- Email: [email protected]
Incorporation of in vitro enzyme data into the physiologically-based pharmacokinetic (PB-PK) model for methylene chloride: implications for risk assessment
Toxicology Letters, 43 (1988) 97-l 16
97
Elsevier
TXL 02034
Incorporation of in vitro enzyme data into the physiologically-based pharmacokinetic (PB-PK) model for methylene chloride: implications for risk assessment R.H. Reitz’, A.L. Mendrala’, F.P. Guengerich3
C.N. Park’, M.E. Andersen2 and
‘Dow Chemical Company, Health and Environmental Sciences, Midland, MI (U.S.A.), ‘Wright Patterson Air Force Base, Armstrong Aerospace Medical Research Laboratory, Dayton, OH (U.S.A.), ‘Vanderbilt University, Department of Biochemistry, NashviNe, TN (U.S.A.) (Received
19 April,
(Revision
received
(Accepted
6 June,
1988) 1 June,
1988)
1988)
Key words: PB-PK;
Pharmacokinetics;
Methylene
chloride;
Risk assessment
SUMMARY Physiologically-based tainty inherent for incorporating Methylene
pharmacokinetic
in extrapolating data
chloride
(PB-PK)
models provide
a mechanism
for reducing
the results of animal toxicity tests to man. This paper discusses
from
in vitro
studies
is used as an example,
of xenobiotic
and carcinogenic
metabolism risk estimates
the uncera technique
into in vivo PB-PK incorporating
models.
PB-PK principles
are presented.
INTRODUCTION
Human beings are exposed to many chemicals, natural and synthetic. To assess whether those chemicals are likely to produce adverse effects in human populations, toxicity tests are conducted in laboratory animals. Inherent in this procedure is the assumption that laboratory animals are reasonable surrogates for humans; that is, Address Chemical
for correspondence: Company,
Abbreviations:
PB-PK,
glutathione-S-transferase; threshold
Midland,
limit value.
R.H.
Reitz,
Health
and Environmental
Sciences,
1803 Building,
Dow
MI 48674, U.S.A.
physiologically-based QRA,
quantitative
pharmacokinetics;
MFO, mixed function
risk
LMS,
assessment;
linearized
oxidase;
multi-stage;
GST, TLV,
98
most materials that fail to produce adverse effects in animals will not produce adverse effects in human populations. However, significant differences usually exist between the conditions prevailing in laboratory tests and those encountered by humans. Animals
used in laboratory
tests come from carefully
bred homogeneous
strains
and are carefully screened to be free of disease. At the start of the test, all the animals are young adults. For practical reasons, small groups are used, and a limited number of doses are evaluated. Doses are usually high, and chemicals are administered by the most experimentally convenient route. In contrast, human populations are heterogeneous and contain healthy and diseased individuals of all ages. Human exposures typically involve doses much lower than those employed in the animal tests and frequently occur through one or more routes different than the animal studies. Therefore, a variety of extrapolations are necessary in order to estimate human hazard from animal studies. Each extrapolation introduces uncertainty into the final risk assessment. Physiologically-based pharmacokinetic (PB-PK) models are useful for reducing some of this uncertainty because they take into account actual physiological properties of the various species, compound-specific partitioning of the chemical between tissues and blood, and concentration-dependent rates of metabolism. PB-PK models are particularly useful for high dose/low dose extrapolations, dose route extrapolations, and interspecies extrapolations. We have reported a procedure for incorporating the techniques of PB-PK modeling into the risk assessment for CHzClz [l]. A PB-PK model was developed providing quantitative descriptions of the rates of metabolism and levels of CHzClz in various organs of four mammalian species (mouse, hamster, rat, and human). The model was validated by comparing computer simulations with experimental data gathered in mice, rats, and humans. Levels of reactive metabolites produced in target tissues were correlated with tumor incidences in two chronic bioassays of CH2C12 in the B6C3Fl mouse [2,3]. Finally, relevant measures of tissue dose were calculated for use in a pharmacokinetically-based risk assessment. Development of this PB-PK model required the acquisition of 3 types of model parameters: (1) physiological constants, (2) partition coefficients, and (3) metabolic rate constants. Physiological parameters for the various species were available from the scientific literature [4-61, and partition coefficients were determined experimentally at Wright Patterson Air Force Base (WPAFB) according to the procedures of Sato and Nakajima [7]. In vivo gas uptake studies conducted at WPAFB according to the method of Gargas et al. [8] were used to estimate metabolic rate constants for two pathways of CH2C12 metabolism in animals [ 11. These pathways involve oxidation of CHzClz by cytochrome P450-dependent enzymes (mixed function oxidase, MFO) [9] and conjugation of CH2C12 with glutathione by glutathione-S-transferase enzymes (GST) [lo]. Metabolic rate constants for the MFO pathway in humans were derived from
99
measurements of carboxyhemoglobin, exhaled carbon monoxide, and CHzClz uptake in human volunteers exposed to either 100 or 350 ppm of CH2Cl2 at the Dow Chemical Company [ 11. However, there were no in vivo data available for direct calculation of the rate constants for the GST pathway in humans. Since MFO enzymes have a higher affinity for CH2C12 than the GST enzymes [ 111, most of the metabolism of CHzClz occurs through MFO at low concentrations. Therefore, the human studies of CHzClz metabolism (which all used relatively low concentrations of CHX12) are not appropriate for estimating the rate of metabolism by the GST pathway, Consequently, the rate constants for the GST pathway were estimated by allometric scaling of the in vivo rate constants for GST metabolism in animals [ 11. Although allometric scaling has been used by others for estimating metabolic rates in different species [12], it was clear that data supporting the validity of such scaling for CHzCl2 would be valuable. Additional experiments to estimate the value of the metabolic rate constants for the GST pathway in humans have now been conducted. These experiments involved in vitro measurements of the glutathione-dependent metabolism of CH2C12 in tissue samples from humans as well as tissue samples from F-344 rats, B6C3Fl mice, and Syrian Golden hamsters. Similar procedures were employed to make in vitro measurements of the rate of oxidation of CHzClz by the MFO enzymes, and the reliability of the in vitro to in vivo extrapolation was assessed by comparing observed in vivo rates of MFO metabolism with those predicted from the in vitro data. This paper will describe the methods employed to collect the in vitro data and show how they may be incorporated into PB-PK risk assessments for CH2C12. METHODS
Test materials Spectroscopic grade CH2Cl2 (< 99.9% purity) was obtained from Fisher Chemical Co. CH~~~c12(0.5 Ci/mol) was synthesized by New England Nuclear (Boston, MA). This material was checked for radiochemical purity by gas chromatography/mass spectrometry prior to use and was found to be 95.0% CHZ~%J~, with small amounts of CH336C1 (0.6%) and CH3%Z13(2.8%) identified as minor impurities. Preparation of enzymes Cytosolic and microsomal fractions of lungs and livers were prepared from male F-344 rats, B6C3Fl mice, Syrian Golden hamsters, and tissues of otherwise healthy human accident victims selected for organ transplantation through the Nashville Regional Organ Procurement Agency, Nashville, TN. Tissues were removed from animals or humans and small pieces were rapidly frozen either by immersion in liquid nitrogen (humans) or with dry ice (animals). Soluble enzymes (cytosol) and particulate enzymes (microsomes) were prepared as described elsewhere [ 13-181.
100
Samples of human liver from four different individuals were processed individually. Availability of human lung samples was limited, so lung samples from two individuals were pooled ( - 8 g total tissue weight) for the enzyme preparations from human lung. (None of the lung samples were from the same individuals as the liver samples). Assay of A4FO Incubations (30 min, 37°C) were conducted in 5.0 ml vials sealed with Teflon@coated rubber septa (4.0 ml headspace). Vials contained an NADPH regenerating system in addition to phosphate buffer (pH 7.4) and CHzClz. At the end of incubation, the contents of the vial were transferred to a glass tube containing 0.5 ml of a solution of 2% NaX03 and 1% NaCl as well as 3.0 ml of unlabeled CHZCIZ, and vortexed vigorously for 10 sec. Centrifugation served to separate the unreacted substrate (in the organic layer) from the water-soluble 36C1- formed by enzyme reaction. Assay of GST Incubations (30 min, 37°C) were conducted in 1.8 ml glass vials sealed with Teflon@-coated rubber septa (< 0.1 ml headspace in vial). Each incubation contained (in addition to CHzC12) potassium phosphate buffer (So-100 mM, pH 7.4) and reduced glutathione (10 mM). 36C1_ was separated from the unreacted CHzC12 as outlined for the MFO assay. RESULTS
A4FO enzymes
A series of incubations
with microsomes
prepared
from liver were conducted
at
CHzClz concentrations varying from 1 mM to 10 mM. The reaction rates observed (nmol/min per mg protein) are listed in Table I. The highest rates of oxidation were observed in the hamster liver microsomes, with slightly lower rates in the mouse liver microsomes. Lower rates were observed in rat and human liver microsomes (Table I). Microcomes prepared from lung tissues in the various species were assayed at a single substrate concentration: 5 mM CHzClz. Microsomes from mouse lungs were much more active than microsomes from the lungs of the other species (Table I). No MFO activity was detected in pooled samples of human lung microsomes, with a detection limit of approximately 0.1 nmol/min per mg protein. Double reciprocal plots (l/velocity versus l/substrate; Lineweaver and Burk [19]) of the results obtained with liver microsomes are presented in Fig. 1A (B6C3Fl mice), Fig. 1B (F-344 rats), Fig. 1C (Syrian Golden hamsters), and Fig. 1D (human sample HL-99). Kinetic data for three other human liver samples were analyzed but are not shown.
101
TABLE
I
REACTION Enzyme
RATE
activities
(nmol/min/mg
(GST, and MFO) observed
from liver tissue of male B6C3Fl reported Cont.
protein)
as nmol product
formed/min/mg
Mouse
(mM)
at various
concentrations
mice, F-344 rats, Syrian
Golden
of CHzClz with enzymes prepared
hamsters,
and humans.
All values are
protein. Rat
Hamster
Human
Human
(HL-99)
(HL-109)
MFO assays: Liver 1.00
5.87a
2.40=
7.18=
1.49
1.66”
1.23
6.52a
2.42a
8.16=
1.85
2.06a 2.26a
1.64
7.18”
2.93”
8.54a
2.09
2.60
9.02”
3.44a
11 .02a
2.28
2.87a
5.00
11.40a
4.10a
14.47a
3.34
4.46”
10.00
14.40”
4.91a
18.18a
4.58
4.80a
4.62
0.16
0.99
6.1
4.17
0.75
0.182
10.0
7.24
1.11
0.307
Lung 5.00
GST assays: Liver
1.36
0.383
1.53
16.7
12.3
2.08
0.482
1.95
1.66
25.0
18.5
3.19
0.76
2.58
2.24
9.92
12.5
50.0
33.2
100.0
48.6
12.1
6.17
1.3
1 .o
1.28
1.24
3.95
3.51
2.64
4.74
3.94
0.0
0.37
Lung 40.0
a Mean value from two independent b Pooled
human
lung samples,
experiments.
no activity
observed
at detection
limit of 0.05-O. 1 nmol/min/mg
protein.
Values of Km (the affinity constant) and V,,, (the maximum rate of the enzyme reaction) were determined by computer optimization (method of least squares) of the data listed in Table I. The results (+ standard deviation) are listed in Table II. The values of Km were almost identical in the 7 sets of liver microsomes analyzed, ranging from 0.92 mM (Human-105) to 2.8 mM (Human-log). However, the values of V,,, in the various species differed significantly. The highest values of V,,, were observed in microsomes from hamster liver ( Vmax=20.8), with lower rates in mice (V,,, = 15.9), rats (V,,, = 5.39), and humans (range= 1.5-13.0, mean V,,,=6.5). GST enzymes
Assays of GST in cytosol prepared from the livers of the various species were con-
102
ducted at CH2C12 concentrations from 6.7 mM to 100 mM. (The limited solubility of CH2Cl2 in water prevented testing concentrations higher than 100 mM). Enzymes prepared from mouse liver contained the highest levels of the GST enzyme (Table I). Tissues from the rat liver contained intermediate levels of GST, while tissues from the hamster and human livers contained the lowest levels of GST (Table I). Cytosol preparations from the lungs of the different species were assayed at a single substrate concentration: 40 mM CHzClz. The highest levels of GST activity in lung tissue were observed in the mouse. The rat lung had intermediate activity, and the lowest activity was observed in cytosol prepared from human and hamster lungs.
