A population approach to enzyme characterization and identification: Application to phenacetin O-deethylation

Article abstract of DOI:10.1023/A:1007665310830

Purpose. To determine the enzyme kinetics (EK) and identify the human cytochrome(s) P450 (CYP) involved in the deethylation of phenacetin to acetaminophen using a population-based method. Methods. A sparse data set was generated from incubations containing human liver microsomes (n = 19) with phenacetin. Estimates of the EK parameters were obtained by fitting the concentration-velocity data to Michaelis-Menten models by using nonlinear mixed effects modeling. Relationships between the EK parameters and the CYP activities determined for these liver microsomes were examined. Results. A two-enzyme kinetic model with a saturated, low K_M enzyme and an unsaturated, high K_M enzyme capable of forming acetaminophen best fit the data. The population estimates of the EK parameters were V_max1, 911 pmol/min/mg protein; K_M1, 11.3 μM; and Cl_int2 0.4 μl/min/mg. The coefficients of variation for interliver variability in V_max1 and residual error of the model were 39% and 15%, respectively. When the selective catalytic activities were examined as potential covariates, 7-ethoxyresorufin O-deethylation (CYP1A2) activity was found to be associated with the low K_M enzyme, however, the high K_M enzyme(s) could not be identified. Conclusions. The population approach characterized the EK parameters and identified the low K_M enzyme responsible for phenacetin O-deethylation as CYP1A2. Population modeling of EK provides valuable information on inter- and intraliver variability in CYP dependent activities.

Full text of DOI:10.1023/A:1007665310830

Pharmaceutical Research, Vol. 17, No. 12, 2000
Research Paper
age or disease state) and PK parameters, and determine inter-
and intraindividual variability associated with these param-
eters (1,2). There is an increasing interest in the application of
population approaches to preclinical drug development pro-
grams, particularly in toxicokinetics (3,4). Furthermore, the
principles of population analysis can be related to other fields
such as enzyme kinetics (EK).
A Population Approach to Enzyme
Characterization and Identification:
Application to
Phenacetin O-Deethylation
The cytochromes P450 (CYP) play an important role in
the metabolism of drugs. A number of in vitro techniques
have been developed to characterize and identify the CYP(s)
responsible for the metabolism of a drug. These studies serve
as the foundation for making predictions concerning the in
vivo potential for drug-drug interactions and population vari-
ability with the use of that drug. Characterization of the ap-
parent EK for a given CYP biotransformation involves using
nonlinear least squares regression to fit Michaelis-Menten
models of EK. Identification of the enzyme(s) involved in the
biotransformation employs several approaches including cor-
relation of metabolite formation rate with CYP activity of a
1
,5
1
Donna J. Belle, Barbara J. Ring,
1
2
3
Sandra R. B. Allerheiligen, Michael A. Heathman,
Lisa M. O’Brien, Vikram Sinha, Lorin K. Roskos,
and Steven A. Wrighton
2
2
1
,4
Received June 27, 2000; accepted September 12, 2000
Purpose. To determine the enzyme kinetics (EK) and identify the
human cytochrome(s) P450 (CYP) involved in the deethylation of
phenacetin to acetaminophen using a population-based method.
Methods. A sparse data set was generated from incubations contain- panel of human liver microsomes, metabolism by cDNA-
ing human liver microsomes (n ס 19) with phenacetin. Estimates of expressed enzymes, and selective inhibition of CYPs in liver
the EK parameters were obtained by fitting the concentration- microsomes (5).
velocity data to Michaelis-Menten models by using nonlinear mixed
effects modeling. Relationships between the EK parameters and the
CYP activities determined for these liver microsomes were examined.
Results. A two-enzyme kinetic model with a saturated, low KM en-
zyme and an unsaturated, high KM enzyme capable of forming acet-
aminophen best fit the data. The population estimates of the EK
parameters were Vmax1, 911 pmol/min/mg protein; KM1, 11.3 ␮M; and
Clint2, 0.4 ␮l/min/mg. The coefficients of variation for interliver vari-
ability in Vmax1 and residual error of the model were 39% and 15%,
This in vitro approach to enzyme characterization and
identification is a multifaceted and labor-intensive process.
