high concentrations were dilutedwith blank matrix. Methanol (150 mL) containingthe internal standard was added to the plasmasamples (50 mL) followed by vigorous mixing andcentrifugation at 10000g for 15 min at 48C.Supernatant (20 mL) was introduced into theLC-MS/MS. In the protein binding study, metha-nol (50 mL) containing the internal standard wasadded to the supernatant sample and then themixture (20 mL) was introduced into the LC-MS/MS. Samples for the calibration curves werefreshly prepared by adding several concentrationsof D01-4582 standard solution to blank matrixsamples.

LC-MS/MS Analysis

A Waters HPLC system (Aliance 2695, Waters,Milford, MA) coupled to a Quattro Ultima(Waters) via an electrospray ionization interfacewas used for mass analysis and detection.Chromatographic separation was achieved on aSymmetry Shield RP8 column (50� 2.1 mm i.d.;Waters) with gradient elution using an 10 mMammonium formate buffer and methanol as themobile phase at a flow rate of 0.2 mL/min. Thecolumn temperature was 408C and the total runtime was 12 min for each sample. The massspectrometer was operated in positive ion mode.The MS–MS transition selected to monitor D01-4582 was from m/z 540.2 to a product ion at m/z401.6. The internal standard was monitored usingthe transition from m/z 488.2 to m/z 350.2.Each transition was alternately monitored at adwell time of 0.3 sec. Calibration curves wereobtained with a weighting of 1/x. The coefficientof determination was >0.998. The limit ofquantification was 9.26 nM for plasma samplesand 1.85 nM for supernatant samples. The upperlimit of quantification was 3700 nM for the plasmasamples and 1850 nM for the supernatantsamples.

DOI 10.1002/jps JO

Data Analysis

Pharmacokinetics

The total plasma clearance (CL), apparent volumeof distribution (Vdss), and half-life (T1/2) werecalculated by the non-compartmental methodwithout extrapolation using a program developedin-house. This program is written in Visual Basicfor Application and runs on Microsoft Excel. It isvalidated to output the same results as WinNonlinversion 3.1. The harmonized mean was calculatedfor T1/2. RB was estimated by dividing thenominal blood concentration by the plasmaconcentration of D01-4582.

The relationship between CL and the unboundfraction in the pharmacokinetic/pharmacogeneticstudy was analyzed based on a well-stirred model.11

Since D01-4582 is not eliminated from the kidneybut is excreted into bile in rats as either theunchanged form or as a metabolite8, D01-4582 isassumed to be predominantly eliminated from theliver so that the total clearance is accounted for bythe hepatic clearance, as follows:

CL ¼ QHRBfuCLint

QHRB þ fuCLint(1)

where QH represents the hepatic blood flow rate(30.9 mL/min/kg),12 CLint represents the intrinsicclearance, and fu represents the unbound fractionof D01-4582 in plasma. A nonlinear least-squareprogram (WinNonlin from Pharsight, MountainView, CA) was used to fit the model. The co-efficient of determination (r2) was calculated toevaluate the performance of the model.

To evaluate the predictive performance of thewell-stirred model, the predicted CL for eachindividual dog was calculated by Eq (1) usingCLint estimated by a fitting analysis and theindividual fu. The predictability was expressed asthe mean absolute error (MAE) as follows:

MAE ¼X predicted CL � observed CLj j

n(2)

Binding Analysis

The binding characteristics were examinedinitially by a Scatchard plot to determine theapplicability of only one principle class of bindingsite. The equilibrium dissociation constant (Kd)and concentration of the binding site (nPt) wereestimated by fitting the results to the followingequation. A nonlinear least-square program(WinNonlin) was used to fit the following

URNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

Figure 2. Unbound fraction of D01-4582 in beagles inrelation to the albumin gene genotype. Symbols repre-sent individual data. Horizontal lines represent themean of each group and vertical lines represent the95% confidence intervals. A wild-type allele that isidentical to the reference sequence AB090854 is desig-

1550 ITO ET AL.

equation:

Cb ¼ nPtCu

Kd þ Cu; (3)

where Cb and Cu represent the molar concentra-tions of bound and unbound D01-4582, respectively.The unbound fraction of D01-4582 was estimatedby dividing the unbound concentration of D01-4582by the nominal incubation concentration.