0.50
0.20
I
0.40 0.15 E ‘5 0 5 z
0.30 0.10 0.20 0.05
0.10
0.00
0.00 200
0
P .z :: 5 c
400 600 1 /Substrate
800
0
1OOC
0.20
0.8
0.15
0.6
0.10
0.4
0.05
0.2
f
0.00
200
0
400
600
800
I 1000
200
activity
in microsomes
ture varied
from
plots (l/velocity prepared
versus
from various
1 mM to 10 mM, and the incubations
800
1000
I /Substrate
0 0
l/substrate)
species.
600
D
200
400 600 1 /Substrate
1 /Substrate
Fig. 1. Lineweaver-Burk
I
400
of the concentration
Concentrations
800
1000
dependence
of MFO
of CHsClz in the incubation
also contained
50 mmol potassium
mix-
phosphate
- 1 unit of glucose-6-phosphate debuffer (pH 7.4). 10 mmol glucose 6-phosphate, 0.5 mmol NADP+, hydrogenase, CHZ%Z, and l-2 mg of microsomal protein in a total volume of 1 .O ml. Plots are shown for (A) mouse
microsomes,
(B) rat microsomes, prepared
(C) hamster microsomes, from liver #99.
and (D) human
microsomes
103
TABLE
II
KINETIC
CONSTANTS
Kinetic constants
obtained
from in vitro experiments
from the livers of male B6C3Fl obtained
by relative
the best estimate
least squares f
Species
with GST and MFO assays in enzyme preparations
mice, F-344 rats, Syrian Golden weighting
SD of the estimate
hamsters,
using the SimuSolv@ computer
as calculated
by the computer
Km (mM)
n
and humans. program.
Constants
were
Values reported
are
program. If,,,,, (nmol product formed/min/mg
prot.)
MFO assays Mouse
12
1.s4+
0.33
15.90+
1.10
Rat
12
1.42&
0.74
5.39+
0.94
Hamster
12
2.07k
0.30
20.8Ok
1.15
Human-99
6
2.57+
2.17
5.27k
1.85
Human-103
6
1.95+
5.24
1.53k
1.56
Human-105
6
0.92+
0.29
13.00*
1.13
Human-109
6
2.82+
1.47
6.24+
1:39
GST assays Mouse
5
137
Rat
5
N.S.a
Hamster Human-99
5 5
N.S.a 43.8
Human-109
5
44.1
a Data collected
but no acceptable
solution
obtained
+21
118.2
k 14.4
N.S.” N.S.” + 4.5
7.05*
0.44
f
6.04+
0.67
during
8.1 computer
optimization.
GST activity was not detectable in cytosol from hamster lungs (detection limit approx. 0.2 nmol/min per mg protein), but a low level of GST activity was detected in cytosol prepared from human lungs. Cytosol prepared from lung was less active than cytosol prepared from liver in all species studied (Table I). Double reciprocal plots of GST activity were prepared for the four species studied (mouse, Fig.2A; rat, Fig.2B; hamster, Fig.2C; human, Fig. 2D). Examination of these plots suggested that the Km’s were much higher than the K, for MFO activity. The K,,, values obtained by computer optimization of data collected with human and mouse tissues were 44 and 137 mM, respectively (Table II). Computer optimization failed to find a satisfactory ‘solution’ for K, and V,,,,, with enzyme preparations from rat and hamsters, but examination of the double reciprocal plots (Figs. 2B,C) suggested that the K,,, was much higher than 100 mM. DISCUSSION
MFO enzymes
The in vivo values of Vmaxand K,,, for MFO reported by Andersen et al. [l] were converted to rates of metabolism/g liver tissue ( I/max’) by dividing the in vivo I/max
104
0.60
0.00
. 0
40
80
120
160
0
40
80
120
160
1 /Substrate
1 /Substrate
0
0
40
80 t/Substrate
120
160
0
40
80
120
160
1 /Substrate
Fig. 2. Lineweaver-Burk plots (l/velocity versus l/substrate) of the concentration dependence activity in cytosols prepared from various species. Concentrations of CHzCla in the incubation varied from 6.7 mM to 100 mM, and the incubations also contained 8 mM glutathione, and phosphate buffer (pH = 7.4). Plots are shown for (A) mouse cytosol, (B) rat cytosol, (C) cytosol, and (D) human cytosol prepared from liver #99.
of GST mixture 60 mM hamster
for the whole animal by the grams of liver tissue/animal (Table III). Liver sizes for the three species of animals were estimated from historical data collected at the Dow Chemical Company (Table III). In vivo rates of metabolism in liver tissue (assuming that the liver is the primary site of metabolism in the animal) were calculated for a concentration in the linear portion of the velocity versus concentration curve (I FM or 10T6 molar) according to the Michaelis-Menten equation: Rate =
Vmax* Cone
K, + ConC
(1)
and are listed in row 5 of Table III (with units of nmol/g liver per h). This calcuiated rate of in vivo metabolism was divided by the rate of metabolism observed in in vitro
105
TABLE
III
COMPARISON VITRO
ENZYME
OF MFO ACTIVITY
CALCULATED
Rates of in viva metabolism are calculated (mg/h/animal) to V,,,,,l (nmol/g liver/h), equation
FROM
IN VW0
EXPERIMENTS
AND
IN
ASSAYS
with a concentration
from the reference of Andersen et al. [l] by converting V,,, K,,, from mg/l to mmol/l and solving the Michaelis-Menten
of 1 PM. In vitro rates are given as nmol converted/h/mg
microsomal
protein. Mouse Body wt. (g) Liver wt. (g) I’,,,,,1 (nmol/g/h) Km (ILM) Calc (nmol/g/h) Obs (nmol/mg/h) Ratio
(Calc/Obs)
34.5 1.84 6700 4.66 1180 352 3.36
Rat 233.0 5.41 3260 9.08 324 144 2.25
Hamster
Human
140.0
70000
4.34 5560 7.64 644 431 1.49
637 6.83 81.4 31-433 0.2-2.6 1.13
2.37
Mean ratio
2198
k1.04
+ 0.94
experiments at a concentration of 1 mM (a concentration that was also in the nearly linear portion of the in vitro velocity versus substrate curve), and the ratio of the two activities is listed in row 7 of Table III. The ratio of the in vivo and in vitro activities is reasonably consistent across the three animal species (mouse = 3.36, rat = 2.25, hamster = 1.49). This suggests that the in vitro and in vivo results are proportional to each other in spite of the obvious variations in assay conditions (e.g., cofactor concentrations, ionic strength, oxygen tension, substrate concentrations, etc). In vitro measurements of MFO activity in microsomes prepared from human livers revealed considerably more variation than was present in microsomes prepared from the animal species. In addition to the two samples of human liver microsomes reported in Table I, two additional samples of microsomes were assayed: HL-103 and HL-105. One of these (HL-103) had about one-fourth the activity observed in HL-99 and HL-109, and the other (HL-105) had about twice as much activity as HL-99 and HL-109. Therefore, variations in the level of MFO in human tissues spanned almost one order of magnitude. However, the mean ratio of in vivo to in vitro activity in human liver microsomes (1.13 f 1.04) was fairly close to the mean ratio observed in the animal liver microsomes (2.37 f 0.94), and this comparison provides support for the concept of incorporating in vitro enzyme data into the in vivo PB-PK model for CHzClz.
GST enzymes Saturation
of MFO occurred
in vivo during exposure
to concentrations
of 300-500
106
ppm CHzClz [l]. In vitro K, values for MFO activity were in the range of l-2 mM CHZCL (Table II). Since the in vitro K, values for GST in animal tissues were at least two orders of magnitude higher than those for MFO (Table II), it appears unlikely that saturation of GST in vivo would occur at concentrations of CHzCl2 roughly one order
of magnitude
higher
than those causing
saturation
of MFO in vivo. Conse-
quently, the in vitro studies reported here are consistent with the description of GST metabolism as a first order pathway as described by Andersen et al. [l]. The rate of GST reaction (dAM2/dt) in this paper was calculated according to the equation: y
(2)
= (KF) (Conc)(Vol)
where KF= first order rate constant, Cone zconcentration of CHzClz in venous blood leaving the liver, and Vol = size of the liver in liters. This equation is equivalent to a simplified form of the Michaelis-Menten equation: dAM2 dt
( Vmax)(Conc)(Vol) - (K, + Cone)
V,,, K (Conc)(Vol) m
(3)
where I’,,, has units of mg/h per liter of tissue and Cone is small relative to K,. Thus, the constant KF in Eqn. (2) is equivalent to the expression vm,,/K, in Eqn. (3). Assuming that a liter of liver from each species contains roughly equal amounts of cytosolic protein (so that the volume term drops out of the equation), we can rearrange Eqn. (2) to solve for KF: KF= (KI)
dAFoycdt
= (K2) Velocity Substrate
(4)
where KF is equal to a proportionality constant times the rate of the enzyme reaction divided by the substrate concentration (V/S ratio). Eqn. (4) can be written in the same form for either the in vivo reaction or the in vitro reaction; only the proportionality constant is different (KI or Kz). Since the in vivo KF and the in vitro V/S ratio are known with reasonable accuracy for the B6C3Fl mice, the value of KZ can be obtained by dividing KF by the V/S ratio, and then used to calculate the in vivo KF in any species for which an in vitro V/S ratio is known. V/S ratios for GST activity in cytosol prepared from the livers of B6C3Fl mice, F-344 rats, Syrian Golden hamsters, and human livers were determined by dividing the in vitro reaction velocities for GST in Table I (columns 2-6) by the substrate concentration (column 1, Table I). The V/S ratio was essentially constant up to 25 mM in all species, but declined at concentrations of 50 mM or higher in the human liver samples. Consequently, the average V/S ratio for concentrations less than 50 mM in a given species (Table IV, column 1) was used in these calculations. Values of the in vivo rate constant KF calculated for rats, hamsters, and humans, using mouse data as a reference point, are listed in column 2 of Table IV.
107
TABLE
IV
METABOLIC Calculation contains
RATE
of metabolic the velocity
in vitro studies their
PB-PK
calculating specific
CONSTANTS rate constants
to substrate
as outlined model.
of MFO (Al)
model
2 contains
of Andersen
in the rat, hamster,
and human.)
Columns
Calc
Orig
KF
KF
4.01a
1
from the
et al. [l] for
et al. [l] is used as a reference
and GST (A2) in lung and liver in the various
141.4
et al. [l]. Column
the value of KF calculated
3 lists the value of KF cited by Andersen
of KF cited by Andersen
v/s
Mouse
Column
in the text. Column
(The value
the in vivo rate constants
activities
for use in the PB-PK
ratio (V/S).
point
for
4 and 5 show the relative species. Al
A2
4.01”
0.405
0.28
0.63
2.21
0.039
0.14
0.16
1.51
0.068
0.10
0.43s
0.53
0.0014
0.18
k31.1 Rat
116.9 +8.6
Hamster
29.6 + 1.5
Human
105.6 t 12.0
a Used as a reference b Value multiplied
point
for the other
by 0.75 to correct
values.
for the fact that only 3 of the 4 human
liver samples
had detectable
activity.
The estimated value of KF for humans obtained by this procedure (0.43) was very close to that which Andersen et al. [l] obtained by allometric scaling of the mouse data (0.53; Table IV, column 3). The value of KF for rats (0.63) was somewhat lower than that estimated by Andersen et al. [l] in gas uptake experiments (2.21), and the value of KF in hamsters (0.16) was almost an order or magnitude less than the value that Andersen et al. [l] derived from gas uptake (1.5 1). It seems likely that the rather large discrepancy in the hamster constants is related to an inherent limitation of the gas uptake procedure from which the in vivo constant was estimated. The gas uptake technique depends upon measuring overall rates of loss of CHzC12 from the chamber [8]. Consequently, uncertainty in the values of rate constants estimated with this procedure increases when one of the two pathways is much more active than the other. In the mouse, the ratio of GST to MFO activity, measured at substrate concentrations of 6.7 and 5.0 mM CH&lz, respectively, was 0.420 (Table I). Consequently, the gas uptake procedure should give reasonable accuracy in estimating both the GST and MFO rate constants. In contrast, the ratio of GST to MFO activity in
108
hamsters was only 0.013, and large errors in the estimation of in vivo GST rate constants are possible (although estimation of in vivo MFO rate constants by gas uptake should be reliable in this species). Clearly the values for KF estimated from gas uptake studies in mice will be the most reliable, the values estimated from gas uptake studies
in rats sowewhat
will contain
the highest
less reliable,
and the values
for KF in the hamster
studies
levels of uncertainty.
