‘6ZKB]/\[I10S['8U11'10[1123.I0‘S[ELI.II’l0I18dSB'pIJ.Ip{I101}PSPEOIIIAAOG
`
`
`
`
`
`
`
`0090-9556/99/271l—l350—l359$02.00/0
`DRUG METABOLISM AND DISPOSITION
`Copyright © 1999 by The American Society for Pharmacology and Experimental Therapeutics
`
`Vol. 27, No. 11
`Printed in U.S.A.
`
`PREDICTION OF HUMAN CLEARANCE OF TWENTY-NINE DRUGS FROM HEPATIC
`
`MICROSOMAL INTRINSIC CLEARANCE DATA: AN EXAMINATION OF IN VITRO
`
`HALF-LIFE APPROACH AND NONSPECIFIC BINDING TO MICROSOMES
`
`R. SCOTT OBACH
`
`Drug Metabolism Department, Candidate Synthesis, Enhancement, and Evaluation, Central Research Division, Pfizer, Inc., Groton, Connecticut
`
`(Received April 9, 1999; accepted July 30, 1999)
`
`This paper is available online at http://www.dmd.org
`
`ABSTRACT:
`
`Twenty-nine drugs of disparate structures and physicochemical
`properties were used in an examination of the capability of human
`liver microsomal Iability data (“in vitro T,,2" approach) to be useful
`in the prediction of human clearance. Additionally, the potential
`importance of nonspecific binding to microsomes in the in vitro
`incubation milieu for the accurate prediction of human clearance
`was investigated. The compounds examined demonstrated a wide
`range of microsomal metabolic labilities with scaled intrinsic clear-
`ance values ranging from less than 0.5 ml/min/kg to 189 ml/min/kg.
`Microsomal binding was determined at microsomal protein con-
`centrations used in the Iability incubations. For the 29 compounds
`studied, unbound fractions in microsomes ranged from 0.11 to 1.0.
`Generally, basic compounds demonstrated the greatest extent of
`binding and neutral and acidic compounds the least extent of
`
`binding. In the projection of human clearance values, basic and
`neutral compounds were well predicted when all binding consid-
`erations (blood and microsome) were disregarded, however, in-
`cluding both binding considerations also yielded reasonable pre-
`dictions. Including only blood binding yielded very poor projections
`of human clearance for these two types of compounds. However,
`for acidic compounds, disregarding all binding considerations
`yielded poor predictions of human clearance. It was generally most
`difficult to accurately predict clearance for this class of com-
`pounds; however the accuracy was best when all binding consid-
`erations were included. Overall, inclusion of both blood and micro-
`some binding values gave the best agreement between in vivo
`clearance values and clearance values projected from in vitro
`intrinsic clearance data.
`
`The use of in vitro drug metabolism data in the understanding of in
`vivo pharmacokinetic data has recently become an area of scientific
`interest (Houston, 1994; Houston and Carlile, 1997; Iwatsubo et al.,
`1997). This has partially stemmed from a trend in the pharmaceutical
`industry to use in vitro drug metabolism data, using human—derived
`reagents, as a criterion to select compounds for further development
`(Rodrigues, 1997). Thus, in vitro metabolism data is used in a pro-
`spective manner to choose those compounds for further development
`that are expected to possess commercially acceptable pharmacokinetic
`properties (e.g., half-life permitting once-per—day administration reg-
`imens, low oral clearance to reduce dose, etc.). Several investigators
`have recently described methods whereby preclinical drug metabo-
`lism and pharmacokinetic data can be used to predict human pharma-
`cokinetic parameters (Obach et al., 1997; Lave et al., 1997a,b; Mah-
`mood, l998a,b).
`The first demonstration of the correlation between in vivo clearance
`values and clearance values calculated from liver microsomal metab-
`
`olism intrinsic clearance data was made by Rane et al. (1977) for the
`rat. Intrinsic clearance data were obtained by determination of the
`enzyme kinetic parameters (Vmax and KM). In our work, we described
`two related methods whereby human clearance could be predicted
`from in vitro metabolism data (Obach et al., 1997). In one method, the
`
`
`
`enzyme kinetic parameters Vmax and KM were determined and con-
`verted to intrinsic clearance (CL',,,,)1, which is similar to that de-
`scribed by Rane et al. (1977). In the other method, referred to as the
`“in vitro T1,; method”, CL’,m was determined by measuring the
`first-order rate constant for consumption of the substrate at a low
`concentration. Interestingly, for both of these methods, a better cor-
`relation was observed between the actual and predicted clearance
`values if the free fraction in blood was disregarded in the well-stirred
`or parallel-tube equations describing hepatic extraction.