To streamline the current practice, a “population” approach
was examined to perform the EK analyses and to define re-
lationships between the determined EK parameters and the
form-selective catalytic activities from a characterized liver
bank using a sparse sample paradigm. Thus, the population
approach would combine the EK characterization and corre-
respectively. When the selective catalytic activities were examined as lation analyses into one step. Phenacetin O-deethylation
potential covariates, 7-ethoxyresorufin O-deethylation (CYP1A2) ac- (POD) was used as a test biotransformation. The results ob-
tivity was found to be associated with the low KM enzyme, however, tained with the population approach were compared to tra-
the high KM enzyme(s) could not be identified.
ditional procedures.
Conclusions. The population approach characterized the EK param-
eters and identified the low KM enzyme responsible for phenacetin
O-deethylation as CYP1A2. Population modeling of EK provides
MATERIALS AND METHODS
valuable information on inter- and intraliver variability in CYP de-
Phenacetin and NADPH were purchased from Sigma
pendent activities.
(
St. Louis, MO). 7-Ethoxyresorufin and resorufin were ob-
KEY WORDS: cytochromes P450; NONMEM; sparse sampling;
tained from Molecular Probes (Eugene, OR). Acetamino-
phen was obtained from Kodak (Rochester, NY), and fura-
fylline was obtained from Ultrafine Chemicals (Manchester,
England). Microsomes prepared from insect cells expressing
phenacetin O-deethylation; enzyme kinetics.
INTRODUCTION
Nonlinear mixed effects regression models were intro- CYP1A2 and CYP2C19 were obtained from Gentest Corpo-
duced to study and analyze the sparse data sets acquired from ration (Woburn, MA).
population pharmacokinetic (PK) studies (1,2). These models
Human liver samples (designated HLA through HLS)
estimate the typical values of the population’s PK parameters, were obtained and microsomes prepared and stored as pre-
assess the relationships between potential covariates (such as viously reported (6). Protein content was determined by the
Lowry method (7).
1
2
Drug Disposition or Pharmacokinetics, Eli Lilly and Company,
POD Assay
POD activity was determined by measuring the conver-
sion of phenacetin to acetaminophen. A typical 0.2 ml incu-
bation mixture contained 0.25 mg/ml of microsomes, 100 mM
sodium phosphate buffer (pH 7.4), reduced nicotinamide ad-
enine dinucleotide phosphate (NADPH) (1 mM) and phen-
acetin. After a 3 min 37°C preincubation, the reaction was
initiated with NADPH. After 25 min (linear rate conditions),
Indianapolis, IN 46285.
Abgenix, Fremont, CA 94555.
To whom correspondence should be addressed. (e-mail
wrighton_steven@lilly.com)
Current address: Pharmacokinetics, Schering Plough, Kenilworth,
NJ 07033.
ABBREVIATIONS: CV, coefficient of variation; CYP, cytochrome
P450; EK, enzyme kinetics; EROD, 7-ethoxyresorufin O-de-
ethylation; MLR, multiple linear regression; MOF, minimal value of
3
4
5
the objective function; PK, pharmacokinetics; POD, phenacetin O- the reaction was stopped by the addition of 0.2 ml cold metha-
2
2
deethylation; ␴ , intraliver variance; ␻ , interliver variance.
nol. The mixture was centrifuged at 4000 rpm for 10 min and
1531
0724-8741/00/1200-1531$18.00/0 © 2000 Plenum Publishing Corporation

1532
Belle et al.
acetaminophen in the supernatant was determined by HPLC Inc., Lexington, KY) using the equation for a single enzyme
with UV detection at 254 nm. Metabolites were separated on model (13) for the cDNA expressed CYPs:
a heated (35°C), 250 × 4.6 mm Zorbax-Phenyl column (Mac-
v = ͑Vmax1 × ͓S͔͒
ր
͑KM1 + ͓S͔͒
(1)
Mod Analytical, Inc., Chadds Ford, PA) using a mobile phase
consisting of 25 mM sodium phosphate buffer (pH 3.0)/
methanol (95:5 or 90:10) at a flow rate of 1 ml/min.