Correlation of in vitro results with animal bioassays High incidences of malignant lung and liver tumors have been observed in B6C3Fl mice exposed to high concentrations of CHzClz vapor [2]. However, increased levels of lung and liver tumors were not seen in F-344 rats and Syrian Golden hamsters exposed to similar levels of CHzClz vapor [2,20], and exposure of B6C3Fl mice to CH2C12 in drinking water also failed to induce a tumorigenic response [3]. Andersen et al. [l] have suggested that the reactive metabolites derived from conjugation of CHzClz with glutathione are likely to be responsible for the tumorigenic activity of CHzClz in the B6C3Fl mouse, and the PB-PK model allows us to calculate levels of these metabolite(s) in the target tissues of the various species in either inhalation or drinking water studies to determine whether there is a correlation between GST activity and tumorigenicity. Under the conditions of the NTP inhalation bioassay, the dose (in mg equiv. of GST metabolites formed per day per liter of tissue) to lung and liver tissues of the B6C3Fl mouse was 482 and 1670, respectively (Table V). Under the conditions of the NCA drinking water bioassay in this same strain, the dose to lung and liver tissues was only 2.2 and 16, respectively. This result correlates well with the observed tumor frequencies in the two bioassays; malignant tumors of lung and liver tissue were observed following inhalation exposure, but no significant changes in lung or liver tumor
TABLE
frequencies
were observed
water study
[2,3].
V
GLUTATHIONE
CONJUGATE
Average
of glutathione
amounts
h per liter of tissue in various week study) ppm,
in the drinking
or drinking
conjugate
(in mg equiv.
species during
water exposures
and doses in the drinking
water
inhalation
(24 h/day). study
of CH2C12) produced
exposures Concentrations
(6 h/day,
in target
5 days/week,
in the inhalation
tissues per 24 96 weeks/l04
exposures
are in
are in m&kg/day.
Species
Dose
Route
Lung
Liver
Mouse
4000=
Inh
482
1670
Mouse
250b
Rat Hamster
Water Inh Inh
4000” 3500a
a Dose in ppm of CH2C12. b Dose in mg/kg/day
in drinking
water.
2.2 96 17.1
16.0 677 I67
109
The dose of GST metabolites exposure
to lung and liver tissue in the rat following
(4000 ppm) to CHZCL was 3-5-fold
lower than in the mouse
inhalation exposed
to
4000 ppm (Table V). The rat failed to show a significant increase in the levels of lung and/or liver tumors, although an increase in the number of benign mammary tumors was noted in this species [2,20]. The dose of GST metabolites to lung and liver tissue in the hamster exposed to 3500 ppm CHzClz was l l-31-fold lower than in the mouse (Table V). Hamsters exposed to CHzClz failed to show significant increases in any type of tumor at any site [20]. Thus there appears to be a reasonable correlation between the in vivo activity of GST in the various species and their sensitivity to CH2C12.
Human risk estimations Quantitative
risk assessments
(QRA)
cannot
provide
precise estimates
of human
risk due to the empirical nature of the models. Furthermore, as a policy decision, QRA procedures have been designed to be protective of public health through inclusion of conservative assumptions wherever uncertainty exists. Consequently, as currently practiced, QRA provides estimates of the ‘plausible upper bounds’ on risk. Acutal risks are unknown, and in many instances may be as low as zero. Nevertheless, QRA plays a useful role in hazard evaluations by providing an objective mechanism for ‘ranking’ materials thought to pose a carcinogenic risk to man. This allows limited societal resources to be directed toward those materials that probably pose the greatest risk to human populations. QRA may also provide a mechanism for definitive action in some cases. For example, if a material is thought to have carcinogenic potential, but QRA suggests that even under the most pessimistic assumptions, the risk to human populations will be less than some very small number (e.g., 10p5), then use of that substance in commerce need not be regulated (i.e., the de minimus approach). Within this framework, the PB-PK model can now be used to formulate human cancer risk estimations. Formulation of the risk estimation involves several steps: (1) Selection of appropriate animal bioassay data. (2) Calculation of the delivered dose to the target organ(s) under conditions that produced tumors in animals. (3) Fitting tumor incidence and delivered dose data to one or more empirical dose-response models for carcinogenesis. (4) Defining human exposure conditions of interest and calculating the delivered dose to corresponding organs of humans under those particular conditions. (5) Estimating upper bounds on human risk from the doseresponse model previously fitted to animal bioassay/target organ delivered dose data. This process differs from that used by default when pharmacokinetic data on delivered dose are not available. In the absence of pharmacokinetic data, delivered dose is considered to be proportional to nominal dose throughout the entire exposure range. Extrapolation from animal to human is performed by assuming that equipotent doses are expressed in mg of chemical per unit of surface area in the
110
various species. The latter process always assumes that humans are more sensitive to a given dose of test chemical (in mg/kg per day) than rodents because of the lower surface area to volume ratio (humans are assumed to be 5-6-fold more sensitive than rats and 13-fold more sensitive than mice). Following the default procedure for CH2Cl2 (without use of PB-PK data), Singh et al. [21] estimated that the lifetime risk for continuous exposure to a concentration of 1 pg/m3 CHzCl2 in inhaled air was 4.1 x 10m6 *. This estimate was presented as a ‘plausible upper limit’ on risk, along with the qualifier that ‘. ..the true value is unknown and may be as low as zero’. The uncertainty and upper limit aspect of this assessment reflect the fact that a series of conservative assumptions were made in the face of scientific uncertainty, including uncertainty as to the magnitude of the delivered dose. Development of a unified PB-PK model for CH2Cl2 permitted the reduction of uncertainty in the original risk estimation through replacement of default assumptions with results from the PB-PK model [l]. Delivered doses for the female B6C3Fl mice in the drinking water and inhalation chronic bioassays [2,3] were calculated with the PB-PK model of Andersen et al. [l] and are summarized in Table VI. Model parameters used by Andersen et al. [l] were employed for these calculations with the exception of KF (the metabolic rate constant for the GST pathway) and the proportionality constants Al and A2 (which give the relative levels of GST in lung and liver tissue). Al and A2 were calculated from the data presented in Table I of this report, and KF was calculated as outlined earlier (Table IV). Tumor incidences in treated and control groups of female mice in these studies also are summarized in Table VI [2,3]. (In the interest of brevity, data from male mice are not presented since these gave estimates of risk similar to those derived from female mice.) Once delivered dose had been calculated, it was necessary to choose a doseresponse model to describe the animal responses. The choice of the appropriate high dose to low dose model has been the subject of considerable discussion (e.g., see Brown [23]). Some models, including the linearized multi-stage (LMS) model, impose conservative assumptions on the QRA process (in this case by forcing the model to be linear at low doses). Other models such as the Probit model make fewer assumptions about the shape of the dose-response curve. The LMS has been widely used by regulatory agencies because of the feeling that the estimates produced by this model are very unlikely to underpredict the true risk. However, current regulatory practice generally includes presentation of a range of risks estimated from other models as well. Consequently, the delivered dose data were fit to four different dose-response models currently used by the toxicological community. The models chosen were (1)
* 4.1 x 10e6 was the original subsequently
been lowered
‘unit risk’ for CHzCll based
in the EPA document
on use of PB-PK
modeling
[22].
of 1985, but this estimate
has
111
TABLE
VI
INTERNATIONAL The ‘internal
DOSE AND TUMOR
dose’ calculated
by the GST pathway of 6 h/day,
INCIDENCES
by the PB-PK
model is in average
per day per liter volume
5 days/week,
corrected
for the fraction
Values reported
for humans
for the fraction
of the year (48152) that humans
Nominal
dose
mg equivalents
of tissue. The values reported of a lifetime
are for one year of occupational
(96/102)
exposure,
of CHzCls
metabolized
for mice are for exposures that animals
6 h/day,
were exposed.
5 days/week,
corrected
are in the workplace.
PB-PK
PB-PK
Lung
Liver
(lung)
(liver)
tumors
tumors
Mouse: Drinking 0
water
mg/kg
0.0
60 mg/kg
0.395
125 mg/kg
0.902
185 mg/kg
1.46
250 mg/kg
2.19
0.0
3.00 6.71 10.8 16.0
5/100
6/100
3/100
4/99
l/50
2/50
3/50
5/50
4/50
3/50
Inhalation 0.0
0 ppm 2000 ppm
0.0
321 482
4000 ppm
3150
3/50
785
30/48
16/48
1670
41/48
40/48
Human: Occupational
exposure
1 mm 2 mm
5.36 x lO-5 1.07x
1o-4
1.73 x 1o-4
5 ppm
2.69x
1O-4
4.38 x 1O-4
8.61 x 10-s
10 ppm
5.40 x 1om4
8.94x
10m4
20 ppm
1.09 x 10-s
1.86x
1o-3
50 ppm
2.80 x lo-’
5.34x
1o-3
5.98x
1.37 x 10-z
100 ppm Continuous
inhalation
1 wg/m’ Drinking
1O-3
6.16x 10-6
9.87 x 1O-6
4.26 x lo-’
3.66 x 1O-6
_
water
1 pg/l
the Linearized Multi-Stage (LMS), (2) the Probit model, (3) the Weibull model, and (4) the Logit model. All of the models predict excess risk (ExRisk) as a function of administered dose, except the Probit model which predicts the probit of the excess risk. ExRisk is defined as the excess lifetime risk resulting from the independent effect of the chemical exposure in question:
112 TABLE
VII
CONSTANTS Constants
obtained
Linearized
Multi-Stage
female
when
computerized
(LMS), Logit,
optimizations Weibull,
mice (given in Table VI). These constants
excess risk for a given PB-PK
were used to fit four
and Probit
- to the PB-PK
dose-response
dose and tumor
can be used with the equations
models incidence
for
in the text to estimate
dose.
Independent
Background
B
A
background LMS-Lung
0.0453
LMS-Liver
0.0550
Logit-Lung
0.0473
- 8.788
1.696
Logit-Liver
0.0576
- 22.258
3.207
Weibull-Lung
0.0434
- 5.072
0.922
Weibull-Liver
0.0576
- 15.273
2.132
Probit-Lung
0.0475
- 5.303
1.024
Probit-Liver
0.0576
- 13.555
1.952
ExRisk
= (TumFr
0.00495
0.00685
- BckGrnd) (5)
(1 .O - BckGrnd)
In this equation, TumFr is the tumor frequency observed in the treated groups and BckGrnd is the tumor frequency observed in control groups. The mathematical forms of the four dose-response models are given for the LMS model [Eqn.(6)]*, the Logit model [Eqn. (7)], the Weibull model [Eqn. (S)], and the Probit model [Eqn. (9)l. ExRisk
= A * Dose
ExRisk
= ~
ExRisk
= 1 - EXP [-A
Probit
(6)
~~~~.. 1 1 + EXP(-A -B*LN(Dose)
(7)
* (Dose)B]
(8)
= A + B * LN (Dose)
(9)
Constants for these equations were derived by computer optimization of the tumor incidences and PB-PK doses from Table VI. All of the treated and control female mice from the two studies (500 animals) were used in the analysis. The values of the constants obtained from these optimizations are listed in Table VII. The models can now be used to estimate the risk for any exposure condition within
* The LMS model fidence
is first fit in the form
level on the linear
of an exponential
term is calculated
polynomial,
to give the low dose form
and then the upper shown
in Eqn.
(6).
95% con-
113
the scope of the PB-PK
model.