`One possible reason for the observation that a better prediction of
`human clearance was made when disregarding plasma protein binding
`was that the substrates were bound in the microsomal incubations, and
`that the extent of this binding could be great enough so as to almost
`cancel out the plasma protein binding term in the well-stirred and
`parallel-tube equations (Obach, 1996). This possibility was further
`substantiated in an examination of probe substrates propranolol, imip-
`ramine, and warfarin (Obach, 1997). In this report, it was demon-
`strated that the lipophilic amines propranolol and imipramine were
`bound to microsomes, and that incorporation of this binding term
`aided in the accurate prediction of human clearance from in vitro
`intrinsic clearance data. The acidic drug, warfarin, exhibited this
`phenomenon to a much lesser extent. However, for all three drugs
`overall, incorporation of both plasma protein and microsome binding
`
`Send reprint requests to: R. Scott Obach, Ph.D., Drug Metabolism Depart-
`ment, Candidate Synthesis, Enhancement, and Evaluation, Central Research
`Division, Pfizer, lnc., Groton, CT 06340. E-mail: obachr@pfizer.com
`
`1 Abbreviations used are: CL',,,,, intrinsic clearance; f,,(,,,,C,, unbound fraction in
`microsomal incubation mixtures; f,,(b.,,,,,,,, unbound traction in blood; Q, hepatic
`blood flow; ISTD, internal standard.
`1350
`
`WOCK- EXHIBIT 1024
`
`Z 9 [
`
`-1
`27:‘
`
`O a
`
`t E ,
`
`0
`L23
`
`2V
`
`J
`
`1'3
`
`0 i
`
`t
`
`+-
`H-1
`
`2 Ba
`
`t D
`
`

`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`MICROSOME BINDING IMPACT ON CLEARANCE PREDICTIONS
`
`1351
`
`FIG. 1.Chemical structures of the 29 drugs examined in this study.
`
`terms generally yielded more accurate predictions of human clear-
`ance.
`The objective of the experiments described herein is to more
`exhaustively test the hypothesis that microsomal binding is an impor-
`tant phenomenon in the prediction of in vivo pharmacokinetics from
`in vitro drug metabolism data. To this end, human hepatic microsomal
`metabolism data were gathered for 29 drugs, using the in vitro T1/2
`
`approach. Additionally, the extent of nonspecific binding to micro-
`somes in the in vitro matrix was measured for each drug. The drugs
`used in these experiments span a broad range of structural types (Fig.
`1) and include basic compounds (positively charged at pH 7.5), acidic
`compounds (negatively charged at pH 7.5), and neutral compounds
`(no charge at pH 7.5). The data set was used to project human
`clearance from the in vitro intrinsic clearance data to determine
`
`

`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`1352
`
`OBACH
`
`TABLE 1
`
`Sample processing and HPLC-MS conditions for 29 drugs used in this analysis
`
`Drug
`
`Internal Standard
`
`Incubation
`Termination
`
`Mobile Phase
`System
`
`CH3CN
`
`MS
`Polarity
`
`Basic compounds
`Chlorpromazine
`Propafenone
`Verapamil
`Diphenhydramine
`Lorcainide
`Diltiazem
`Amitriptyline
`Desipramine
`Imipramine
`Ketamine
`Quinidine
`Clozapine
`Neutral compounds
`Dexamethasone
`Prednisone
`Diazepam
`Midazolam
`Methoxsalen
`Alprazolam
`Triazolam
`Zolpidem
`Acidic compounds
`Diclofenac
`Ibuprofen
`Tolbutamide
`Warfarin
`Tenidap
`Tenoxicam
`Amobarbital
`Hexobarbital
`Methohexital
`
`Amitriptyline
`Verapamil
`Propafenone
`Propafenone
`Propafenone
`Propafenone
`Imipramine
`Amitriptyline
`Amitriptyline
`Metoprolol
`Ondansetron
`Diltiazem
`
`Prednisone
`Dexamethasone
`Midazolam
`Diazepam
`Diazepam
`Triazolam
`Alprazolam
`Quinine
`
`Ibuprofen
`Diclofenac
`Warfarin
`Tolbutamide
`Warfarin
`Piroxicam
`Methohexital
`Methohexital
`Amobarbital
`
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`NaOH
`
`NaOH
`NaOH
`NaOH
`NaOH
`CH3CN
`NaOH
`NaOH
`NaOH
`
`HCl
`HCl
`HCl
`HCl
`HCl
`HCl
`HCl
`HCl
`HCl
`
`1
`1
`1
`1
`1
`1
`1
`1
`1
`1
`1
`1
`
`1
`1
`1
`1
`1
`1
`1
`1
`
`2
`2
`2
`2
`2
`2
`2
`2
`2
`
`%
`
`36.5
`32.0
`32.0
`32.0
`32.0
`32.0
`36.5
`36.5
`36.5
`18.5
`18.5
`27.5
`
`32.0
`32.0
`50.0
`50.0
`50.0
`41.0
`41.0
`23.0
`
`32.0
`32.0
`27.5
`27.5
`32.0
`27.5
`45.5
`45.5
`45.5
`
`1
`1
`1
`1
`1
`1
`1
`1
`1
`1
`1
`1
`
`1
`1
`1
`1
`1
`1
`1
`1
`
`2
`2
`2
`2
`2
`2
`2
`2
`2
`
`m/z
`
`318.8
`341.9
`455.1
`256.0
`371.0
`415.0
`278.0
`266.5
`281.0
`237.8
`325.0
`326.9
`
`393.1
`359.1
`284.9
`325.8
`217.0
`309.0
`342.9
`308.0
`
`294.0
`205.1
`269.0
`307.3
`319.1
`336.1
`225.2
`235.1
`261.1
`
`Rt
`
`min
`
`1.2
`1.4
`1.6
`0.8
`1.4
`1.0
`1.0
`0.8
`1.0
`0.8
`1.5
`1.2
`
`1.8
`1.1
`1.4
`0.8
`1.0
`0.9
`1.0
`1.5
`
`1.1
`1.3
`1.2
`1.2
`1.3
`0.8
`0.8
`0.8
`1.5
`
`whether the most accurate projections are made by disregarding all
`binding data, including only blood binding values, or including both
`blood and microsomal binding values.