Statistics

A one-way analysis of variance (ANOVA) with aDunnett post test and a Student’s t-test wasperformed using a statistical program (EXSASver. 7.10 based on SAS release 8.2, ArmSYSTEXCorp., Osaka, Japan & SAS Institute Japan Ltd.,Tokyo, Japan). The allele frequencies wereestimated by gene counting. Departure from theHardy–Weinberg equilibrium was tested by a x2

test with 1 degree of freedom (df).

nated as H1, whereas a mutant allele is designated asH2. H1/H1: 0.015� 0.003% (CV¼ 20%, n¼ 10), H1/H2:0.028� 0.011% (CV¼ 39%, n¼ 20), H2/H2: 0.059� 0.016%(CV¼ 27%, n¼ 5), All: 0.029� 0.017% (CV¼ 59%,n¼ 35) (Mean�SD), �Significantly different (p< 0.05)from H1/H1 group by a Dunnett’s test.

RESULTS

Protein Binding Study and Determination of thePartition Coefficient

The protein binding was analyzed for 35 dogs(Fig. 2). The unbound fraction of D01-4582 rangedfrom 0.010% to 0.081%. RB for D01-4582 was0.55� 0.04 (Mean�SD, n¼ 6).

Table 2. Summary of Genotyping Results in EachAlbumin Genotype

Genetic Polymorphism

G1075T(Ala335Ser)

A1422T(Glu450Asp)

Number ofObservation

H1/H1 G/G A/A 14H1/H2 G/T A/T 28H2/H2 T/T T/T 5All 47

A wild-type allele that is identical to the reference sequenceAB090854 is designated as H1, whereas a mutant allele isdesignated as H2.

Determination of Dog Albumin Genotype

The cDNA sequences of the albumin genes weredetermined for 26 dogs from liver samples. Twopoint mutations, G1075T and A1422T, wereidentified and none of the other mutations,including synonymous mutations, were found.Both mutations caused amino acid changes.G1075T changes Ala335 to Ser, and A1422Tchanges Glu450 to Asp. An additional 21 dogswere genotyped from whole blood samples focus-ing on the two mutations (Tab. 2). The two pointmutations observed in the present study werecompletely linked with each other. Assuming thatthe two point mutations are on the identical allele,no significant deviation from the Hardy–Wein-berg equilibrium was observed for the polymorph-ism (p> 0.05). The estimated prevalence of themutant allele was 40%. Hereafter, the wild-typeallele that is identical to the reference sequence

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

(AB090854) is referred to as H1, whereas themutant allele is referred to as H2.

Relationship between Protein Binding andAlbumin Genotype

Large interindividual variability (CV¼ 59%) wasobserved in the unbound fraction of D01-4582 forthe total population (n¼ 35) (Fig. 2). However,

DOI 10.1002/jps

Table 3. Binding Parameters of D01-4582 in Plasma from Dogs Carrying Wild-Type or Mutant Albumin Genes

Genotype Number of Dogs

Kd nPt

Meana (nM) SD (nM) Meana (mM) SD (mM)

H1/H1 3 15.2b 1.4 258c 14H2/H2 3 88.8b 8.1 291c 17

aStatistical significance was analyzed by a Student’s t-test.bSignificantly different (p<0.05).cNot significant (p> 0.05).