To illustrate
this, 3 specific exposure
scenarios
were
evaluated: (1) lifetime exposure to 1 pg/m3 CHzClz in the atmosphere (EPA’s standard Unit Risk condition), (2) lifetime consumption of 2 l/day of water containing 1 ,ug CH2C12/1, and (3) one year’s occupational exposure to CHzClz at the proposed Threshold Limit Value (TLV, 50 ppm). The PB-PK doses equivalent to these exposures are listed in Table VI. Low dose risk estimates obtained with the Logit, Weibull, and Probit models may vary depending upon assumptions made about the additivity or independence of the background tumors. Since the LMS model assumes an additive background, this consideration does not apply to this model, The question of whether to use independent or additive backgrounds is a difficult one, and is beyond the scope of this paper. However, the effect of incorporating pharmacokinetic data into the risk estimation can be visualized without resolving this question. The LMS model, which assumes an additive background, generally gave the highest estimates of low dose risk for a given dose of CH2C12 (Table VIII). Low dose risk estimates from the other three models were similar to the risk estimates from the LMS when additive backgrounds were assumed (data not shown). However, when independent backgrounds were assumed, the 4 models give very different predictions of low dose risk (shown in Table VIII). Although each model gave an excellent description of the observable data from the two bioassays, the models differed by 5 to 15 orders of magnitude in predicted risk for the exposure conditions depicted in Table VIII. The effect of incorporating pharmacokinetic principles into the risk assessment process may be visualized by comparing the total risk (lung + liver) predicted by the LMS model. Under conditions of continuous inhalation of 1 pg/m3, the LMS predicted an upper bound risk of 4.1 x 10e6 for lifetime exposure [21] when administered dose was used instead of delivered dose. However, use of a pharmacokinetically calculated delivered dose predicted a risk more than two orders of magnitude lower: 3.7 x lo-’ (Table VIII). There are two components of this difference. The first is the non-linearity in the activity of the metabolic pathways. The primary metabolic pathway (MFO) becomes saturated at high exposure concentrations of CHX12, and a disproportionate increase in the activity of the secondary toxic pathway (GST) occurs. The second component has to do with calculation of species-specific doses rather than generic surface area ‘corrections’. Humans possess relatively lower levels of the enzyme(s) responsible for CHzClz toxicity, and at equivalent tissue concentrations of CHZCIZ will generate lower levels of the toxic GST metabolite. At the same time, it needs to be pointed out that use of a PB-PK model in QRA does not always result in estimation of lower risks for humans than the default procedures mentioned earlier. For instance, if a test chemical were directly toxic and were inactivated by metabolic enzymes, then a PB-PK model would predict more risk to humans than rodents because of the lower levels of detoxifying enzymes in
114
TABLE
VIII
EXCESS RISKS EXPOSURE
FOR INHALATION,
Excess risks calculated tinuous
inhalation
WATER
by four dose-response
models
of 1 @g/m’ in air for a lifetime,
1 @g/liter for a lifetime,
CONSUMPTION
and (3) occupational
are structured
with the non-additive
under three specific
(2) consumption
exposure
at 50 ppm for one year. The four models are: Linearized
AND OCCUPATIONAL
(6 h/day,
exposure
5 days/week,
Multi-Stage,
conditions:
of 2 liters of water/day Logit,
48 weeks/year) Weibull,
(1) concontaining to CH&Zlz
and Probit.
Models
background. Inhai,
Water
occup.
1 pgg/m3
I pglliter
50 ppm
LMS (w/P-K)
3.0x
10-x
2.1 x 10-9
1.4x 10-S
Logit
2.1 x lo-‘3
2.3 x lo-‘5
7.0x
Weibull
9.8 x IO-*
8.4 x 1o-9
2.8 x 10-5
Risk of lung tumor 10-9
Probit
< lo-‘5
< lo-‘5
< IO-i5
Geo. mean
2.8 x io-”
2.5 x lo-i2
7.2 x 10-9
Risk of liver tumor LMS (w/P-K)
6.8 x IO-’
2.5 x 1O-9
3.7 x 10-e
Logit
2.1 x 10-26
8.8 x 1O-28
1.2x 10-i’
Weibull
< lo-i7
Probit
< lo-‘5
< 10-15
< lo-‘5
Geo. mean
3.5 x 10-17
1.2x 10-l’
2.6 x IO- i4
Total
risk (lung
LMS (w/o
f
liver)
P-K)’
4.1 x 10-s
LMS (w/P-K)
3.7x
4.6 x lO-9
1.8x10-5
Logit Weibull
2.1 x lo-”
10-S
2.3 x lo-is
7.0x
9.8 x 10-8
8.4 x 10-a
2.8 x 10-S
10-9
Probit
< lo-‘”
< lo-i5
< 1o-‘5
Geo. mean
2.9 x 10-i’
3.1 x10-i*
7.7 x 10-9
’ Singh et al. [21] estimated risk of developing
the risk of developing
only a lung tumor
either a lung or liver tumor,
but did not estimate
-the
or only a liver tumor.
this species. Similarly, if toxicity resulted from reactive metabolitets) produced by a saturable process, then low dose risk extrapolated from high, saturating doses by a PB-PK model could be more than predicted by linear extrapolations. In summary, development of a PB-PK for CHA&, and subsequent incorporation of in vitro enzyme studies, has provided a valuable tool for improving risk estimations for CH$&. The model permitted identification of a probable mechanism of carcinogenesis for CH2C12, provided a consistent explanation for the differences between chronic bioassays of CH2CL in different species and by different routes, and provided a quantitative basis for evaluating the effect of two competing metabolic pathways upon the production of the putative toxic metabolite in situ at various
115
doses. We believe that risk estimations that incorporate these principles will prove significantly more reliable than risk estimations that do not consider pharmacokinetics. ACKNOWLEDGEMENTS
The authors would like to thank Dr. Daniel Krewski (Health Protection Branch, Canada) and Dr. Murray Cohn (Consumer Products Safety Commission) for their expert assistance in fitting the PB-PK delivered dose data to the various cancer doseresponse models. REFERENCES Andersen,
M.E.,
ly based
Clewell, H.J.,
pharmacokinetics
Pharm.
87, 185-205.
National
Toxicology
III, Gargas,
Program
of Dichloromethane
TR-306
(board
draft).
Serota,
D., Ulland,
Caster,
March
J., Simon,
determined
process
for methylene
chloride.
Toxicol.
on the Toxicology Mice (Inhalation
Chronic
8-9, Bethesda,
Appl.
and CarcinoStudies).
NTP-
Oral Study in Mice. Food Solvents
MD.
A.B. and Armstrong,
by dissection
Report
Rats and B6C3Fl
F. (1984) Hazelton
Chloride,
Poncelet,
values
in F-344/N
B. and Carlborg,
I: Methylene
W.O.,
Normal
Smith, F.A. and Reitz, R.H. (1987) Physiological-
(NTP) (1985) NTP Technical
genesis Studies
Workshop
M.L.,
and the risk assessment
W.D.
and chemical
(1956) Tissue weights
methods.
Proc.
Sot.
Exp.
of the rat. I.
Biol.
Med.
91,
model
of
122-126. Davis,
N.R.
and Mapleson,
W.W.
of injected
agents
the distribution International Man.
Commission
In: W.S.
Snyder,
ICRP
Publication
(Eds.),
on Radiation M.J.
Gargas,
blood,
Anders,
dihalomethanes 10 Ahmed,
Protection
E.S. Nasset, Snyder
T. (1979) Partition
metabolism
V.L.,
and quantification
anaesthetics.
(1975) Report L.R.
coefficients
53, 399-404.
of the task group
Karhausen,
et al., Eds.),
of a physiological
Br. J. Anaesthesiol. G.P.
Pergamon
Howells
Press,
of some aromatic
on Reference
and I.H.
Tipton
New York. hydrocarbons
and ketones
and oil. Brit. J. Ind. Med. 36, 231-234.
M.L. and Andersen,
of chemical Kubic,
Cook,
23 (W.S.
Sato, A. and Nakajima, in water,
(1981) Structure and inhaled
M.E. (1988) A gas phase technique
in the rat. Toxicologist M.W.,
to carbon
Engel,
A.E.
and Anders,
M.W.
organic
halide.
I. In vitro
studies.
Gargas,
M.L.,
Clewell,
viva: differentiation
R.R.,
monoxide.
H.J.
Barlow,
(1976) Metabolism Metab.
and Andersen,
of kinetic
C.H.
I. In vivo studies. Drug
constants
for determining
the kinetic constants
8, 112. and Caughey,
W.S. (1974) Metabolism
Drug Metab.
Dispos.
of dihalomethanes
Dispos.
of
2, 53-57.
to formaldehyde
and in-
4, 357-361.
M.E. (1986) Metabolism
for two independent
of inhaled
pathways.
dihalomethanes
Toxicol.
Appl.
in
Pharmacol.
82, 21 l-223. Lindstedt, Council
S.L.
(1987) Allometry:
of the National
Academy
and Health,
Vol. 8, National
Beaune,
Kremers,
parison
P.,
microsomes. Backer,
R.H.
Drug Metab. and
Academy
P.G.,
of monooxygenase
body
size constraints
of Sciences, Press,
Kaminsky, activities Dispos.
Guengerich,
L.W.,
evidence
Washington,
dosing.
In: National
in Risk Assessment,
Research
Drinking
Water
DC, pp. 65-79.
DeGraeve,
and cytochrome
L. and Guengerich,
P-450 isozyme
F.P.
concentrations
(1986) Comin human
liver
14, 437-442. F.P.
(1986)
methyl-3,5-bis(alkoxycarbonyl)-1,4-dihydropyridines chemical
in animal
Pharmacokinetic
for the involvement
Oxidation by
of 4-arylhuman
of a form of cytochrome
and liver
4-alkyl-substituted microsomes
and
2,6-diimmuno-
P-450. J. Med. Chem. 29, 1596-1603.
116
15 Hall, S.D., Guengerich, tion of mephenytoin 216-222. 16 Knodell,
R.G.,
tolbutamide:
F.P.,
Branch,
Hall,
S.D.,
in human
Wilkinson,
in viva characterization
tion and relationship 17 Guengerich,
F.P.,
Characterization oxidation,
R.A. and Wilkinson,
4-hydroxylase
G.R.
M.V.,
Beaune,
of rat and human
a prototype
and Guengerich,
for genetic
J. Pharmacol. P.H.,
Kremers,
liver microsomal polymorphism
J. Pharmacol. F.P.
of the form of cytochrome
to in vivo disposition. Martin,
G.R. (1987) Characterization
liver microsomes.
(1987) Hepatic
P-450 involved
Exp. Ther. P., Wolff,
cytochrome
in oxidative
and inhibi-
Exp.
Ther.
240,
metabolism
of
in methyl hydroxyla-
241, 1112-1119. T. and Waxman,
P-450 forms
D.J. (1986)
involved
drug metabolism.
in nifedipine
J. Biol. Chem.
261,
505 l-5060. 18 Guengerich,
F.P.,
cytochrome
Muller-Enoch,
P-450.
19 Lineweaver,
H. and Burk,
Chem.
Sot.
56, 658-666.
20 Burek,
J.D.,
Nitschke,
Rampy,
study D.V.,
K.D.,
Bell, T. J., Wackerle,
for dichloromethane
H.L.
Fundam.
and White,
(methylene
P.D.
chloride).
of enzyme
D.L.,
M.J. (1984) Methylene
in rats and hamsters. Spitzer,
I.A. (1986) Oxidation
of quinidine
by human
liver
30, 287-295.
D. (1934) The determination
L.W. and McKenna,
genicity 21 Singh,
D. and Blair,
Mol. Pharmacol.
App.
Childs,
R.C.,
constants.
Beyer, J.D.,
chloride:
A 2-year inhalation
Toxicol.
4, 30-47.
(1985) Addendum
Updated
dissociation
to the health
carcinogenicity
assessment
J. Amer.
Dittenber, toxicity
assessment
D.A.,
and oncodocument
of dichloromethane.
EPA/600&82/004F. 22 Blancato,
J.N.,
addendum epidemiology. 23 Brown,
Hopkins,
J. and Rhomberg,
for dichloromethane C.C.
External
Review
(1983) Learning
(methylene Draft, about
L. (1987) Update chloride):
to the health assessment
pharmacokinetics,
mechanism
document
and
of action,
and
EPA/600/8-87/030A. toxicity
in humans
from
studies
on animals.