or with the human liver microsomes, using the equations for
a two enzyme model (13):
Investigation of POD in Human Liver Microsomes
v = ͑Vmax1 × ͓S͔͒ ͑KM1 + ͓S͔͒ + ͑Vmax2 × ͓S͔͒ ͑KM2 + ͓S͔͒ (2)
ր ր
A sparse data set for POD by microsomes from the char-
acterized liver bank was accomplished by sequentially assign-
ing HLA–HLS to one of three incubation groups. All incu-
bations were performed in duplicate. Group one incubations
or
v = ͑Vmax1 × ͓S͔͒
ր
͑KM1 + ͓S͔͒ + Clint2 × ͓S͔
(3)
contained phenacetin concentrations of 5, 30, 80, and 300 ␮M; where v is the velocity of POD; S is the initial concentration
group two contained phenacetin concentrations of 10, 40, 100, of phenacetin in the incubation; KM1 and KM2 represent the
and 500 ␮M; group three contained phenacetin concentra- high and low affinity enzymes, respectively; Vmax1 and Vmax2
tions of 20, 60, 200, and 1000 ␮M.
represent the maximum enzyme velocities for the high and
For traditional analysis of enzyme kinetics, POD activity low affinity enzymes; and Clint2 is equal to Vmax2/KM2 for an
was determined in three samples; samples with low (HLP), unsaturated, low affinity enzyme where KM2 > S.
moderate (HLG), or high (HLS) CYP1A2 activity. Phenac-
etin concentrations from 12.5 to 2000 ␮M (10 concentrations/
liver performed in duplicate) were used in the microsomal
incubations. Incubations with human recombinant CYPs
Population Analyses
The population analyses were conducted utilizing NON-
were performed using the same conditions as described MEM (version V, level 1.1, NONMEM Project Group, Uni-
above, except the mixtures contained 25 pmol CYP1A2, or 50 versity of California, San Francisco, CA). The base model was
pmol CYP2C19 and were incubated for up to 30 min. Control determined (i.e., no covariates) by fitting Michaelis-Menten
incubations were performed with microsomes from cells models of EK [Eqs. (1)–(3)] to sparse concentration-velocity
transfected with the expression vector alone and showed no data. First-order conditional estimation was implemented
detectable activity.
(14). Interliver variability was estimated according to addi-
tive, proportional, and exponential error models (14). Re-
sidual variability between the observed response (i.e., veloc-
ity) and those predicted by the model(s) was also estimated
Determination of CYP Activities
The selective CYP activities were determined by moni- using NONMEM. Intraliver variability was attributable to er-
toring 7-ethoxyresorufin O-deethylase (EROD) (5), couma- rors in the POD incubation and/or analytical methods and
rin 7-hydroxylase (5), S-mephenytoin 4Ј-hydroxylase (5), bu- was estimated using additive, proportional, and exponential
furalol 1Ј-hydroxylase (5), midazolam 1Ј-hydroxylase (6), error models (14).
chlorzoxazone 6-hydroxylase (8), diclofenac 4Ј-hydroxylase
In the covariate structural model, the vector population
(
9), and taxol 6-hydroxylase (10). CYP2B6 levels were deter- parameters are denoted with ␪. All available data (concen-
mined as previously reported (11).
tration-velocity) were fitted simultaneously to obtain the best
2
estimates of ␪ parameters, intraliver variances (␻ ) for each
2
random effect, and residual error variance (␴ ) that charac-
Traditional Analyses
terizes the EK statistical model. The iterative extended-least-
For correlation analyses, microsomes from livers HLA– squares fitting routine continues until a minimal value of the
HLS were incubated with 40 or 1000 ␮M phenacetin with or objective function (MOF) is reached. The best fit model in the
without 10 ␮M furafylline, a mechanism-based inhibitor of series was designated as the base model.