`
`Experimental Procedures
`
`Materials. The 29 drugs examined in these experiments were obtained from
`Sigma Chemical Co. (St. Louis, MO) with the exception of lorcainide (ob-
`tained from ICN, Aurora, OH), methoxsalen (obtained from Aldrich Chemical,
`Milwaukee, WI), zolpidem (obtained from Research Biochemicals Interna-
`tional, Natick, MA), and methohexital (obtained from Radian Inc., Dallas,
`TX). NADPH was obtained from Sigma. Solvents and other reagents were
`from common sources and were of HPLC grade or better. Human liver
`microsomes were from an in-house bank of liver microsomes maintained at
`Pfizer Central Research (Groton, CT). A pool was prepared from six liver
`microsomal preparations from six individual donors that were selected on the
`basis of having average activities for five of the major drug metabolizing
`cytochrome P-450 (CYP) enzymes (CYP1A2, CYP2C9, CYP2C19, CYP2D6,
`and CYP3A) normalized per microsomal protein content. Microsomes from
`putative CYP2D6 and CYP2C19 poor metabolizers were excluded. The P-450
`content of this pool, as determined by spectral means (Omura and Sato, 1964)
`was 0.26 nmol/mg microsomal protein. CYP isoform specific marker substrate
`activities were as follows: CYP1A2, phenacetin O-deethylase of 0.147 nmol/
`min/mg protein (at 50 mM phenacetin); CYP2C9, tolbutamide 4-hydroxylase
`of 0.23 nmol/min/mg protein (at 1.0 mM tolbutamide); CYP2C19, S-mephe-
`nytoin 49-hydroxylase of 0.093 nmol/min/mg protein (at 1.0 mM S-mepheny-
`toin); CYP2D6, bufuralol 19-hydroxylase of 0.075 nmol/min/mg protein (at 10
`mM bufuralol); and CYP3A4,
`testosterone 6b-hydroxylase of 2.7 nmol/
`min/mg protein (at 250 mM testosterone). All glassware was subjected to gas
`phase silylation before use.
`Metabolic Incubations. Human liver microsomal incubations were con-
`ducted in triplicate. General conditions are described as follows with details
`specific to each drug listed in Table 1. Incubation mixtures consisted of liver
`microsomes (0.3–10 mg microsomal protein/ml), substrates (1.0 mM), MgCl2
`
`(3.3 mM), and NADPH (1.3 mM) in a total volume of 0.5 ml potassium
`phosphate buffer (25 mM, pH 7.5). Reactions were commenced with the
`addition of NADPH and shaken in a water bath open to the air at 37°C. At T 5
`0 and at five time points ranging to 40 min, aliquots (50 ml) were removed and
`added to termination mixtures containing internal standards as listed in Table
`1. The samples were processed by extraction into methy t-butyl ether (3 ml),
`the aqueous layer was frozen in a dry ice-acetone bath, the organic solvent was
`decanted and evaporated under N2 at 30°C. The residue was reconstituted in 50
`ml HPLC mobile phase A (see below). For methoxsalen samples, the work-up
`procedure consisted of precipitation of protein with CH3CN (100 ml), removal
`of precipitated materials by centrifugation, and analysis of the supernatant by
`HPLC-mass spectrometry (MS).
`Equilibrium Dialysis. Drugs (1.0 mM) were mixed with human liver
`microsomes (at protein concentrations used for the respective metabolic incu-
`bations), MgCl2 (3.3 mM) and potassium phosphate buffer (25 mM; pH 7.5).
`The mixtures were subjected to equilibrium dialysis versus buffer/MgCl2 at
`37°C using a Spectrum apparatus (Spectrum Industries, Los Angeles, CA) as
`per instructions of the manufacturer. Spectra-Por no. 4 membranes, with
`molecular mass cutoff of 12 to 14 kDa, were used and the cells were rotated
`at 20 rpm for 5 h. (These dialysis conditions had been previously shown to give
`equilibrium for this dialysis apparatus; Obach, 1997). Dialysis experiments
`were done in triplicate. On completion of the dialysis period, the microsome
`and buffer samples were removed, processed as described above, and analyzed
`by HPLC-MS. Microsome samples (50 ml) were mixed with control buffer
`(100 ml), and buffer samples (100 ml) were mixed with control microsomes (50
`ml) to yield an identical matrix before sample work-up. Drug recovery through
`the dialysis procedure was determined by analyzing samples of the mixtures
`that were not subjected to dialysis, and recovery values were 86% or greater.