Chemtech
13, 6.
97
Elsevier
TXL 02034
Incorporation of in vitro enzyme data into the physiologically-based pharmacokinetic (PB-PK) model for methylene chloride: implications for risk assessment R.H. Reitz’, A.L. Mendrala’, F.P. Guengerich3
C.N. Park’, M.E. Andersen2 and
‘Dow Chemical Company, Health and Environmental Sciences, Midland, MI (U.S.A.), ‘Wright Patterson Air Force Base, Armstrong Aerospace Medical Research Laboratory, Dayton, OH (U.S.A.), ‘Vanderbilt University, Department of Biochemistry, NashviNe, TN (U.S.A.) (Received
19 April,
(Revision
received
(Accepted
6 June,
1988) 1 June,
1988)
1988)
Key words: PB-PK;
Pharmacokinetics;
Methylene
chloride;
Risk assessment
SUMMARY Physiologically-based tainty inherent for incorporating Methylene
pharmacokinetic
in extrapolating data
chloride
(PB-PK)
models provide
a mechanism
for reducing
the results of animal toxicity tests to man. This paper discusses
from
in vitro
studies
is used as an example,
of xenobiotic
and carcinogenic
metabolism risk estimates
the uncera technique
into in vivo PB-PK incorporating
models.
PB-PK principles
are presented.
INTRODUCTION
Human beings are exposed to many chemicals, natural and synthetic. To assess whether those chemicals are likely to produce adverse effects in human populations, toxicity tests are conducted in laboratory animals. Inherent in this procedure is the assumption that laboratory animals are reasonable surrogates for humans; that is, Address Chemical
for correspondence: Company,
Abbreviations:
PB-PK,
glutathione-S-transferase; threshold
Midland,
limit value.
R.H.
Reitz,
Health
and Environmental
Sciences,
1803 Building,
Dow
MI 48674, U.S.A.
physiologically-based QRA,
quantitative
pharmacokinetics;
MFO, mixed function
risk
LMS,
assessment;
linearized
oxidase;
multi-stage;
GST, TLV,
98
most materials that fail to produce adverse effects in animals will not produce adverse effects in human populations. However, significant differences usually exist between the conditions prevailing in laboratory tests and those encountered by humans. Animals
used in laboratory
tests come from carefully
bred homogeneous
strains
and are carefully screened to be free of disease. At the start of the test, all the animals are young adults. For practical reasons, small groups are used, and a limited number of doses are evaluated. Doses are usually high, and chemicals are administered by the most experimentally convenient route. In contrast, human populations are heterogeneous and contain healthy and diseased individuals of all ages. Human exposures typically involve doses much lower than those employed in the animal tests and frequently occur through one or more routes different than the animal studies. Therefore, a variety of extrapolations are necessary in order to estimate human hazard from animal studies. Each extrapolation introduces uncertainty into the final risk assessment. Physiologically-based pharmacokinetic (PB-PK) models are useful for reducing some of this uncertainty because they take into account actual physiological properties of the various species, compound-specific partitioning of the chemical between tissues and blood, and concentration-dependent rates of metabolism. PB-PK models are particularly useful for high dose/low dose extrapolations, dose route extrapolations, and interspecies extrapolations. We have reported a procedure for incorporating the techniques of PB-PK modeling into the risk assessment for CHzClz [l]. A PB-PK model was developed providing quantitative descriptions of the rates of metabolism and levels of CHzClz in various organs of four mammalian species (mouse, hamster, rat, and human). The model was validated by comparing computer simulations with experimental data gathered in mice, rats, and humans. Levels of reactive metabolites produced in target tissues were correlated with tumor incidences in two chronic bioassays of CH2C12 in the B6C3Fl mouse [2,3]. Finally, relevant measures of tissue dose were calculated for use in a pharmacokinetically-based risk assessment. Development of this PB-PK model required the acquisition of 3 types of model parameters: (1) physiological constants, (2) partition coefficients, and (3) metabolic rate constants. Physiological parameters for the various species were available from the scientific literature [4-61, and partition coefficients were determined experimentally at Wright Patterson Air Force Base (WPAFB) according to the procedures of Sato and Nakajima [7]. In vivo gas uptake studies conducted at WPAFB according to the method of Gargas et al. [8] were used to estimate metabolic rate constants for two pathways of CH2C12 metabolism in animals [ 11. These pathways involve oxidation of CHzClz by cytochrome P450-dependent enzymes (mixed function oxidase, MFO) [9] and conjugation of CH2C12 with glutathione by glutathione-S-transferase enzymes (GST) [lo]. Metabolic rate constants for the MFO pathway in humans were derived from
99
measurements of carboxyhemoglobin, exhaled carbon monoxide, and CHzClz uptake in human volunteers exposed to either 100 or 350 ppm of CH2Cl2 at the Dow Chemical Company [ 11. However, there were no in vivo data available for direct calculation of the rate constants for the GST pathway in humans. Since MFO enzymes have a higher affinity for CH2C12 than the GST enzymes [ 111, most of the metabolism of CHzClz occurs through MFO at low concentrations. Therefore, the human studies of CHzClz metabolism (which all used relatively low concentrations of CHX12) are not appropriate for estimating the rate of metabolism by the GST pathway, Consequently, the rate constants for the GST pathway were estimated by allometric scaling of the in vivo rate constants for GST metabolism in animals [ 11. Although allometric scaling has been used by others for estimating metabolic rates in different species [12], it was clear that data supporting the validity of such scaling for CHzCl2 would be valuable. Additional experiments to estimate the value of the metabolic rate constants for the GST pathway in humans have now been conducted. These experiments involved in vitro measurements of the glutathione-dependent metabolism of CH2C12 in tissue samples from humans as well as tissue samples from F-344 rats, B6C3Fl mice, and Syrian Golden hamsters. Similar procedures were employed to make in vitro measurements of the rate of oxidation of CHzClz by the MFO enzymes, and the reliability of the in vitro to in vivo extrapolation was assessed by comparing observed in vivo rates of MFO metabolism with those predicted from the in vitro data. This paper will describe the methods employed to collect the in vitro data and show how they may be incorporated into PB-PK risk assessments for CH2C12. METHODS
Test materials Spectroscopic grade CH2Cl2 (< 99.9% purity) was obtained from Fisher Chemical Co. CH~~~c12(0.5 Ci/mol) was synthesized by New England Nuclear (Boston, MA). This material was checked for radiochemical purity by gas chromatography/mass spectrometry prior to use and was found to be 95.0% CHZ~%J~, with small amounts of CH336C1 (0.6%) and CH3%Z13(2.8%) identified as minor impurities. Preparation of enzymes Cytosolic and microsomal fractions of lungs and livers were prepared from male F-344 rats, B6C3Fl mice, Syrian Golden hamsters, and tissues of otherwise healthy human accident victims selected for organ transplantation through the Nashville Regional Organ Procurement Agency, Nashville, TN. Tissues were removed from animals or humans and small pieces were rapidly frozen either by immersion in liquid nitrogen (humans) or with dry ice (animals). Soluble enzymes (cytosol) and particulate enzymes (microsomes) were prepared as described elsewhere [ 13-181.
100
Samples of human liver from four different individuals were processed individually. Availability of human lung samples was limited, so lung samples from two individuals were pooled ( - 8 g total tissue weight) for the enzyme preparations from human lung. (None of the lung samples were from the same individuals as the liver samples). Assay of A4FO Incubations (30 min, 37°C) were conducted in 5.0 ml vials sealed with Teflon@coated rubber septa (4.0 ml headspace). Vials contained an NADPH regenerating system in addition to phosphate buffer (pH 7.4) and CHzClz. At the end of incubation, the contents of the vial were transferred to a glass tube containing 0.5 ml of a solution of 2% NaX03 and 1% NaCl as well as 3.0 ml of unlabeled CHZCIZ, and vortexed vigorously for 10 sec. Centrifugation served to separate the unreacted substrate (in the organic layer) from the water-soluble 36C1- formed by enzyme reaction. Assay of GST Incubations (30 min, 37°C) were conducted in 1.8 ml glass vials sealed with Teflon@-coated rubber septa (< 0.1 ml headspace in vial). Each incubation contained (in addition to CHzC12) potassium phosphate buffer (So-100 mM, pH 7.4) and reduced glutathione (10 mM). 36C1_ was separated from the unreacted CHzC12 as outlined for the MFO assay. RESULTS
A4FO enzymes
A series of incubations
with microsomes
prepared
from liver were conducted
at
CHzClz concentrations varying from 1 mM to 10 mM. The reaction rates observed (nmol/min per mg protein) are listed in Table I. The highest rates of oxidation were observed in the hamster liver microsomes, with slightly lower rates in the mouse liver microsomes. Lower rates were observed in rat and human liver microsomes (Table I). Microcomes prepared from lung tissues in the various species were assayed at a single substrate concentration: 5 mM CHzClz. Microsomes from mouse lungs were much more active than microsomes from the lungs of the other species (Table I). No MFO activity was detected in pooled samples of human lung microsomes, with a detection limit of approximately 0.1 nmol/min per mg protein. Double reciprocal plots (l/velocity versus l/substrate; Lineweaver and Burk [19]) of the results obtained with liver microsomes are presented in Fig. 1A (B6C3Fl mice), Fig. 1B (F-344 rats), Fig. 1C (Syrian Golden hamsters), and Fig. 1D (human sample HL-99). Kinetic data for three other human liver samples were analyzed but are not shown.
101
TABLE
I
REACTION Enzyme
RATE
activities
(nmol/min/mg
(GST, and MFO) observed
from liver tissue of male B6C3Fl reported Cont.
protein)
as nmol product
formed/min/mg
Mouse
(mM)
at various
concentrations
mice, F-344 rats, Syrian
Golden
of CHzClz with enzymes prepared
hamsters,
and humans.
All values are
protein. Rat
Hamster
Human
Human
(HL-99)
(HL-109)
MFO assays: Liver 1.00
5.87a
2.40=
7.18=
1.49
1.66”
1.23
6.52a
2.42a
8.16=
1.85
2.06a 2.26a
1.64
7.18”
2.93”
8.54a
2.09
2.60
9.02”
3.44a
11 .02a
2.28
2.87a
5.00
11.40a
4.10a
14.47a
3.34
4.46”
10.00
14.40”
4.91a
18.18a
4.58
4.80a
4.62
0.16
0.99
6.1
4.17
0.75
0.182
10.0
7.24
1.11
0.307
Lung 5.00
GST assays: Liver
1.36
0.383
1.53
16.7
12.3
2.08
0.482
1.95
1.66
25.0
18.5
3.19
0.76
2.58
2.24
9.92
12.5
50.0
33.2
100.0
48.6
12.1
6.17
1.3
1 .o
1.28
1.24
3.95
3.51
2.64
4.74
3.94
0.0
0.37
Lung 40.0
a Mean value from two independent b Pooled
human
lung samples,
experiments.
no activity
observed
at detection
limit of 0.05-O. 1 nmol/min/mg
protein.
Values of Km (the affinity constant) and V,,, (the maximum rate of the enzyme reaction) were determined by computer optimization (method of least squares) of the data listed in Table I. The results (+ standard deviation) are listed in Table II. The values of Km were almost identical in the 7 sets of liver microsomes analyzed, ranging from 0.92 mM (Human-105) to 2.8 mM (Human-log). However, the values of V,,, in the various species differed significantly. The highest values of V,,, were observed in microsomes from hamster liver ( Vmax=20.8), with lower rates in mice (V,,, = 15.9), rats (V,,, = 5.39), and humans (range= 1.5-13.0, mean V,,,=6.5). GST enzymes
Assays of GST in cytosol prepared from the livers of the various species were con-
102
ducted at CH2C12 concentrations from 6.7 mM to 100 mM. (The limited solubility of CH2Cl2 in water prevented testing concentrations higher than 100 mM). Enzymes prepared from mouse liver contained the highest levels of the GST enzyme (Table I). Tissues from the rat liver contained intermediate levels of GST, while tissues from the hamster and human livers contained the lowest levels of GST (Table I). Cytosol preparations from the lungs of the different species were assayed at a single substrate concentration: 40 mM CHzClz. The highest levels of GST activity in lung tissue were observed in the mouse. The rat lung had intermediate activity, and the lowest activity was observed in cytosol prepared from human and hamster lungs.