CYP1A2 (12). During the inhibition studies, microsomes in
CYP activities were introduced as covariates into the
buffer were preincubated with or without furafylline at 37°C base model by testing each covariate on each EK parameter.
for 15 min prior to the initiation of the reaction with phen- Those covariates that exhibited a significant relationship with
acetin. Univariate or stepwise multiple linear regression the EK parameters were added to the base model in combi-
(
MLR) analysis (JMP, version 3.21, SAS Institute Inc., Cary, nation so that a full model contained all possible covariates.
NC) was performed to compare POD activity with the CYP- To confirm the relationship between the EK parameters and
selective activities or protein levels from the microsomal covariates, the process was reversed, with covariates being
bank. The activities for CYP1A2, CYP2A6, CYP2C8, removed individually from the model. Covariates retained in
CYP2C9, CYP2D6, CYP2E1, and CYP3A4 were log-trans- the final model were those which when removed from the
formed to normalize their distribution prior to correlation model resulted in a significant increase in MOF (⌬MOF м10
analysis. Stepwise MLR analyses were performed in a for- points for 1 degree of freedom, p < 0.01).
ward mode with a probability to enter of 0.15.
The goodness-of-fit of each NONMEM analysis was also
Eadie-Hofstee plots were constructed to assess whether assessed by examining the scatterplot of predicted versus
one or two enzymes contributed to POD (Excel, Microsoft measured POD velocities, the scatterplot of predicted POD
Corp., Redmond, WA). Apparent Michaelis-Menten param- velocities versus weighted residuals, the percent standard er-
eters were determined from nonlinear, least-squares regres- rors of the mean parameter estimates (%SEM). Empirical
sion analysis (WINNONLIN version 1.5, Scientific Consulting Bayesian estimates of the model parameters were obtained

A Population Approach to Enzyme Characterization and Identification
1533
for each liver by means of the POSTHOC option of NON- Population Analyses
MEM.
The database consisted of 150 POD velocity measure-
ments from 19 livers. The final base model for POD incorpo-
rated a two enzyme kinetic model with a low KM site and an
unsaturated high KM site [Eq. (3)], an exponential error
model for interliver variability in Vmax1, as well as a propor-
RESULTS
Traditional Analyses
Analyses of POD EK in the three livers investigated tional error model for intraliver variability. Interliver variabil-
demonstrated biphasic curves with respect to substrate con- ity was estimated for the Vmax1 term of the EK model only.
centration (data not shown). Data from HLG and HLP were The interliver variability in V
fit to Eq. (3), whereas data from HLS were fit to Eq. (2). The the square root of the ␻ value (i.e., ␻ Vmax1) obtained by
was calculated by taking
max1
2
2
KM1 values for HLG, HLP, and HLS were 37 ± 6, 67 ± 4, and NONMEM and expressing it as percent coefficient of varia-
1
8
3 ± 3 ␮M, respectively, with corresponding Vmax1 values of tion (% CV) (14). Intraliver variability was estimated by tak-
01 ± 70, 269 ± 10, and 1601 ± 110 pmol/min/mg protein. ing the square root of the estimated error variance, ␴ (15).
2
Estimates of Clint2 for HLG and HLP were 0.34 ± 0.01, and Interliver variability terms for KM1 and Clint2 could not be
.24 ± 0.07 ␮l/min/mg protein, respectively. For HLS, the KM2 determined with good precision. Table 2 summarizes the pa-
value was 624 ± 274 ␮M with a Vmax2 value of 1248 ± 119 rameter, standard error, and residual variability estimates for
pmol/min/mg protein. the base model. Figures 1A and B show scatterplots of the
0
POD at concentrations representing the low KM1 (40 measured versus predicted concentrations and weighted re-
␮M) and high KM2 enzyme (1000 ␮M) were correlated with siduals versus predicted concentrations from the base model.
eight CYP activities and CYP2B6 content in a liver bank This model was then used as a basis for evaluating covariate
Table 1). Of the CYP parameters examined, only EROD, a effects.