`HPLC-MS Analysis. The HPLC-MS system consisted of a Hewlett-
`Packard 1100 quaternary gradient HPLC pump with membrane degasser
`(Hewlett-Packard, Palo Alto, CA), a CTC PAL autoinjector (Leap Technolo-
`gies, Carrboro, NC), and a PE-Sciex API 100 single quadrupole mass spec-
`
`

`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`MICROSOME BINDING IMPACT ON CLEARANCE PREDICTIONS
`
`1353
`
`Values for systemic clearance, fraction unbound in plasma, and blood-to-plasma ratio for 29 drugs examined in this analysis
`
`TABLE 2
`
`Drug
`
`Fraction Unbound
`in Plasma (fu)
`
`Blood-to-Plasma
`Ratio
`
`Basic compounds
`Chlorpromazine
`Propafenone
`Verapamil
`Diphenhydramine
`Lorcainide
`Diltiazem
`Amitriptyline
`Desipramine
`Imipramine
`Ketamine
`Quinidine
`Clozapine
`Neutral compounds
`Dexamethasone
`Prednisone
`Diazepam
`Midazolam
`Methoxsalen
`Alprazolam
`Triazolam
`Zolpidem
`Acidic compounds
`Diclofenac
`Ibuprofen
`Tolbutamide
`Warfarin
`Tenidap
`Tenoxicam
`Amobarbital
`Hexobarbital
`Methohexital
`
`0.05
`0.04
`0.10
`0.22
`0.15
`0.22
`0.05
`0.18
`0.10
`0.88
`0.13
`0.05
`
`0.32
`0.25
`0.013
`0.05
`0.09
`0.32
`0.10
`0.08
`
`0.005
`0.01
`0.04
`0.01
`0.0007
`0.009
`0.39
`0.53
`0.27
`
`0.78
`0.70
`0.77c
`0.65c
`0.77
`1.0
`0.86
`0.96
`1.1
`0.82c
`0.92
`0.87
`
`0.93
`0.83c
`0.71
`0.53
`0.67
`0.78c
`0.62c
`0.76c
`
`0.55c
`0.55c
`0.55c
`0.55
`0.56
`0.67
`1.5
`1.0
`0.70c
`
`Nonrenal Clearancea
`
`Plasma
`
`Blood
`
`ml/min/kg
`
`8.6b
`13
`15
`6.2
`14
`12
`10
`12
`13
`16
`2.5
`2.5
`
`3.5
`4.1
`0.4
`4.6
`12
`0.59
`2.9
`4.3
`
`4.2
`0.8
`0.2
`0.045
`0.058
`0.02
`0.53
`3.6
`11
`
`11
`19
`19
`9.5
`18
`12
`12
`12
`12
`20
`2.7
`2.9
`
`3.8
`4.9
`0.6
`8.7
`18
`0.76
`4.7
`5.7
`
`7.6
`1.5
`0.36
`0.081
`0.10
`0.03
`0.35
`3.6
`16
`
`References
`
`Dahl and Strandjard, 1974; Maxwell et al., 1972; Lund, 1980
`Bryson et al., 1993
`Eichelbaum et al., 1984
`Blyden et al., 1986
`Somani et al., 1987; Klotz et al., 1978
`Echizen and Eichelbaum, 1986; Smith et al., 1983
`Schulz et al., 1983
`Brosen and Gram, 1988
`Sallee and Pollack, 1990; Abernathy et al., 1985
`White et al., 1985
`Greenblatt et al., 1977; Rakhit et al., 1984; Hughes et al., 1975
`Cheng et al., 1988
`
`Tseui et al., 1979; Peterson et al., 1983
`Schalm et al., 1977
`Greenblatt et al., 1980; Maguire et al., 1980
`Heizmann et al., 1983
`Billard et al., 1995; Pibouin et al., 1987
`Smith et al., 1984
`Smith et al., 1987
`Durand et al., 1992
`
`Willis et al., 1979; Chan et al., 1987
`Martin et al., 1990
`Balant, 1981; Scott and Poffenbarger, 1979
`O’Reilly, 1972
`Gardner et al., 1995
`Heintz et al., 1984
`Bachmann, 1987; Sawada et al., 1985
`Breimer et al., 1975; Sawada et al., 1985
`Breimer, 1976; Gillis et al., 1976
`
`a All clearance values from the literature were from i.v. dosing. In the case of dependence of clearance on genetic polymorphism of drug-metabolizing enzymes, data from poor metabolizers was
`excluded. Nonrenal clearance values were calculated by: Clnon-renal 5 Cltotal z (1 2 fraction of the dose excreted unchanged in urine).
`b Chlorpromazine clearance values from i.m. dose; assumes complete absorption from i.m. route.
`c Denotes blood-to-plasma ratios that were unavailable in the scientific literature. Values were determined in duplicate after incubation of drug at 1.0 mg/mL in whole blood at ambient temperature
`for 45 min.
`
`trometer (Sciex, Thornhill, Ontario, Canada) with a turbo ionspray interface.