0.50
0.20
I
0.40 0.15 E ‘5 0 5 z
0.30 0.10 0.20 0.05
0.10
0.00
0.00 200
0
P .z :: 5 c
400 600 1 /Substrate
800
0
1OOC
0.20
0.8
0.15
0.6
0.10
0.4
0.05
0.2
f
0.00
200
0
400
600
800
I 1000
200
activity
in microsomes
ture varied
from
plots (l/velocity prepared
versus
from various
1 mM to 10 mM, and the incubations
800
1000
I /Substrate
0 0
l/substrate)
species.
600
D
200
400 600 1 /Substrate
1 /Substrate
Fig. 1. Lineweaver-Burk
I
400
of the concentration
Concentrations
800
1000
dependence
of MFO
of CHsClz in the incubation
also contained
50 mmol potassium
mix-
phosphate
- 1 unit of glucose-6-phosphate debuffer (pH 7.4). 10 mmol glucose 6-phosphate, 0.5 mmol NADP+, hydrogenase, CHZ%Z, and l-2 mg of microsomal protein in a total volume of 1 .O ml. Plots are shown for (A) mouse
microsomes,
(B) rat microsomes, prepared
(C) hamster microsomes, from liver #99.
and (D) human
microsomes
103
TABLE
II
KINETIC
CONSTANTS
Kinetic constants
obtained
from in vitro experiments
from the livers of male B6C3Fl obtained
by relative
the best estimate
least squares f
Species
with GST and MFO assays in enzyme preparations
mice, F-344 rats, Syrian Golden weighting
SD of the estimate
hamsters,
using the SimuSolv@ computer
as calculated
by the computer
Km (mM)
n
and humans. program.
Constants
were
Values reported
are
program. If,,,,, (nmol product formed/min/mg
prot.)
MFO assays Mouse
12
1.s4+
0.33
15.90+
1.10
Rat
12
1.42&
0.74
5.39+
0.94
Hamster
12
2.07k
0.30
20.8Ok
1.15
Human-99
6
2.57+
2.17
5.27k
1.85
Human-103
6
1.95+
5.24
1.53k
1.56
Human-105
6
0.92+
0.29
13.00*
1.13
Human-109
6
2.82+
1.47
6.24+
1:39
GST assays Mouse
5
137
Rat
5
N.S.a
Hamster Human-99
5 5
N.S.a 43.8
Human-109
5
44.1
a Data collected
but no acceptable
solution
obtained
+21
118.2
k 14.4
N.S.” N.S.” + 4.5
7.05*
0.44
f
6.04+
0.67
during
8.1 computer
optimization.
GST activity was not detectable in cytosol from hamster lungs (detection limit approx. 0.2 nmol/min per mg protein), but a low level of GST activity was detected in cytosol prepared from human lungs. Cytosol prepared from lung was less active than cytosol prepared from liver in all species studied (Table I). Double reciprocal plots of GST activity were prepared for the four species studied (mouse, Fig.2A; rat, Fig.2B; hamster, Fig.2C; human, Fig. 2D). Examination of these plots suggested that the Km’s were much higher than the K, for MFO activity. The K,,, values obtained by computer optimization of data collected with human and mouse tissues were 44 and 137 mM, respectively (Table II). Computer optimization failed to find a satisfactory ‘solution’ for K, and V,,,,, with enzyme preparations from rat and hamsters, but examination of the double reciprocal plots (Figs. 2B,C) suggested that the K,,, was much higher than 100 mM. DISCUSSION
MFO enzymes
The in vivo values of Vmaxand K,,, for MFO reported by Andersen et al. [l] were converted to rates of metabolism/g liver tissue ( I/max’) by dividing the in vivo I/max
104
0.60
0.00
. 0
40
80
120
160
0
40
80
120
160
1 /Substrate
1 /Substrate
0
0
40
80 t/Substrate
120
160
0
40
80
120
160
1 /Substrate
Fig. 2. Lineweaver-Burk plots (l/velocity versus l/substrate) of the concentration dependence activity in cytosols prepared from various species. Concentrations of CHzCla in the incubation varied from 6.7 mM to 100 mM, and the incubations also contained 8 mM glutathione, and phosphate buffer (pH = 7.4). Plots are shown for (A) mouse cytosol, (B) rat cytosol, (C) cytosol, and (D) human cytosol prepared from liver #99.
of GST mixture 60 mM hamster
for the whole animal by the grams of liver tissue/animal (Table III). Liver sizes for the three species of animals were estimated from historical data collected at the Dow Chemical Company (Table III). In vivo rates of metabolism in liver tissue (assuming that the liver is the primary site of metabolism in the animal) were calculated for a concentration in the linear portion of the velocity versus concentration curve (I FM or 10T6 molar) according to the Michaelis-Menten equation: Rate =
Vmax* Cone
K, + ConC
(1)
and are listed in row 5 of Table III (with units of nmol/g liver per h). This calcuiated rate of in vivo metabolism was divided by the rate of metabolism observed in in vitro
105
TABLE
III
COMPARISON VITRO
ENZYME
OF MFO ACTIVITY
CALCULATED
Rates of in viva metabolism are calculated (mg/h/animal) to V,,,,,l (nmol/g liver/h), equation
FROM
IN VW0
EXPERIMENTS
AND
IN
ASSAYS
with a concentration
from the reference of Andersen et al. [l] by converting V,,, K,,, from mg/l to mmol/l and solving the Michaelis-Menten
of 1 PM. In vitro rates are given as nmol converted/h/mg
microsomal
protein. Mouse Body wt. (g) Liver wt. (g) I’,,,,,1 (nmol/g/h) Km (ILM) Calc (nmol/g/h) Obs (nmol/mg/h) Ratio
(Calc/Obs)
34.5 1.84 6700 4.66 1180 352 3.36
Rat 233.0 5.41 3260 9.08 324 144 2.25
Hamster
Human
140.0
70000
4.34 5560 7.64 644 431 1.49
637 6.83 81.4 31-433 0.2-2.6 1.13
2.37
Mean ratio
2198
k1.04
+ 0.94
experiments at a concentration of 1 mM (a concentration that was also in the nearly linear portion of the in vitro velocity versus substrate curve), and the ratio of the two activities is listed in row 7 of Table III. The ratio of the in vivo and in vitro activities is reasonably consistent across the three animal species (mouse = 3.36, rat = 2.25, hamster = 1.49). This suggests that the in vitro and in vivo results are proportional to each other in spite of the obvious variations in assay conditions (e.g., cofactor concentrations, ionic strength, oxygen tension, substrate concentrations, etc). In vitro measurements of MFO activity in microsomes prepared from human livers revealed considerably more variation than was present in microsomes prepared from the animal species. In addition to the two samples of human liver microsomes reported in Table I, two additional samples of microsomes were assayed: HL-103 and HL-105. One of these (HL-103) had about one-fourth the activity observed in HL-99 and HL-109, and the other (HL-105) had about twice as much activity as HL-99 and HL-109. Therefore, variations in the level of MFO in human tissues spanned almost one order of magnitude. However, the mean ratio of in vivo to in vitro activity in human liver microsomes (1.13 f 1.04) was fairly close to the mean ratio observed in the animal liver microsomes (2.37 f 0.94), and this comparison provides support for the concept of incorporating in vitro enzyme data into the in vivo PB-PK model for CHzClz.
GST enzymes Saturation
of MFO occurred
in vivo during exposure
to concentrations
of 300-500
106
ppm CHzClz [l]. In vitro K, values for MFO activity were in the range of l-2 mM CHZCL (Table II). Since the in vitro K, values for GST in animal tissues were at least two orders of magnitude higher than those for MFO (Table II), it appears unlikely that saturation of GST in vivo would occur at concentrations of CHzCl2 roughly one order
of magnitude
higher
than those causing
saturation
of MFO in vivo. Conse-
quently, the in vitro studies reported here are consistent with the description of GST metabolism as a first order pathway as described by Andersen et al. [l]. The rate of GST reaction (dAM2/dt) in this paper was calculated according to the equation: y
(2)
= (KF) (Conc)(Vol)
where KF= first order rate constant, Cone zconcentration of CHzClz in venous blood leaving the liver, and Vol = size of the liver in liters. This equation is equivalent to a simplified form of the Michaelis-Menten equation: dAM2 dt
( Vmax)(Conc)(Vol) - (K, + Cone)
V,,, K (Conc)(Vol) m
(3)
where I’,,, has units of mg/h per liter of tissue and Cone is small relative to K,. Thus, the constant KF in Eqn. (2) is equivalent to the expression vm,,/K, in Eqn. (3). Assuming that a liter of liver from each species contains roughly equal amounts of cytosolic protein (so that the volume term drops out of the equation), we can rearrange Eqn. (2) to solve for KF: KF= (KI)
dAFoycdt
= (K2) Velocity Substrate
(4)
where KF is equal to a proportionality constant times the rate of the enzyme reaction divided by the substrate concentration (V/S ratio). Eqn. (4) can be written in the same form for either the in vivo reaction or the in vitro reaction; only the proportionality constant is different (KI or Kz). Since the in vivo KF and the in vitro V/S ratio are known with reasonable accuracy for the B6C3Fl mice, the value of KZ can be obtained by dividing KF by the V/S ratio, and then used to calculate the in vivo KF in any species for which an in vitro V/S ratio is known. V/S ratios for GST activity in cytosol prepared from the livers of B6C3Fl mice, F-344 rats, Syrian Golden hamsters, and human livers were determined by dividing the in vitro reaction velocities for GST in Table I (columns 2-6) by the substrate concentration (column 1, Table I). The V/S ratio was essentially constant up to 25 mM in all species, but declined at concentrations of 50 mM or higher in the human liver samples. Consequently, the average V/S ratio for concentrations less than 50 mM in a given species (Table IV, column 1) was used in these calculations. Values of the in vivo rate constant KF calculated for rats, hamsters, and humans, using mouse data as a reference point, are listed in column 2 of Table IV.
107
TABLE
IV
METABOLIC Calculation contains
RATE
of metabolic the velocity
in vitro studies their
PB-PK
calculating specific
CONSTANTS rate constants
to substrate
as outlined model.
of MFO (Al)
model
2 contains
of Andersen
in the rat, hamster,
and human.)
Columns
Calc
Orig
KF
KF
4.01a
1
from the
et al. [l] for
et al. [l] is used as a reference
and GST (A2) in lung and liver in the various
141.4
et al. [l]. Column
the value of KF calculated
3 lists the value of KF cited by Andersen
of KF cited by Andersen
v/s
Mouse
Column
in the text. Column
(The value
the in vivo rate constants
activities
for use in the PB-PK
ratio (V/S).
point
for
4 and 5 show the relative species. Al
A2
4.01”
0.405
0.28
0.63
2.21
0.039
0.14
0.16
1.51
0.068
0.10
0.43s
0.53
0.0014
0.18
k31.1 Rat
116.9 +8.6
Hamster
29.6 + 1.5
Human
105.6 t 12.0
a Used as a reference b Value multiplied
point
for the other
by 0.75 to correct
values.
for the fact that only 3 of the 4 human
liver samples
had detectable
activity.
The estimated value of KF for humans obtained by this procedure (0.43) was very close to that which Andersen et al. [l] obtained by allometric scaling of the mouse data (0.53; Table IV, column 3). The value of KF for rats (0.63) was somewhat lower than that estimated by Andersen et al. [l] in gas uptake experiments (2.21), and the value of KF in hamsters (0.16) was almost an order or magnitude less than the value that Andersen et al. [l] derived from gas uptake (1.5 1). It seems likely that the rather large discrepancy in the hamster constants is related to an inherent limitation of the gas uptake procedure from which the in vivo constant was estimated. The gas uptake technique depends upon measuring overall rates of loss of CHzC12 from the chamber [8]. Consequently, uncertainty in the values of rate constants estimated with this procedure increases when one of the two pathways is much more active than the other. In the mouse, the ratio of GST to MFO activity, measured at substrate concentrations of 6.7 and 5.0 mM CH&lz, respectively, was 0.420 (Table I). Consequently, the gas uptake procedure should give reasonable accuracy in estimating both the GST and MFO rate constants. In contrast, the ratio of GST to MFO activity in
108
hamsters was only 0.013, and large errors in the estimation of in vivo GST rate constants are possible (although estimation of in vivo MFO rate constants by gas uptake should be reliable in this species). Clearly the values for KF estimated from gas uptake studies in mice will be the most reliable, the values estimated from gas uptake studies
in rats sowewhat
will contain
the highest
less reliable,
and the values
for KF in the hamster
studies
levels of uncertainty.