probe for CYP1A2, showed a strong correlation (r ס 0.83, p When the CYP activities and CYP2B6 content (Table 1)
0.01) with POD at 40 and 1000 ␮M phenacetin. To further were tested as potential covariates on the EK parameters of
(
<
elucidate the high KM2 enzyme(s) involved in POD, furafyl- the base model, only the model incorporating EROD activity
line (10 ␮M) was preincubated in human liver microsomes to on Vmax1 resulted in a statistically significant improvement in
eliminate CYP1A2 involvement in this biotransformation fit (⌬MOF ס 117, p < 0.01) (Table 2). Interliver variability
(
data not shown). In the furafylline inhibited samples, POD decreased from 70% CV in the base model to 39% CV. The
activity at 40 ␮M phenacetin was almost completely abol- estimated intraliver variability improved from 20.3% CV in
ished. In incubations containing 1000 ␮M phenacetin- and the base model to 15.4% CV (Table 2). A dramatic improve-
furafylline-inhibited microsomes, remaining POD activity ment in the scatterplots of observed versus predicted POD
strongly correlated with S-mephenytoin 4Ј-hydroxylation (r velocities (Fig. 1C) and weighted residuals versus predicted
ס 0.93, p < 0.01), a probe for CYP2C19.
velocities (Fig. 1D) was noted when comparing the results
Kinetic analyses were performed in microsome cells ex- from the base model to the final model. The scatterplots from
pressing human CYP1A2 and CYP2C19. The apparent KM the final model showed that the differences between the ob-
values obtained with CYP1A2 and CYP2C19 were 10.4 and served and predicted values were small, further demonstrat-
8
18 ␮M, respectively.
ing a good fit. Little if any bias was evident in the weighted
a
Table 1. Form-Selective Catalytic Activites and Immunoquantified Levels of CYPs in a Human Liver Bank
7-Hydroxy 6-Hydroxy 4Ј-Hydroxy 4Ј-Hydroxy S- 1Ј-Hydroxy
6-Hydroxy
1Ј-Hydroxy
Microsomal EROD
coumarin
(2A6)
taxol
(2C8)
diclofenac
(2C9)
mephenytoin
(2C19)
bufuralol
(2D6)
chlorzoxazone midazolam CYP2B6
sample
(1A2)
(2E1)
(3A)
levels
HLA
116.1
41.5
53.1
29.7
35.4
41.5
37.2
40.6
43.2
69.5
36.4
28.7
21.3
45.6
42.0
25.2
31.8
21.3
114.4
470.2
308.3
116.2
836.0
484.3
545.7
412.7
323.7
1338.0
387.7
1031.0
79.7
406.0
428.2
1468.0
728.0
509.8
137.7
1594.0
156.8
279.3
184.7
495.8
456.4
141.3
114.0
4.7
183.5
110.3
76.3
52.2
94.1
0.0
597.0
337.0
254.0
659.0
517.0
305.0
373.0
546.0
501.0
308.0
532.0
150.0
591.0
463.0
298.0
292.0
677.0
413.0
441.0
58.0
14.0
65.6
0.0
158.4
0.0
44.0
49.2
46.8
87.6
34.8
0.0
20.8
99.2
26.0
13.6
13.2
31.6
125.6
17.3
30.4
18.6
41.1
97.3
28.0
44.0
24.9
38.2
34.3
10.7
15.0
69.5
13.9
91.8
44.3
50.4
12.5
36.7
2210.0
1210.0
2440.0
2790.0
1890.0
1470.0
920.0
1630.0
3400.0
920.0
989.0
833.0
595.0
100.0
81.8
79.0
230.1
508.9
28.5
183.2
39.2
387.0
91.4
60.2
23.8
116.6
78.2
HLB
HLC
HLD
HLE
HLF
HLG
HLH
HLI
439.0
4257.0
2382.0
1632.0
1143.0
5836.0
1365.0
1020.0
681.0
HLJ
HLK
HLL
HLM
HLN
HLO
HLP
HLQ
HLR
HLS
880.0
1030.0
1650.0
1280.0
1300.0
3800.0
1010.0
1080.0
1890.0
909.0
1244.0
7897.0
3157.0
734.0
722.0
2209.0
0.0
0.0
0.0
0.0
2543.6
1512.3
148.5
127.0
287.4
0.0
a
All activities are reported as pmol/min/mg protein. EROD ס 7-ethoxyresorufin O-deethylation.