`There were various mobile phases used for the different drugs as listed in Table
`1. Mobile phase system 1 consisted of 20 mM acetic acid (adjusted to pH 4
`with NH4OH) and CH3CN used at various percentages of organic solvent (as
`listed in Table 1). System 2 consisted of 5 mM NH4OAc (pH unadjusted) and
`CH3CN at various percentages as listed in Table 1. The column used was a
`Phenomenex Luna C18 narrow bore column (2.5 3 50 mm) with a 3-mm
`particle size (Phenomenex, Torrance, CA). The flow rate was 0.5 ml/min and
`the mobile phase composition was held isocratically for each analyte. The
`injection volume was 25 ml.
`The effluent was split with approximately 0.25 ml/min introduced into the
`turbo ionspray source of the mass spectrometer. Source parameters (e.g.,
`orifice voltage, temperature, gas flow rates, etc.) were individually optimized
`for each drug, and the molecular ion (either M 1 H1 or M 2 H2, depending
`on the orifice polarity) was followed for each compound and internal standard
`in the selected ion monitoring mode.
`Calculations. In the determination of the in vitro t1/2, the analyte/ISTD peak
`height ratios were converted to percentage drug remaining, using the T 5 0
`peak height ratio values as 100%. The slope of the linear regression from log
`percentage remaining versus incubation time relationships (2k) was used in
`5 20.693/k. Conversion
`the conversion to in vitro T1/2, values by in vitro T1/2
`to in vitro CL9
`int (in units of ml/min/kg) was done using the following formula
`(Obach et al., 1997):
`
`CL9int 5
`
`0.693
`in vitro T1/ 2
`
`z
`
`ml incubation
`mg microsomes
`
`z
`
`45 mg microsomes
`gm liver
`
`z
`
`20 gm liver
`kg b.w.
`
`For microsomal binding, the fraction unbound in the incubation mixture was
`calculated by:
`
`fu(mic) 5
`
`drug/ISTD peak height ratio in buffer sample
`2 z drug/ISTD peak height ratio in microsome sample
`
`with the factor of 2 in the denominator because the aliquot volume of buffer
`samples analyzed was twice that analyzed for the microsome samples (see
`above).
`The overall accuracies of clearance prediction methods were determined by
`(Obach et al., 1997):
`
`average fold error 5 10
`
`U(logSpredicted
`
`actual
`N
`
`D
`
`U
`
`Literature values for i.v. clearance, plasma binding, and blood-to-plasma ratio
`for the 29 compounds are listed in Table 2. For those compounds in which
`renal excretion of unchanged drug represents a significant component of total
`clearance, clearance values were corrected to nonrenal clearance values by:
`
`CLnonrenal 5 CLtotal z ~1 2 fraction of dose excreted unchanged in urine)
`
`Results
`The use of HPLC-atmospheric pressure ionization-MS was an
`important tool in the gathering of these metabolic lability and micro-
`somal binding data. The selectivity and sensitivity of this instrumen-
`tation permitted facile quantitation of a wide variety of drug struc-
`tures. Chromatographic methods were developed for each compound
`
`

`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`1354
`
`OBACH
`
`TABLE 3
`
`In vitro intrinsic clearance values and fraction unbound in the incubation conditions for 29 drugs examined
`Each in vitro T1/2 and microsomal binding value represents mean 6 S.D. for triplicate determinations. Intrinsic clearance values were calculated from in vitro T1/2 data as described in
`Experimental Procedures.
`
`Drug
`
`Basic compounds
`Chlorpromazine
`Propafenone
`Verapamil
`Diphenhydramine
`Lorcainide
`Diltiazem
`Amitriptyline
`Desipramine
`Imipramine
`Ketamine
`Quinidine
`Clozapine
`Neutral compounds
`Dexamethasone
`Prednisone
`Diazepam
`Midazolam
`Methoxsalen
`Alprazolam
`Triazolam
`Zolpidem
`Acidic compounds
`Diclofenac
`Ibuprofen
`Tolbutamide
`Warfarin
`Tenidap
`Tenoxicam
`Amobarbital
`Hexobarbital
`Methohexital
`
`Microsomal
`Concentration
`
`mg/ml
`
`1.0
`0.5
`0.5
`6.0
`1.0
`2.0
`0.5
`0.5
`0.5
`1.0
`5.0
`5.0
`
`5.0
`5.0
`5.0
`1.0
`0.5
`5.0
`1.0
`5.0
`
`0.3
`2.0
`10
`10
`3.0
`10
`10
`5.0
`1.0
`
`In Vitro
`T1/2
`
`min
`
`25 6 6
`8.0 6 0.4
`10 6 0.2
`49 6 24
`13 6 2
`21 6 3
`92 6 13
`74 6 24
`66 6 5
`23 6 3
`37 6 5
`27 6 5
`
`42 6 3
`47 6 1
`54 6 19
`3.9 6 0.1
`31 6 3
`105 6 66
`33 6 2
`44 6 5
`
`11 6 3
`36 6 4
`71 6 12
`.120
`26 6 2
`38 6 11
`66 6 5
`48 6 6
`13 6 2
`
`CL9int
`
`ml/min/kg
`
`25 6 6
`166 6 8
`122 6 2
`2.1 6 0.9
`50 6 6
`15 6 2
`14 6 2
`17 6 7
`19 6 2
`27 6 4
`3.4 6 0.5
`4.6 6 0.9
`
`3.0 6 0.2
`2.7 6 0.0
`2.3 6 0.7
`160 6 3
`40 6 3
`1.6 6 1.0
`19 6 1
`2.8 6 0.3
`
`189 6 39
`8.8 6 0.9
`0.90 6 0.15
`,0.52
`8.3 6 0.7
`1.7 6 0.4
`0.94 6 0.07
`2.3 6 0.3
`49 6 8
`
`fu(mic)
`
`0.11 6 0.02
`0.26 6 0.04
`0.43 6 0.10
`0.29 6 0.02
`0.52 6 0.03
`0.76 6 0.10
`0.15 6 0.04
`0.21 6 0.01
`0.18 6 0.04
`0.49 6 0.02
`0.32 6 0.17
`0.13 6 0.01
`
`1.00 6 0.07
`0.20 6 0.02
`0.28 6 0.05
`0.88 6 0.12
`0.94 6 0.11
`0.66 6 0.04
`0.78 6 0.09
`0.58 6 0.10
`
`1.00 6 0.13
`0.84 6 0.13
`0.95 6 0.03
`0.47 6 0.05
`0.32 6 0.01
`0.78 6 0.03
`0.76 6 0.08
`0.81 6 0.05
`0.86 6 0.13
`
`using the same column and only two types of mobile phases, with
`virtually the only customization required for each compound being
`determination of an optimal percentage of organic modifier (CH3CN)
`to effect elution of drug and internal standard within a reasonable run
`time.