CANINE PK-PG WITH ALBUMIN POLYMORPHISMS 1551

when analyzed by genotype, the variability of allthe genotypic classes diminishes considerably(Fig. 2). The unbound fractions were significantlyhigher for plasma prepared from both individualsof the H2/H2 mutant genotype and the H1/H2genotype than for plasma prepared from indivi-duals of the H1/H1 wild-type genotype. Plasmasamples prepared from beagles of the H1/H2genotype exhibit intermediate values for theunbound fraction. To investigate the effect ofthe mutant allele on the binding characteristics ofD01-4582, the Kd and nPt were determined usingplasma prepared from genotyped dogs (Tab. 3).The Kd was approximately sixfold greater forplasma prepared from individuals of the H2/H2genotype than that for plasma prepared fromindividuals of the H1/H1 genotype. In contrast, nosignificant difference was observed for the nPtbetween the two genotypes. The albumin concen-trations were 4.21� 0.59 (n¼ 5), 4.19� 0.46(n¼ 5), and 4.14� 0.21 mg/mL (n¼ 3)(Mean�SD), which are not significantly differentfrom those by a Dunnett’s test.

Figure 3. Plasma concentration-time curves ofD01-4582 in genotyped beagles after intravenousadministration at a dose of 1 mg/kg. Circles representH1/H1; triangles, H1/H2; and squares, H2/H2. Theconcentrations of H2/H2 at 8 h after administrationwere under the low limit of quantification.

Pharmacokinetic/Pharmacogenetic Study

To investigate the relationship between thealbumin genotype and the D01-4582 pharmaco-kinetics in dogs, the plasma concentration profileswere assessed prospectively in genotyped dogs.The mean plasma concentration during themonitoring period was high in H1/H1 dogs andlow in H2/H2 dogs (Fig. 3). The CL wassignificantly higher for both H1/H2 and H2/H2than for H1/H1 (p< 0.05) (Tab. 4). The albumingenotype explained 72% of the variability in CL.The coefficient of variance of CL in each genotypegroup was lower than that in the total population(Tab. 4). The differences in Vdss and T1/2 were notstatistically significant among the three groups.The unbound fraction of D01-4582 was signifi-

DOI 10.1002/jps JO

cantly higher for H2/H2 than that for H1/H1. Thetotal clearance is plotted against the unboundfraction in Figure 4. The intrinsic clearanceestimated by a fitting analysis was 15000� 1200mL/min/kg (Mean�SE). This model explained88% of the variability in CL. When the well-stirredmodel was applied, MAE was 0.87 (n¼ 9), whereasMAE was 3.11 (n¼ 9) when the mean of theobserved CL (3.66 mL/min/kg) was used as thepredicted CL.

DISCUSSION

This study was performed to uncover the mechan-ism behind the interindividual pharmacokineticvariability of D01-4582 in beagles. In a precedingstudy, it was found that D01-4582 exhibits strain

URNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

Table 4. Pharmacokinetics of D01-4582 in Genotyped Beagles at a Dose of 1 mg/kg and the Unbound Fraction at aConcentration of 92.6 mM

GenotypeCL

(mL/min/kg)Mean (CV%)(mL/min/kg)

Vdss(L/kg)

Mean(CV%) (L/kg)

T1/2(h)

Mean(h)

UnboundFraction (%)

Mean(SD) (%)

H1/H1 2.16 0.150 1.4 0.0131.93 1.97 (9) 0.0992 0.118 (23) 0.8 1.2 0.012 0.014 (0.002)1.83 0.106 1.6 0.016

H1/H2 5.41 0.327 1.3 0.0503.97 4.07a (32) 0.118 0.179b (72) 0.8 0.8b 0.019 0.030b (0.018)2.83 0.0914 0.7 0.020

H2/H2 4.35 0.134 0.6 0.0454.34 4.95a (21) 0.165 0.135b (22) 0.7 0.6b 0.046 0.053a (0.014)6.15 0.105 0.5 0.069

Overall 3.66 (43) 0.144 (51)

aSignificantly different (p<0.05) from the H1/H1 group by a Dunnett’s test.bNot significant (p>0.05) from the H1/H1 group by a Dunnett’s test.

1552 ITO ET AL.

differences in its pharmacokinetics in rats. Thiseffect is thought to be caused by an albuminpolymorphism that alters the D01-4582 bindingaffinity.8 Therefore, the hypothesis that thedifference in protein binding of D01-4582 mightbe also associated with the interindividual varia-bility in canine pharmacokinetics was examined.