1534
Belle et al.
Table 2. Phenacetin O-Deethylation Population Model Parameter the individual predicted velocities and the observed velocities
Estimates
for each of the 19 livers (Fig. 2).
Base model
Final model
DISCUSSION
Parameter
Estimate
% SEM
Estimate
% SEM
9.8
a
The objectives of this study were to use a population
approach to characterize and identify the CYPs that catalyze
POD using a sparse data set generated in a bank of 19 mi-
crosomal samples and to compare the results to the tradi-
tional approach. A two enzyme model with a low KM1 en-
zyme and an unsaturated high KM2 enzyme site best fit the
sparse concentration-velocity data set. In the final population
model, Vmax1 was related to EROD activity (CYP1A2). The
addition of EROD into the model explained approximately
␪
19.8
1
Vmax1 (pmol/min/mg
protein)
KM1 (␮M)
Clin12 (␮l/min/mg
protein)
Interliver variability
on Vmax1 (% CV)
Intraliver variability
791
6.9
15.0
38.5
911.4
11.3
18.3
21.6
31.2
29.0
0.56
70.1
25.3
33.1
26.6
0.41
39
(
% CV)
20.3
1732
15.4
1615
30% of the interliver variability. The final model predicted
MOF
Vmax1 values for the 19 livers ranged from 214.9 to 2315.4
a
Structural model: Vmax ס ␪1 *ethoxyresorufin O-deethylation; pmol/min/mg protein (data not shown). The typical values for
where ␪1 is the coefficient for ethoxyresorufin O-deethylation ac- KM1 and Clint2 were 11.3 ␮M and 0.41 ␮l/min/mg protein,
tivities in Vmax1
.
respectively. Because of the relatively small data set, model-
predicted estimates for KM1 and Clint2 for each liver could not
be determined because of the inability to estimate interliver
variability for these parameters with adequate precision.
Validation of the population model was obtained from
traditional analyses and the literature. The analyses of POD
in the three human livers investigated demonstrated biphasic
Abbreviations used: CV(%) ס coefficient of variation, SEM ס stan-
dard error of the mean, MOF ס minimum objective function, Vmax1
ס maximal velocity of phenacetin O-deethylation of the low KM
enzyme, KM1 ס phenacetin concentration at 1/2 Vmax1, and Clint2 is
equivalent to Vmax2/KM2
.
residuals (Fig. 1D). The final population model predicted EK suggesting the involvement of at least two CYPs. Involve-
Vmax1 for HLG, HLP, and HLS were 745, 215, and 2079 ment of multiple CYPs in POD as demonstrated by both the
pmol/min/mg protein, respectively. These values compare population and traditional studies is consistent with published
well to the Vmax1 values obtained for HLG, HLP, and HLS by in vitro kinetic studies (15,16). Estimates for Vmax1 in HLG,
traditional methods. An excellent fit was observed between HLP, and HLS for POD from both the population and tradi-
Fig. 1. Scatterplots of (A) base model predicted versus measured phenacetin O-deethylation
velocity measurements, (B) base model weighted residuals versus predicted velocity measure-
ments, (C) final model predicted versus measured phenacetin O-deethylation velocity measure-
ments, and (D) final model weighted residuals versus predicted velocity measurements.

A Population Approach to Enzyme Characterization and Identification
1535
Fig. 2. The individual predicted velocity-phenacetin concentration profiles (solid line) for each liver. The mean measured velocities (open
square) for each liver are indicated.
tional approaches demonstrate that the two methods are in tions indicate that CYP1A2 is the low KM enzyme and
good agreement. The population estimate of 11.3 ␮M for KM1 CYP2C19 is the high KM enzyme responsible for POD in liver
was slightly lower than the estimates obtained from the tra- microsomes.