`In vitro T1/2 data in pooled human liver microsomes for the 29
`compounds examined are listed in Table 3. Metabolic lability of this
`set of compounds spanned a wide range, the most stable compound
`being warfarin (in vitro T1/2 was immeasurably long at a microsomal
`protein concentration of 10 mg/ml), and the most labile being diclofe-
`nac, propafenone, and midazolam (scaled CL9
`int values of 160 ml/
`min/kg or greater). Within each general class of compounds (weak
`bases, weak acids, and neutral compounds), intrinsic clearance values
`spanned a broad range. Bases ranged from low intrinsic clearance
`values of 3.4 and 4.6 ml/min/kg for quinidine and clozapine, respec-
`tively, to high intrinsic clearance values of 122 and 166 ml/min/kg for
`verapamil and propafenone, respectively. Intrinsic clearance values
`for acids ranged from less than 0.52 ml/min/kg for warfarin and 0.90
`and 0.94 ml/min/kg for tolbutamide and amobarbital, respectively, up
`to 189 ml/min/kg for diclofenac. Intrinsic clearance values for the
`neutral compounds ranged from 1.6 ml/min/kg for alprazolam to 160
`ml/min/kg for midazolam.
`The extent of microsomal binding was determined for each com-
`pound using a microsomal protein concentration equal to that used in
`the metabolic incubations (Table 3). Because different protein con-
`centrations were used, the compounds cannot be rank ordered with
`regard to extent of binding to microsomes. The values ranged from no
`binding to approximately 90% bound. Furthermore, those compounds
`
`that exhibited the greatest extent of binding were not necessarily those
`in which the microsomal protein concentration was highest. In gen-
`eral, the weak bases demonstrated greater binding to microsomes,
`despite the fact that microsomal concentrations used for the bases
`were, on average, lower than those used for the neutral and acidic
`compounds.
`A summary of human blood clearance predictions from the in vitro
`data is presented in Table 4 and predicted clearance values are plotted
`versus actual clearance values in Fig. 2. Equations for both the
`well-stirred and the parallel-tube models of hepatic extraction were
`applied under three variations: disregarding all binding values (Table
`4, eqs. 1 and 4), including only blood binding (Table 4, eqs. 2 and 5),
`and including both blood and in vitro microsome binding (Table 4,
`eqs. 3 and 6). Overall accuracy values, determined as described in
`Experimental Procedures, are listed in Table 5. For all compounds
`examined (n 5 29), average fold error values were just over 2-fold in
`the cases in which either no binding values were considered or all
`binding values were considered. The most accurate method was the
`use of the parallel-tube model with both blood and microsome binding
`incorporated (average fold error of 2.13). Using only the blood bind-
`ing value in either model of hepatic extraction yielded very poor
`predictions of human clearance. When subsets of compounds were
`considered, some differences as to which were the most accurate
`methods were observed. For weak bases and neutral compounds,
`disregarding all binding in either model of hepatic extraction yielded
`the best agreement between actual human clearance values and those
`projected from in vitro intrinsic clearance data. However, for the
`acidic compounds, the most accurate clearance prediction methods
`
`

`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`MICROSOME BINDING IMPACT ON CLEARANCE PREDICTIONS
`
`1355
`