The relationship between the genotype andprotein binding was examined. A total of 35 dogswas both genotyped and phenotyped with respectto the albumin gene and the protein binding ofD01-4582. There was a clear relationship betweenthe genotype and protein binding (Fig. 2). The H1and H2 allele seemed to be responsible for high

Figure 4. Relationship between CL and unboundfraction of D01-4582 in beagles. Circles represent H1/H1; triangles, H1/H2; and squares, H2/H2. A solid linerepresents the fitted line of a well-stirred model to thedata.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

protein binding and low protein binding, respec-tively. This was strongly supported by the resultsshowing that the binding affinity of D01-4582 washigher in plasma prepared from the H1/H1genotype dogs than in plasma prepared fromthe H2/H2 genotype dogs (Tab. 3). This geneticpolymorphism would not have an effect on thebinding capacity because the nPt and albuminconcentrations were not changed (Tab. 3). Allthese results suggested that the variability inprotein binding was caused by albumin poly-morphisms that alter the binding affinity.

From the results of the prospective pharmaco-kinetic/pharmacogenetic study, where genotypeddogs were used, it was demonstrated that thegenotype of the albumin had a strong associationwith the pharmacokinetic variability of D01-4582(Fig. 3, Tab. 4). The pharmacokinetic variabilitywas considered to be related to the variability inCL. The relationship between CL and fu of D01-4582 was well described by the well-stirred model(Fig. 4). Although the disposition route of D01-4582 was not thoroughly investigated and otherpossible mechanisms were not considered at thistime, the above analysis strongly suggested thatthe interindividual pharmacokinetic variabilitywas not due to the variability in metabolicactivity, but was due to albumin polymorphism.

Since 72% of the variance in CL was accountedfor by the albumin genotype, the interindividualvariability in CL was decreased when thegenotype was taken into account (Tab. 4). CLcould be predicted more accurately because theMAE was improved when a well-stirred modelwas applied. In contrast to the effect of thealbumin genotype on CL, the effect of the genotype

DOI 10.1002/jps

CANINE PK-PG WITH ALBUMIN POLYMORPHISMS 1553

on Vdss was not significant. This would be partlybecause D01-4582 is a compound with a smalldistribution volume, so that its distributionvolume is less sensitive to the change in proteinbinding. As a result, T1/2 for the heterozygote andhomozygote groups was relatively shorter thanthat for the wild-type group. Consequently,whereas the predictability of the total clearancecould be much improved by taking the albumingenotype or the variability in the unboundfraction into consideration, the predictability ofthe volume of distribution could not be improved.

Recently, canine genetic polymorphisms thataffect pharmacokinetics of drugs and drug candi-dates have been reported. For instance, CYP2D15and CYP1A2 have been demonstrated to bepolymorphic.13–18 Additionally, a certain popula-tion of collies has been reported to be deficient formdr1.19–21 These findings, as well as results of thepresent study, suggest that laboratory strains ofdogs are so heterozygous that the pharmacoki-netics in dogs could be varied among individuals bygenetic polymorphisms, for example, drug meta-bolizing enzymes and transporters.22 Albumingenetic polymorphism should be one of determi-nants to be considered. In the present study,neither the metabolism of D01-4582 in dogs nor thegenotype of drug metabolizing enzymes in the dogsemployed has been investigated yet, but albumingenetic polymorphism would predominantly gov-ern the pharmacokinetic variability of D01-4582.

The mutations observed in the present studyare two of six mutations registered in the Swiss-Prot database (http://ca.expasy.org/uniprot/p49822)where the mutation sites are numbered as pre-proalbumin that contains a signal peptide at theN-terminal. Interestingly, the two mutationsobserved in the present study were completelylinked with each other, suggesting the existence oftwo distinct alleles, H1, the wild-type allele andH2, the mutant allele (Tab. 2). Although thenumber of individuals was too small to estimatethe allele frequency precisely, the estimatedprevalence of the mutant allele was 40%(n¼ 94), suggesting that about 16% of the beagleswere homozygous mutants and that about 36% ofthe beagles were wild-type, assuming the Hardy–Weinberg equilibrium.