ditional analysis, which ranged from 13 to 67 ␮M. However, a
KM1 value of 11.3 ␮M is well within in the reported literature to estimate interliver variability in KM1 and Clint2 with sub-
values, which span from 6 to 68 ␮M (15,17). sequent predictions of these parameters in each liver. This
One limitation of the population model was the inability
Incorporation of EROD into the model for Vmax1 sig- probably stems from the fact that the sparse data set was not
nificantly improved the objective function. EROD is a known large enough to estimate these EK parameters, their respec-
marker reaction for CYP1A2 activity (5). Because EROD tive interliver variabilities, and the intraliver variability term.
was a significant covariate with Vmax1, this indicates that Another shortcoming in the study design was that the data set
CYP1A2 is the low KM enzyme responsible for POD. This was biased to ensure identification of the low KM1 enzyme. In
finding was corroborated with the traditional analyses and the one-third of the population (Group 1), the highest phenacetin
literature (16,17). As additional confirmation of this conclu- concentration examined was 300 ␮M. Perhaps there were not
sion, recombinant CYP1A2 had a KM value for POD similar enough data points at high substrate concentrations (S > KM2
)
to the low KM enzyme in liver microsomes. to adequately characterize that section of the concentration-
Both population and traditional EK analyses indicated velocity curve. The KM term is dependent upon factors that
that low and high KM enzymes contributed to POD. To elu- influence substrate binding (18). Because all microsomal
cidate the high KM enzyme, furafylline was preincubated with samples were incubated under the same experimental condi-
the microsomes. In incubations pretreated with furafylline tions, the KM of the enzyme(s) should remain relatively con-
and containing 1000 ␮M phenacetin, POD activity correlated stant between the livers and therefore a composite KM may
with S-mephenytoin 4-hydroxylase activity. Supplementary be sufficient. However, given this information, Clint2, which
evidence for the involvement of CYP2C19 was provided using reflects Vmax2/KM2, should be directly related toVmax2
.
microsomes derived from cells expressing CYP2C19. The KM Hence, in the future, it would be beneficial to estimate Clint2
value obtained in recombinant CYP2C19 was similar to the for each liver, although this may require a different experi-
high KM2 POD component in HLS. Together, these observa- mental design or a larger data set. Simulation studies are

1536
Belle et al.
planned to determine an optimal study size and design for
complete EK analyses. If a larger data set is necessary, then
the population method may not be practical. The advantages
of the population approach is that it requires roughly half the
number of incubations (140–150 incubations) as the tradi-
tional analyses, but provides additional information such as
EK parameter estimates for each liver examined and inter-
liver variability for the parameter.
In conclusion, similar EK parameters for POD were ob-
tained with both the traditional and population approach.
Both approaches identified CYP1A2 as the low KM1 CYP
responsible for POD. Traditional studies indicated that
CYP2C19 is the enzyme responsible for the high KM2 com-
assessing human hepatic drug metabolism: their use in drug de-
velopment. Drug Metab. Rev. 25:453–484 (1993).
. B. J. Ring, S. N. Binkley, L. Roskos, and S. A. Wrighton. Effect
of fluoxetine, norfluoxetine, sertraline and desmethyl sertraline
on human CYP3A catalyzed 1Ј-hydroxy midazolam formation in
vitro. J. Pharmacol. Exp. Ther. 275:1131–1135 (1995).
7. O. H. Lowry, N. H. Rosenbrough, A. L. Farr, and R. J. Randall.
Protein measurement with the folin phenol reagent. J. Biol.
Chem. 193:265–275 (1951).
6
8
. S. V. Chittur and T. S. Tracy. Rapid and sensitive high-perfor-
mance liquid chromatographic assay for 6-hydroxychlorzoxazone
and chlorzoxazone in liver microsomes. J. Chromatogr. B
Biomed. Sci. Appl. 693:479–483 (1997).
9
. T. Leemann, C. Transon, and P. Dayer. Cytochrome P450TB
(
CYP2C): a major mononoxygenase catalyzing diclofenac 4Ј-
hydroxylation in human liver. Life Sci. 52:29–34 (1993).
ponent, whereas the population method failed to identify the 10. J. W. Harris, A. Rahman, B.-R. Kim, F. P. Guengerich, and J. M.
high KM2 component. A major advantage of the population
approach is the ability to estimate the typical EK parameter
values for the population and predict the EK parameters for
each liver. In addition, the population approach provided in-
formation on inter- and intraliver variability for a given bio-
transformation. In the future, this information may be used to
better understand interindividual variability in drug metabo-
lism and better predict drug-drug interactions with new drug
entities.