`Summary of human clearance values predicted from in vitro intrinsic clearance, protein binding, and microsome binding values using well-stirred and parallel-tube
`models of hepatic extraction
`
`TABLE 4
`
`Actual
`Human
`CLblood
`
`mL/min/kg
`
`No Binding
`(eq. 1)
`
`Well-Stirred
`
`fu(bl) Only
`(eq. 2)
`
`Predicted Human CLblood
`
`a
`
`Parallel-Tube
`
`fu(bl) and fu(mic)
`(eq. 3)
`
`No binding
`(eq. 4)
`
`fu(bl) Only
`(eq. 5)
`
`fu(bl) and fu(mic)
`(eq. 6)
`
`11
`19
`19
`9.5
`18
`12
`12
`12
`12
`20
`2.7
`2.9
`
`3.8
`4.9
`0.6
`8.7
`18
`0.76
`4.7
`5.7
`
`7.6
`1.5
`0.36
`0.08
`0.10
`0.03
`0.35
`3.6
`16
`
`11
`19
`18
`1.9
`15
`8.7
`8.2
`9.4
`10
`12
`2.9
`3.8
`
`2.6
`2.4
`2.1
`19
`14
`1.5
`10
`2.5
`
`19
`6.2
`0.86
`0.46
`5.9
`1.6
`0.90
`2.1
`15
`
`1.5
`6.5
`9.0
`0.7
`6.7
`2.9
`0.8
`2.8
`1.6
`12
`0.5
`0.3
`
`1.0
`0.8
`0.04
`8.8
`4.3
`0.64
`2.7
`0.3
`
`1.6
`0.2
`0.07
`0.01
`0.01
`0.02
`0.24
`1.2
`9.9
`
`8.6
`13
`13
`2.2
`9.9
`3.6
`4.2
`8.8
`6.6
`15
`1.4
`1.9
`
`1.0
`3.4
`0.2
`9.4
`4.5
`0.95
`3.3
`0.5
`
`1.6
`0.2
`0.07
`0.02
`0.03
`0.03
`0.32
`1.40
`11
`
`15
`21
`21
`2.0
`19
`11
`10
`12
`12
`15
`3.1
`4.1
`
`2.8
`2.5
`2.2
`21
`18
`1.5
`12
`2.6
`
`21
`7.2
`0.88
`0.46
`6.8
`1.6
`0.92
`2.2
`19
`
`1.5
`7.6
`11
`0.7
`7.8
`3.0
`0.8
`2.9
`1.6
`15
`0.5
`0.3
`
`1.0
`0.8
`0.04
`11
`4.7
`0.64
`2.8
`0.3
`
`1.6
`0.2
`0.07
`0.01
`0.01
`0.02
`0.24
`1.2
`12
`
`11
`17
`17
`2.3
`12
`3.9
`4.6
`11
`7.7
`20
`1.4
`1.9
`
`1.0
`3.6
`0.2
`12
`5.0
`0.97
`3.6
`0.5
`
`1.6
`0.2
`0.07
`0.02
`0.03
`0.03
`0.32
`1.45
`14
`
`Drug
`
`Basic compounds
`Chlorpromazine
`Propafenone
`Verapamil
`Diphenhydramine
`Lorcainide
`Diltiazem
`Amitriptyline
`Desipramine
`Imipramine
`Ketamine
`Quinidine
`Clozapine
`Neutral compounds
`Dexamethasone
`Prednisone
`Diazepam
`Midazolam
`Methoxsalen
`Alprazolam
`Triazolam
`Zolpidem
`Acidic compounds
`Diclofenac
`Ibuprofen
`Tolbutamide
`Warfarin
`Tenidap
`Tenoxicam
`Amobarbital
`Hexobarbital
`Methohexital
`
`Equations used are as follows:
`
`CLblood 5
`
`Q z CL9int
`Q 1 CL9int
`
`(1)
`
`CLblood 5
`
`Q z fu(blood) z CL9int
`Q 1 fu(blood) z CL9int
`
`(2)
`
`CLblood 5
`
`(3)
`
`Q z fu(blood) z
`
`Q 1 fu(blood) z
`
`CL9int
`fu(mic)
`CL9int
`fu(mic)
`S 2CLint
`D!
`S 2fu(blood)zCLint
`CLblood 5 Q z ~1 2 eS 2fu(blood)zCLint
`
`~5!
`
`~6!
`
`CLblood 5 Q z ~1 2 e
`
`Q
`
`CLblood 5 Q z ~1 2 e
`
`Q
`
`Qzfu(mic)
`
`~4!
`D!
`D!
`a A value of 21 ml/min/kg was used for human hepatic blood flow (Q).
`
`incorporated both blood and microsome binding. Figure 3 contains
`plots of accuracy of predicted clearance values using the six equations
`versus the respective human clearance values.
`
`Discussion
`The prediction of human pharmacokinetic parameters for new
`chemical entities has become an important approach in the drug
`discovery process. For any given drug discovery approach, hun-
`dreds of compounds will satisfy potency objectives, however few
`
`can be examined in humans. New chemical entities require exten-
`sive investigation and investment of resources prior to administra-
`tion to humans, and therefore it is desirable to be able to exclude
`compounds from this process that would be expected to exhibit
`unsatisfactory human pharmacokinetic properties. Recently, sev-
`eral investigators have described various methods and approaches
`whereby human pharmacokinetic parameters can be predicted from
`in vitro and/or preclinical pharmacokinetic data (Hoener, 1994;
`Gobburu and Shelver, 1995; Lave et al., 1997a,b; Obach et al.,
`
`

`
`1356
`
`OBACH
`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`FIG. 2.Plots of human clearance values predicted from in vitro intrinsic clearance data versus actual human clearance values.