The mutant allele is associated with a reductionof albumin’s binding affinity for D01-4582 in beagles.In HSA, Ala335 is located in subdomain IIB andis a constituent of the eighth a-helix. Glu450 islocated in subdomain IIIA and is a constituent ofthe fourth a-helix. Since albumin sequences are

DOI 10.1002/jps JO

highly conserved among species, Ala335 andGlu450 in dog albumin are thought to composethe same construct of human albumin. Glu450 ishighly conserved among species and is locatednear binding site II in HSA, which is known as thediazepam site.23 This amino acid has beenreported to be involved directly in the binding ofligands to HSA.24,25 In contrast, Ala335 is locatedfar from both binding site I, known as the warfarinsite, and site II. This position is thought not to beinvolved in drug binding. Although it is notcertain that identical amino acid residues areinvolved in ligand binding to both human and dogalbumin, the Glu450!Asp change in dog albuminwould alter the binding affinity.

Similar to the case of rats, the pharmacokineticvariability of D01-4582 in dogs is related to analteration in the binding affinity which is asso-ciated with albumin genetic polymorphisms.Amino acid changes lead to an increase in theequilibrium dissociation constant in both rats anddogs. However, the mutation sites are different inrats and dogs. The mutation sites that affect D01-4582 binding in rats are Val238 and Thr293,which are located near binding site I.8 It might bepossible that D01-4582 binds different bindingsites in rats and dogs, though the site at whichD01-4582 binds rat and dog albumin remainsunclear. Kosa et al.26 reported that the bindingcharacteristic of site I in dog albumin is notsimilar to those of human albumin or rat albumin,suggesting the possibility that a particular ligandthat binds to site I in human and rat albuminbinds at a different site in dog albumin other thansite I. To clarify the effect of the polymorphisms onthe albumin binding affinity for D01-4582, acomprehensive study to reveal the binding siteof D01-4582 is needed.

The results of the present study provided ourdrug discovery program with valuable informa-tion. First, it was demonstrated that interindivi-dual pharmacokinetic differences can be causedby albumin genetic polymorphisms as well aspolymorphic drug metabolizing enzymes. Differ-ences in the mechanism of variability can affectthe preclinical study design. In a case wherevariability results from albumin polymorphism,the study design would not have to be optimizedbecause for most drugs the systemic exposure tothe parent drug is independent of changes inprotein binding.9 In contrast, if variability resultsfrom the polymorphism of a given drug metaboliz-ing enzyme, preclinical study designs need tobe optimized because exposure to the unbound

URNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

1554 ITO ET AL.

parent drug would be different between the twophenotypes, that is, the extensive metabolizersand poor metabolizers.27 Second, it becomespossible to evaluate total clearance precisely fordrug candidates, chemical derivatives of D01-4582, most of which also showed interindividualpharmacokinetic differences similar to D01-4582,because pharmacokinetic data could be obtainedwith small variability by assigning dogs based onthe relevant genotype (Tab. 4). If required,intrinsic clearance itself could be estimated bydetail analysis (Fig. 4). Although it is not commonto conduct preclinical pharmacokinetic studiesusing genotyped dogs, it would be worthwhile toconduct such studies using dogs with a knowngenetic background so that we can obtain datawith minimal pharmacokinetic variability andestimate the intrinsic clearance more precisely.There is one final point: a reasonable go-decisionwas made for this program because the inter-individual variability in beagles is not consideredto be of clinical relevance, since the frequency ofhuman albumin polymorphisms is low and theexposure of unbound drugs is not affected by achange in protein binding. Collectively, it isessential to understand the mechanism of phar-macokinetic variability during drug discoveryand the development process in order to designappropriate preclinical pharmacokinetic and tox-icologic studies and to make a reasonable decisionby interpreting the results correctly.