Collins. Metabolism of taxol by human hepatic microsomes and
liver slices: participation of cytochrome P450 3A4 and an un-
known P450 enzyme. Cancer Res. 54:4026–4035.
1. S. Ekins, M. Vandenbranden, B. J. Ring, J. S. Gillespie, T. J.
Yang, H. V. Gelboin, and S. A. Wrighton. Further characteriza-
tion of the expression in liver and catalytic activity of CYP2B6. J.
Pharmacol. Exp. Ther. 286:1253–1259 (1998).
2. K. L. Kunze and W. F. Trager. Isoform-selective mechanism-
based inhibition of human cytochrome P450 1A2 by furafylline.
Chem. Res. Toxicol. 6:649–656 (1993).
13. I. H. Segel. Enzyme Kinetics: Behavior and Analysis of Rapid
Equilibrium and Steady-State Enzyme Systems, John Wiley and
Sons, New York, 1993.
4. A. J. Boeckman, S. L. Beal, and L. B. Sheiner (eds). NONMEM
Users Guides. NONMEM project group, University of Califor-
nia, San Francisco, CA, 1992.
15. A. R. Boobis, G. C. Kahn, C. Whyte, M. J. Brodie, and D. S.
Davies. Biphasic O-deethylation of phenacetin and 7-ethoxycou-
marin by human and rat liver microsomal fractions. Biochem.
Pharmacol. 30:2451–2456 (1981).
1
1
REFERENCES
1
1
. L. B. Sheiner, B. Rosenberg, and V. V. Marathe. Estimation of
population characteristics of pharmacokinetic parameters from
routine clinical data. J. Pharmacokinet. Biopharm. 5:445–479
(
1977).
2. C. C. Peck, D. Z. D’Argenio, and J. H. Rodman. Analysis of
pharmacokinetic data for individualized drug dosage regimens. In
W. E. Evans, J. J. Schentag, and W. J. Jusko (eds.), Applied 16. W. Tassaneeyakul, D. J. Birkett, M. E. Veronese, M. E. McMa-
Pharmacokinetics: Principles of Therapeutic Drug Monitoring,
Applied Therapeutic, Inc., Vancouver, 1992, pp. 1–31.
. A. M. Dahlem, S. R. Allerheiligen, and M. J. Vodicnik. Concomi-
nus, R. H. Tukey, L. C. Quattrochi, H. V. Gelboin, and J. O.
Miners. Specificity of substrate and inhibitor probes for human
cytochromes P450 1A1 and 1A2. J. Pharmacol. Exp. Ther. 265:
401–407 (1993).
3
tant toxicokinetics: techniques for and interpretation of exposure
data obtained during the conduct of toxicology studies. Toxicol. 17. L. L. von Moltke, D. J. Greenblatt, S. X. Duan, J. Schmider, L.
Pathol. 23:170–178 (1995).
Kudchadker, S. M. Fogelman, J. S. Harmatz, and R. I. Shader.
Phenacetin O-deethylation by human liver microsomes in vitro:
inhibition by chemical probes, SSRI antidepressants, nefazodone
and venlafaxine. Psychopharmacology 128:398–407 (1996).
18. R. A. Copeland. Enzymes: A Practical Introduction to Structure,
Mechanism and Data Analysis, Wiley-VCH, New York 1996.
4
. S. M. Pai, S. H. Fettner, G. Hajian, M. N. Cayen, and V. K. Batra.
Characterization of AUCs from sparsely sampled populations in
toxicology studies. Pharm. Res. 13:1283–1290 (1996).
. S. A. Wrighton, M. Vandenbranden, J. C. Stevens, L. A. Shipley,
B. J. Ring, A. E. Rettie, and J. R. Cashman. In vitro methods for
5