Correlation of in vitro results with animal bioassays High incidences of malignant lung and liver tumors have been observed in B6C3Fl mice exposed to high concentrations of CHzClz vapor [2]. However, increased levels of lung and liver tumors were not seen in F-344 rats and Syrian Golden hamsters exposed to similar levels of CHzClz vapor [2,20], and exposure of B6C3Fl mice to CH2C12 in drinking water also failed to induce a tumorigenic response [3]. Andersen et al. [l] have suggested that the reactive metabolites derived from conjugation of CHzClz with glutathione are likely to be responsible for the tumorigenic activity of CHzClz in the B6C3Fl mouse, and the PB-PK model allows us to calculate levels of these metabolite(s) in the target tissues of the various species in either inhalation or drinking water studies to determine whether there is a correlation between GST activity and tumorigenicity. Under the conditions of the NTP inhalation bioassay, the dose (in mg equiv. of GST metabolites formed per day per liter of tissue) to lung and liver tissues of the B6C3Fl mouse was 482 and 1670, respectively (Table V). Under the conditions of the NCA drinking water bioassay in this same strain, the dose to lung and liver tissues was only 2.2 and 16, respectively. This result correlates well with the observed tumor frequencies in the two bioassays; malignant tumors of lung and liver tissue were observed following inhalation exposure, but no significant changes in lung or liver tumor
TABLE
frequencies
were observed
water study
[2,3].
V
GLUTATHIONE
CONJUGATE
Average
of glutathione
amounts
h per liter of tissue in various week study) ppm,
in the drinking
or drinking
conjugate
(in mg equiv.
species during
water exposures
and doses in the drinking
water
inhalation
(24 h/day). study
of CH2C12) produced
exposures Concentrations
(6 h/day,
in target
5 days/week,
in the inhalation
tissues per 24 96 weeks/l04
exposures
are in
are in m&kg/day.
Species
Dose
Route
Lung
Liver
Mouse
4000=
Inh
482
1670
Mouse
250b
Rat Hamster
Water Inh Inh
4000” 3500a
a Dose in ppm of CH2C12. b Dose in mg/kg/day
in drinking
water.
2.2 96 17.1
16.0 677 I67
109
The dose of GST metabolites exposure
to lung and liver tissue in the rat following
(4000 ppm) to CHZCL was 3-5-fold
lower than in the mouse
inhalation exposed
to
4000 ppm (Table V). The rat failed to show a significant increase in the levels of lung and/or liver tumors, although an increase in the number of benign mammary tumors was noted in this species [2,20]. The dose of GST metabolites to lung and liver tissue in the hamster exposed to 3500 ppm CHzClz was l l-31-fold lower than in the mouse (Table V). Hamsters exposed to CHzClz failed to show significant increases in any type of tumor at any site [20]. Thus there appears to be a reasonable correlation between the in vivo activity of GST in the various species and their sensitivity to CH2C12.
Human risk estimations Quantitative
risk assessments
(QRA)
cannot
provide
precise estimates
of human
risk due to the empirical nature of the models. Furthermore, as a policy decision, QRA procedures have been designed to be protective of public health through inclusion of conservative assumptions wherever uncertainty exists. Consequently, as currently practiced, QRA provides estimates of the ‘plausible upper bounds’ on risk. Acutal risks are unknown, and in many instances may be as low as zero. Nevertheless, QRA plays a useful role in hazard evaluations by providing an objective mechanism for ‘ranking’ materials thought to pose a carcinogenic risk to man. This allows limited societal resources to be directed toward those materials that probably pose the greatest risk to human populations. QRA may also provide a mechanism for definitive action in some cases. For example, if a material is thought to have carcinogenic potential, but QRA suggests that even under the most pessimistic assumptions, the risk to human populations will be less than some very small number (e.g., 10p5), then use of that substance in commerce need not be regulated (i.e., the de minimus approach). Within this framework, the PB-PK model can now be used to formulate human cancer risk estimations. Formulation of the risk estimation involves several steps: (1) Selection of appropriate animal bioassay data. (2) Calculation of the delivered dose to the target organ(s) under conditions that produced tumors in animals. (3) Fitting tumor incidence and delivered dose data to one or more empirical dose-response models for carcinogenesis. (4) Defining human exposure conditions of interest and calculating the delivered dose to corresponding organs of humans under those particular conditions. (5) Estimating upper bounds on human risk from the doseresponse model previously fitted to animal bioassay/target organ delivered dose data. This process differs from that used by default when pharmacokinetic data on delivered dose are not available. In the absence of pharmacokinetic data, delivered dose is considered to be proportional to nominal dose throughout the entire exposure range. Extrapolation from animal to human is performed by assuming that equipotent doses are expressed in mg of chemical per unit of surface area in the
110
various species. The latter process always assumes that humans are more sensitive to a given dose of test chemical (in mg/kg per day) than rodents because of the lower surface area to volume ratio (humans are assumed to be 5-6-fold more sensitive than rats and 13-fold more sensitive than mice). Following the default procedure for CH2Cl2 (without use of PB-PK data), Singh et al. [21] estimated that the lifetime risk for continuous exposure to a concentration of 1 pg/m3 CHzCl2 in inhaled air was 4.1 x 10m6 *. This estimate was presented as a ‘plausible upper limit’ on risk, along with the qualifier that ‘. ..the true value is unknown and may be as low as zero’. The uncertainty and upper limit aspect of this assessment reflect the fact that a series of conservative assumptions were made in the face of scientific uncertainty, including uncertainty as to the magnitude of the delivered dose. Development of a unified PB-PK model for CH2Cl2 permitted the reduction of uncertainty in the original risk estimation through replacement of default assumptions with results from the PB-PK model [l]. Delivered doses for the female B6C3Fl mice in the drinking water and inhalation chronic bioassays [2,3] were calculated with the PB-PK model of Andersen et al. [l] and are summarized in Table VI. Model parameters used by Andersen et al. [l] were employed for these calculations with the exception of KF (the metabolic rate constant for the GST pathway) and the proportionality constants Al and A2 (which give the relative levels of GST in lung and liver tissue). Al and A2 were calculated from the data presented in Table I of this report, and KF was calculated as outlined earlier (Table IV). Tumor incidences in treated and control groups of female mice in these studies also are summarized in Table VI [2,3]. (In the interest of brevity, data from male mice are not presented since these gave estimates of risk similar to those derived from female mice.) Once delivered dose had been calculated, it was necessary to choose a doseresponse model to describe the animal responses. The choice of the appropriate high dose to low dose model has been the subject of considerable discussion (e.g., see Brown [23]). Some models, including the linearized multi-stage (LMS) model, impose conservative assumptions on the QRA process (in this case by forcing the model to be linear at low doses). Other models such as the Probit model make fewer assumptions about the shape of the dose-response curve. The LMS has been widely used by regulatory agencies because of the feeling that the estimates produced by this model are very unlikely to underpredict the true risk. However, current regulatory practice generally includes presentation of a range of risks estimated from other models as well. Consequently, the delivered dose data were fit to four different dose-response models currently used by the toxicological community. The models chosen were (1)
* 4.1 x 10e6 was the original subsequently
been lowered
‘unit risk’ for CHzCll based
in the EPA document
on use of PB-PK
modeling
[22].
of 1985, but this estimate
has
111
TABLE
VI
INTERNATIONAL The ‘internal
DOSE AND TUMOR
dose’ calculated
by the GST pathway of 6 h/day,
INCIDENCES
by the PB-PK
model is in average
per day per liter volume
5 days/week,
corrected
for the fraction
Values reported
for humans
for the fraction
of the year (48152) that humans
Nominal
dose
mg equivalents
of tissue. The values reported of a lifetime
are for one year of occupational
(96/102)
exposure,
of CHzCls
metabolized
for mice are for exposures that animals
6 h/day,
were exposed.
5 days/week,
corrected
are in the workplace.
PB-PK
PB-PK
Lung
Liver
(lung)
(liver)
tumors
tumors
Mouse: Drinking 0
water
mg/kg
0.0
60 mg/kg
0.395
125 mg/kg
0.902
185 mg/kg
1.46
250 mg/kg
2.19
0.0
3.00 6.71 10.8 16.0
5/100
6/100
3/100
4/99
l/50
2/50
3/50
5/50
4/50
3/50
Inhalation 0.0
0 ppm 2000 ppm
0.0
321 482
4000 ppm
3150
3/50
785
30/48
16/48
1670
41/48
40/48
Human: Occupational
exposure
1 mm 2 mm
5.36 x lO-5 1.07x
1o-4
1.73 x 1o-4
5 ppm
2.69x
1O-4
4.38 x 1O-4
8.61 x 10-s
10 ppm
5.40 x 1om4
8.94x
10m4
20 ppm
1.09 x 10-s
1.86x
1o-3
50 ppm
2.80 x lo-’
5.34x
1o-3
5.98x
1.37 x 10-z
100 ppm Continuous
inhalation
1 wg/m’ Drinking
1O-3
6.16x 10-6
9.87 x 1O-6
4.26 x lo-’
3.66 x 1O-6
_
water
1 pg/l
the Linearized Multi-Stage (LMS), (2) the Probit model, (3) the Weibull model, and (4) the Logit model. All of the models predict excess risk (ExRisk) as a function of administered dose, except the Probit model which predicts the probit of the excess risk. ExRisk is defined as the excess lifetime risk resulting from the independent effect of the chemical exposure in question:
112 TABLE
VII
CONSTANTS Constants
obtained
Linearized
Multi-Stage
female
when
computerized
(LMS), Logit,
optimizations Weibull,
mice (given in Table VI). These constants
excess risk for a given PB-PK
were used to fit four
and Probit
- to the PB-PK
dose-response
dose and tumor
can be used with the equations
models incidence
for
in the text to estimate
dose.
Independent
Background
B
A
background LMS-Lung
0.0453
LMS-Liver
0.0550
Logit-Lung
0.0473
- 8.788
1.696
Logit-Liver
0.0576
- 22.258
3.207
Weibull-Lung
0.0434
- 5.072
0.922
Weibull-Liver
0.0576
- 15.273
2.132
Probit-Lung
0.0475
- 5.303
1.024
Probit-Liver
0.0576
- 13.555
1.952
ExRisk
= (TumFr
0.00495
0.00685
- BckGrnd) (5)
(1 .O - BckGrnd)
In this equation, TumFr is the tumor frequency observed in the treated groups and BckGrnd is the tumor frequency observed in control groups. The mathematical forms of the four dose-response models are given for the LMS model [Eqn.(6)]*, the Logit model [Eqn. (7)], the Weibull model [Eqn. (S)], and the Probit model [Eqn. (9)l. ExRisk
= A * Dose
ExRisk
= ~
ExRisk
= 1 - EXP [-A
Probit
(6)
~~~~.. 1 1 + EXP(-A -B*LN(Dose)
(7)
* (Dose)B]
(8)
= A + B * LN (Dose)
(9)
Constants for these equations were derived by computer optimization of the tumor incidences and PB-PK doses from Table VI. All of the treated and control female mice from the two studies (500 animals) were used in the analysis. The values of the constants obtained from these optimizations are listed in Table VII. The models can now be used to estimate the risk for any exposure condition within
* The LMS model fidence
is first fit in the form
level on the linear
of an exponential
term is calculated
polynomial,
to give the low dose form
and then the upper shown
in Eqn.
(6).
95% con-
113
the scope of the PB-PK
model.