Although the other possible mechanisms werenot fully investigated at this time, the alteration inthe binding affinity caused by albumin geneticpolymorphisms was demonstrated to be associatedwith the interindividual variability in the caninepharmacokinetics of D01-4582. This is the firstreport demonstrating the effect of canine albuminpolymorphism on the pharmacokinetics of a givencompound. The existence of a mutant allele thatpossesses two amino acid changes, from Ala335 toSer and from Glu450 to Asp, was revealed. It wouldbe important to conduct preclinical studies with athorough understanding of the mechanisms of anyunusual experimental results observed in precli-nical studies in order to make valid decisionsduring drug discovery and development.

REFERENCES

1. Roden DM, George AL, Jr. 2002. The genetic basisof variability in drug responses. Nat Rev DrugDiscov 1:37–44.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009

2. Thomas FJ, McLeod HL, Watters JW. 2004. Phar-macogenomics: The influence of genomic variationon drug response. Curr Top Med Chem 4:1399–1409.

3. Shah RR. 2005. Pharmacogenetics in drug regula-tion: Promise, potential and pitfalls. Philos Trans RSoc Lond B Biol Sci 360:1617–1638.

4. Elices MJ, Osborn L, Takada Y, Crouse C, Luhows-kyj S, Hemler ME, Lobb RR. 1990. VCAM-1 onactivated endothelium interacts with the leukocyteintegrin VLA-4 at a site distinct from the VLA-4/fibronectin binding site. Cell 60:577–584.

5. Cunningham SA, Rodriguez JM, Arrate MP,Tran TM, Brock TA. 2002. JAM2 interacts withalpha4beta1. Facilitation by JAM3. J Biol Chem277:27589–27592.

6. Miller DH, Khan OA, Sheremata WA, BlumhardtLD, Rice GP, Libonati MA, Willmer-Hulme AJ,Dalton CM, Miszkiel KA, O’Connor PW. 2003.A controlled trial of natalizumab for relapsing mul-tiple sclerosis. N Engl J Med 348:15–23.

7. Ghosh S, Goldin E, Gordon FH, Malchow HA,Rask-Madsen J, Rutgeerts P, Vyhnalek P, ZadorovaZ, Palmer T, Donoghue S. 2003. Natalizumabfor active Crohn’s disease. N Engl J Med 348:24–32.

8. Ito T, Takahashi M, Sudo K, Sugiyama Y. 2007.Marked strain differences in the pharmacokineticsof an a4b1 integrin antagonist, 4-[1-[3-chloro-4-[N-(2-methylphenyl)-ureido]phenylacetyl]-(4S)-fluoro-(2S)-pyrrolidine-2-yl]-methoxybenzoic Acid (D01-4582), inSprague-Dawley rats are associated with albumingenetic polymorphism. J Pharmacol Exp Ther 320:124–132.

9. Benet LZ, Hoener BA. 2002. Changes in plasmaprotein binding have little clinical relevance. ClinPharmacol Ther 71:115–121.

10. Doumas BT, Watson WA, Biggs HG. 1971. Albuminstandards and the measurement of serum albuminwith bromcresol green. Clin Chim Acta 31:87–96.

11. Pang KS, Rowland M. 1977. Hepatic clearance ofdrugs. I. Theoretical considerations of a ‘‘well-stir-red’’ model and a ‘‘parallel tube’’ model. Influence ofhepatic blood flow, plasma and blood cell binding,and the hepatocellular enzymatic activity on hepa-tic drug clearance. J Pharmacokinet Biopharm 5:625–653.

12. Davies B, Morris T. 1993. Physiological parametersin laboratory animals and humans. Pharm Res10:1093–1095.

13. Paulson SK, Engel L, Reitz B, Bolten S, Burton EG,Maziasz TJ, Yan B, Schoenhard GL. 1999. Evidencefor polymorphism in the canine metabolism of thecyclooxygenase 2 inhibitor, celecoxib. Drug MetabDispos 27:1133–1142.