`A, well-stirred model disregarding all binding parameters. B, well-stirred model incorporating only plasma protein binding. C, well-stirred model incorporating both
`plasma protein and microsome binding. D, parallel-tube model disregarding all binding parameters. E, parallel-tube model incorporating only plasma protein binding. F,
`parallel-tube model incorporating both plasma protein and microsome binding.
`
`1997; Kuhnz and Gieschen, 1998; Sarver et al., 1997; Mahmood,
`1998a,b). These methods vary in complexity and the amount of
`data required for accurate predictions.
`One of the simplest methods described to predict human clearance
`is the use of human hepatic microsomal lability data, termed the in
`vitro T1/2 approach (Obach et al., 1997). In this method, one incubates
`the test compound with human liver microsomes in the presence of
`appropriate cofactors (NADPH for CYP catalyzed reactions) and
`measures the first-order rate of consumption of the test compound.
`This rate is converted to an in vitro CL9
`int value, scaled-up to reflect
`CL9
`int in vivo, and inserted into a model of hepatic extraction. A high
`degree of success was previously reported for this particular approach,
`however the number of test compounds was low, and a majority of the
`
`compounds were of a similar physicochemical class (lipophilic
`amines). One of the objectives of the present experimentation was to
`further test the in vitro T1/2 approach with more compounds and
`greater structural diversity.
`The use of hepatic microsomes in the prediction of clearance
`requires acceptance of several assumptions and caveats: 1) metabolic
`..
`clearance is the major mechanism of clearance (i.e., CLmetabolism
`1 CLbiliary
`1 CLother); 2) the liver is the major organ of
`CLrenal
`.. SCLall other organs); 3) oxidative metabo-
`clearance (i.e., CLhepatic
`lism predominates over other metabolic routes such as direct conju-
`gative metabolism, reduction, hydrolysis, etc.; 4) rates of metabolism
`and enzyme activities in vitro are truly reflective of those that exist in
`vivo. Additionally, a tenet of the well-stirred and parallel-tube models
`
`

`
`Downloaded from
`
`dmd.aspetjournals.org
`
` at ASPET Journals on May 29, 2015
`
`MICROSOME BINDING IMPACT ON CLEARANCE PREDICTIONS
`
`1357
`
`TABLE 5
`Accuracy of clearance predictions from liver microsomal in vitro T1/2 values including and excluding values for blood binding and microsome binding
`
`Average Fold Error (predicted CLblood/actual CLblood)
`
`Basic compounds
`only (n 5 12)
`
`Neutral compounds
`only (n 5 8)
`
`Acidic compounds
`only (n 5 9)
`
`All compounds
`(n 5 29)
`
`Well-stirred model
`No binding considered (eq. 1)
`Including fu(blood) (eq. 2)
`Including fu(blood) and fu(mic) (eq. 3)
`Parallel-tube model
`No binding considered (eq. 4)
`Including fu(blood) (eq. 5)
`Including fu(blood) and fu(mic) (eq. 6)
`
`1.37
`5.17
`1.86
`
`1.31
`4.75
`1.60
`
`1.99
`3.90
`2.55
`
`1.99
`3.90
`2.53
`
`5.05
`3.91
`2.83
`
`5.35
`3.79
`2.74
`
`2.28
`4.39
`2.31
`
`2.28
`4.20
`2.14
`
`of hepatic extraction is that the unbound concentration of drug in the
`plasma is equal to the unbound concentration in the hepatocyte.
`Therefore, facilitated transport processes that could possibly be re-
`sponsible for drug uptake or drug extrusion from hepatocytes are not
`accounted for in these models. The in vitro T1/2 approach has two
`additional inherent assumptions that are not required if intrinsic clear-
`ance were determined using the more rigorous approach of calculating
`intrinsic clearance from enzyme kinetic data (i.e., Vmax/KM). These
`are: 1) the substrate concentration employed (1.0 mM in the case of
`this report) is well below the apparent KM for substrate turnover, and
`2) there is no significant product inhibition, nor is there any mecha-
`nism based inactivation of enzyme. Overall, clearance for these 29
`compounds was generally underestimated using any of the six ap-
`proaches, giving credence to the notion that the aforementioned as-
`sumptions are not completely valid in many cases. It should be noted
`that the 29 compounds chosen for this examination represent a set for
`which other clearance mechanisms (e.g., renal clearance, nonoxida-
`tive clearance, etc.) are known to be less important than hepatic
`oxidative metabolic clearance, but some are known to fall outside the
`scope of the aforementioned assumptions, e.g., nonhepatic metabo-
`lism of dexamethasone (Diederich et al., 1996; Tomlinson et al.,
`1997), nonoxidative components of metabolic clearance such as the
`reductive metabolism of warfarin (Moreland and Hewick, 1975) or
`glucuronidation of ibuprofen (Rudy et al., 1991), and product inhibi-
`tion of diltiazem (Sutton et al., 1997).
`A second objective of this work was to further explore the potential
`importance of nonspecific reversible binding