To illustrate
this, 3 specific exposure
scenarios
were
evaluated: (1) lifetime exposure to 1 pg/m3 CHzClz in the atmosphere (EPA’s standard Unit Risk condition), (2) lifetime consumption of 2 l/day of water containing 1 ,ug CH2C12/1, and (3) one year’s occupational exposure to CHzClz at the proposed Threshold Limit Value (TLV, 50 ppm). The PB-PK doses equivalent to these exposures are listed in Table VI. Low dose risk estimates obtained with the Logit, Weibull, and Probit models may vary depending upon assumptions made about the additivity or independence of the background tumors. Since the LMS model assumes an additive background, this consideration does not apply to this model, The question of whether to use independent or additive backgrounds is a difficult one, and is beyond the scope of this paper. However, the effect of incorporating pharmacokinetic data into the risk estimation can be visualized without resolving this question. The LMS model, which assumes an additive background, generally gave the highest estimates of low dose risk for a given dose of CH2C12 (Table VIII). Low dose risk estimates from the other three models were similar to the risk estimates from the LMS when additive backgrounds were assumed (data not shown). However, when independent backgrounds were assumed, the 4 models give very different predictions of low dose risk (shown in Table VIII). Although each model gave an excellent description of the observable data from the two bioassays, the models differed by 5 to 15 orders of magnitude in predicted risk for the exposure conditions depicted in Table VIII. The effect of incorporating pharmacokinetic principles into the risk assessment process may be visualized by comparing the total risk (lung + liver) predicted by the LMS model. Under conditions of continuous inhalation of 1 pg/m3, the LMS predicted an upper bound risk of 4.1 x 10e6 for lifetime exposure [21] when administered dose was used instead of delivered dose. However, use of a pharmacokinetically calculated delivered dose predicted a risk more than two orders of magnitude lower: 3.7 x lo-’ (Table VIII). There are two components of this difference. The first is the non-linearity in the activity of the metabolic pathways. The primary metabolic pathway (MFO) becomes saturated at high exposure concentrations of CHX12, and a disproportionate increase in the activity of the secondary toxic pathway (GST) occurs. The second component has to do with calculation of species-specific doses rather than generic surface area ‘corrections’. Humans possess relatively lower levels of the enzyme(s) responsible for CHzClz toxicity, and at equivalent tissue concentrations of CHZCIZ will generate lower levels of the toxic GST metabolite. At the same time, it needs to be pointed out that use of a PB-PK model in QRA does not always result in estimation of lower risks for humans than the default procedures mentioned earlier. For instance, if a test chemical were directly toxic and were inactivated by metabolic enzymes, then a PB-PK model would predict more risk to humans than rodents because of the lower levels of detoxifying enzymes in
114
TABLE
VIII
EXCESS RISKS EXPOSURE
FOR INHALATION,
Excess risks calculated tinuous
inhalation
WATER
by four dose-response
models
of 1 @g/m’ in air for a lifetime,
1 @g/liter for a lifetime,
CONSUMPTION
and (3) occupational
are structured
with the non-additive
under three specific
(2) consumption
exposure
at 50 ppm for one year. The four models are: Linearized
AND OCCUPATIONAL
(6 h/day,
exposure
5 days/week,
Multi-Stage,
conditions:
of 2 liters of water/day Logit,
48 weeks/year) Weibull,
(1) concontaining to CH&Zlz
and Probit.
Models
background. Inhai,
Water
occup.
1 pgg/m3
I pglliter
50 ppm
LMS (w/P-K)
3.0x
10-x
2.1 x 10-9
1.4x 10-S
Logit
2.1 x lo-‘3
2.3 x lo-‘5
7.0x
Weibull
9.8 x IO-*
8.4 x 1o-9
2.8 x 10-5
Risk of lung tumor 10-9
Probit
< lo-‘5
< lo-‘5
< IO-i5
Geo. mean
2.8 x io-”
2.5 x lo-i2
7.2 x 10-9
Risk of liver tumor LMS (w/P-K)
6.8 x IO-’
2.5 x 1O-9
3.7 x 10-e
Logit
2.1 x 10-26
8.8 x 1O-28
1.2x 10-i’
Weibull
< lo-i7
Probit
< lo-‘5
< 10-15
< lo-‘5
Geo. mean
3.5 x 10-17
1.2x 10-l’
2.6 x IO- i4
Total
risk (lung
LMS (w/o
f
liver)
P-K)’
4.1 x 10-s
LMS (w/P-K)
3.7x
4.6 x lO-9
1.8x10-5
Logit Weibull
2.1 x lo-”
10-S
2.3 x lo-is
7.0x
9.8 x 10-8
8.4 x 10-a
2.8 x 10-S
10-9
Probit
< lo-‘”
< lo-i5
< 1o-‘5
Geo. mean
2.9 x 10-i’
3.1 x10-i*
7.7 x 10-9
’ Singh et al. [21] estimated risk of developing
the risk of developing
only a lung tumor
either a lung or liver tumor,
but did not estimate
-the
or only a liver tumor.
this species. Similarly, if toxicity resulted from reactive metabolitets) produced by a saturable process, then low dose risk extrapolated from high, saturating doses by a PB-PK model could be more than predicted by linear extrapolations. In summary, development of a PB-PK for CHA&, and subsequent incorporation of in vitro enzyme studies, has provided a valuable tool for improving risk estimations for CH$&. The model permitted identification of a probable mechanism of carcinogenesis for CH2C12, provided a consistent explanation for the differences between chronic bioassays of CH2CL in different species and by different routes, and provided a quantitative basis for evaluating the effect of two competing metabolic pathways upon the production of the putative toxic metabolite in situ at various
115
doses. We believe that risk estimations that incorporate these principles will prove significantly more reliable than risk estimations that do not consider pharmacokinetics. ACKNOWLEDGEMENTS
The authors would like to thank Dr. Daniel Krewski (Health Protection Branch, Canada) and Dr. Murray Cohn (Consumer Products Safety Commission) for their expert assistance in fitting the PB-PK delivered dose data to the various cancer doseresponse models. REFERENCES Andersen,
M.E.,
ly based
Clewell, H.J.,
pharmacokinetics
Pharm.
87, 185-205.
National
Toxicology
III, Gargas,
Program
of Dichloromethane
TR-306
(board
draft).
Serota,
D., Ulland,
Caster,
March
J., Simon,
determined
process
for methylene
chloride.
Toxicol.
on the Toxicology Mice (Inhalation
Chronic
8-9, Bethesda,
Appl.
and CarcinoStudies).
NTP-
Oral Study in Mice. Food Solvents
MD.
A.B. and Armstrong,
by dissection
Report
Rats and B6C3Fl
F. (1984) Hazelton
Chloride,
Poncelet,
values
in F-344/N
B. and Carlborg,
I: Methylene
W.O.,
Normal
Smith, F.A. and Reitz, R.H. (1987) Physiological-
(NTP) (1985) NTP Technical
genesis Studies
Workshop
M.L.,
and the risk assessment
W.D.
and chemical
(1956) Tissue weights
methods.
Proc.
Sot.
Exp.
of the rat. I.
Biol.
Med.
91,
model
of
122-126. Davis,
N.R.
and Mapleson,
W.W.
of injected
agents
the distribution International Man.
Commission
In: W.S.
Snyder,
ICRP
Publication
(Eds.),
on Radiation M.J.
Gargas,
blood,
Anders,
dihalomethanes 10 Ahmed,
Protection
E.S. Nasset, Snyder
T. (1979) Partition
metabolism
V.L.,
and quantification
anaesthetics.
(1975) Report L.R.
coefficients
53, 399-404.
of the task group
Karhausen,
et al., Eds.),
of a physiological
Br. J. Anaesthesiol. G.P.
Pergamon
Howells
Press,
of some aromatic
on Reference
and I.H.
Tipton
New York. hydrocarbons
and ketones
and oil. Brit. J. Ind. Med. 36, 231-234.
M.L. and Andersen,
of chemical Kubic,
Cook,
23 (W.S.
Sato, A. and Nakajima, in water,
(1981) Structure and inhaled
M.E. (1988) A gas phase technique
in the rat. Toxicologist M.W.,
to carbon
Engel,
A.E.
and Anders,
M.W.
organic
halide.
I. In vitro
studies.
Gargas,
M.L.,
Clewell,
viva: differentiation
R.R.,
monoxide.
H.J.
Barlow,
(1976) Metabolism Metab.
and Andersen,
of kinetic
C.H.
I. In vivo studies. Drug
constants
for determining
the kinetic constants
8, 112. and Caughey,
W.S. (1974) Metabolism
Drug Metab.
Dispos.
of dihalomethanes
Dispos.
of
2, 53-57.
to formaldehyde
and in-
4, 357-361.
M.E. (1986) Metabolism
for two independent
of inhaled
pathways.
dihalomethanes
Toxicol.
Appl.
in
Pharmacol.
82, 21 l-223. Lindstedt, Council
S.L.
(1987) Allometry:
of the National
Academy
and Health,
Vol. 8, National
Beaune,
Kremers,
parison
P.,
microsomes. Backer,
R.H.
Drug Metab. and
Academy
P.G.,
of monooxygenase
body
size constraints
of Sciences, Press,
Kaminsky, activities Dispos.
Guengerich,
L.W.,
evidence
Washington,
dosing.
In: National
in Risk Assessment,
Research
Drinking
Water
DC, pp. 65-79.
DeGraeve,
and cytochrome
L. and Guengerich,
P-450 isozyme
F.P.
concentrations
(1986) Comin human
liver
14, 437-442. F.P.
(1986)
methyl-3,5-bis(alkoxycarbonyl)-1,4-dihydropyridines chemical
in animal
Pharmacokinetic
for the involvement
Oxidation by
of 4-arylhuman
of a form of cytochrome
and liver
4-alkyl-substituted microsomes
and
2,6-diimmuno-
P-450. J. Med. Chem. 29, 1596-1603.
116
15 Hall, S.D., Guengerich, tion of mephenytoin 216-222. 16 Knodell,
R.G.,
tolbutamide:
F.P.,
Branch,
Hall,
S.D.,
in human
Wilkinson,
in viva characterization
tion and relationship 17 Guengerich,
F.P.,
Characterization oxidation,
R.A. and Wilkinson,
4-hydroxylase
G.R.
M.V.,
Beaune,
of rat and human
a prototype
and Guengerich,
for genetic
J. Pharmacol. P.H.,
Kremers,
liver microsomal polymorphism
J. Pharmacol. F.P.
of the form of cytochrome
to in vivo disposition. Martin,
G.R. (1987) Characterization
liver microsomes.
(1987) Hepatic
P-450 involved
Exp. Ther. P., Wolff,
cytochrome
in oxidative
and inhibi-
Exp.
Ther.
240,
metabolism
of
in methyl hydroxyla-
241, 1112-1119. T. and Waxman,
P-450 forms
D.J. (1986)
involved
drug metabolism.
in nifedipine
J. Biol. Chem.
261,
505 l-5060. 18 Guengerich,
F.P.,
cytochrome
Muller-Enoch,
P-450.
19 Lineweaver,
H. and Burk,
Chem.
Sot.
56, 658-666.
20 Burek,
J.D.,
Nitschke,
Rampy,
study D.V.,
K.D.,
Bell, T. J., Wackerle,
for dichloromethane
H.L.
Fundam.
and White,
(methylene
P.D.
chloride).
of enzyme
D.L.,
M.J. (1984) Methylene
in rats and hamsters. Spitzer,
I.A. (1986) Oxidation
of quinidine
by human
liver
30, 287-295.
D. (1934) The determination
L.W. and McKenna,
genicity 21 Singh,
D. and Blair,
Mol. Pharmacol.
App.
Childs,
R.C.,
constants.
Beyer, J.D.,
chloride:
A 2-year inhalation
Toxicol.
4, 30-47.
(1985) Addendum
Updated
dissociation
to the health
carcinogenicity
assessment
J. Amer.
Dittenber, toxicity
assessment
D.A.,
and oncodocument
of dichloromethane.
EPA/600&82/004F. 22 Blancato,
J.N.,
addendum epidemiology. 23 Brown,
Hopkins,
J. and Rhomberg,
for dichloromethane C.C.
External
Review
(1983) Learning
(methylene Draft, about
L. (1987) Update chloride):
to the health assessment
pharmacokinetics,
mechanism
document
and
of action,
and
EPA/600/8-87/030A. toxicity
in humans
from
studies
on animals.
Chemtech
13, 6.