14. Tenmizu D, Endo Y, Noguchi K, Kamimura H.2004. Identification of the novel canine CYP1A21117 C>T SNP causing protein deletion. Xenobio-tica 34:835–846.

DOI 10.1002/jps

CANINE PK-PG WITH ALBUMIN POLYMORPHISMS 1555

15. Tenmizu D, Noguchi K, Kamimura H, Ohtani H,Sawada Y. 2006. The canine CYP1A2 deficiencypolymorphism dramatically affects the pharmaco-kinetics of 4-cyclohexyl-1-ethyl-7-methylpyrido[2,3-D]-pyrimidine-2-(1H)-one (YM-64227), a phos-phodiesterase type 4 inhibitor. Drug Metab Dispos34:800–806.

16. Tenmizu D, Noguchi K, Kamimura H. 2006. Eluci-dation of the effects of the CYP1A2 deficiency poly-morphism in the metabolism of 4-cyclohexyl-1-ethyl-7-methylpyrido[2,3-d]pyrimidine-2-(1h)-one(YM-64227), a phosphodiesterase type 4 inhibitor,and its metabolites in dogs. Drug Metab Dispos 34:1811–1816.

17. Mise M, Yadera S, Matsuda M, Hashizume T, Mat-sumoto S, Terauchi Y, Fujii T. 2004. Polymorphicexpression of CYP1A2 leading to interindividualvariability in metabolism of a novel benzodiazepinereceptor partial inverse agonist in dogs. DrugMetab Dispos 32:240–245.

18. Mise M, Hashizume T, Matsumoto S, Terauchi Y,Fujii T. 2004. Identification of non-functional allelicvariant of CYP1A2 in dogs. Pharmacogenetics 14:769–773.

19. Mealey KL, Bentjen SA, Gay JM, Cantor GH. 2001.Ivermectin sensitivity in collies is associated witha deletion mutation of the mdr1 gene. Pharmaco-genetics 11:727–733.

20. Mealey KL, Northrup NC, Bentjen SA. 2003.Increased toxicity of P-glycoprotein-substrate che-motherapeutic agents in a dog with the MDR1

DOI 10.1002/jps JO

deletion mutation associated with ivermectin sen-sitivity. J Am Vet Med Assoc 223:1453–1455.

21. Neff MW, Robertson KR, Wong AK, Safra N, Bro-man KW, Slatkin M, Mealey KL, Pedersen NC.2004. Breed distribution and history of caninemdr1-1Delta, a pharmacogenetic mutation thatmarks the emergence of breeds from the collie line-age. Proc Natl Acad Sci USA 101:11725–11730.

22. Kamimura H. 2006. Genetic polymorphism of cyto-chrome P450s in beagles: Possible influence ofCYP1A2 deficiency on toxicological evaluations.Arch Toxicol 80:732–738.

23. Peters JT. 1996. All about albumin. San Diego:Academic Press.

24. Curry S, Mandelkow H, Brick P, Franks N. 1998.Crystal structure of human serum albumin com-plexed with fatty acid reveals an asymmetric dis-tribution of binding sites. Nat Struct Biol 5:827–835.

25. Ghuman J, Zunszain PA, Petitpas I, BhattacharyaAA, Otagiri M, Curry S. 2005. Structural basis ofthe drug-binding specificity of human serum albu-min. J Mol Biol 353:38–52.

26. Kosa T, Maruyama T, Otagiri M. 1997. Speciesdifferences of serum albumins: I. Drug bindingsites. Pharm Res 14:1607–1612.

27. Paulson SK, Maziasz TJ. 2004. Role of preclinicalmetabolism and pharmacokinetics in the develop-ment of celecoxib. In: Krishna R, editor. Applica-tions of pharmacokinetic principles in drugdevelopment. New York: Kluwer Academic/PlenumPublishes. pp. 405–425.

URNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 4, APRIL 2009