PD) modeling to understanding the mechanism of action of hazardous stances

lbxicology

letters

ELSEVIER

Toxicology Letters 79 (1995) 185-191

The application of physiologically based pharmacokinetic/ pharmacodynamic (PBPK/PD) modeling to understanding the mechanism of action of hazardous substances Michele

A. Medinsky

Chemical Industr,y Institute of Toxicology, P.O. Box 12137, 6 Davis Drive, Research Triangle Park, NC 27709, USA

Accepted 5 April 1995

Abstract Much of toxicology research is focused on elucidating the nature of the mechanisms through which various xenobiotics exert their toxic effects. The central issue in extrapolating laboratory experiments to the human situation is whether mechanisms which are operative in laboratory animals are similar to mechanisms operating in humans. The underlying assumption is that understanding mechanisms permits rational extrapolation between species, across routes of exposure, or from high to low doses. There are two general classes of mechanisms of action. First, there are the mechanisms that result in the translation of an exposure concentration to the effective dose at the target site. The mechanisms that are operative at a pharmacokinetic level include those that are physiologically driven and those that are metabolically based. Second are mechanisms through which the dose at the target site elicits the ultimate adverse response. These are pharmacodynamic in nature and refer to the action of the effective dose at the target site. Altered gene regulation, cytotoxicity, and cell proliferation are examples of processes involving potential adverse effects at the target site. A quantnative understanding of the mechanisms involved in going from exposure to dose and dose to response can aid in answering the question of whether or not these mechanisms in animals and humans are similar or different. Keywords: Physiologically based pharmacokinetic/pharmacodynamic Phenol metabolism; Hydroquinone

1. Introduction Much of toxicology research is focused on elucidating the mechanisms through which various xenobiotics exert their toxic effects. The central issue in extrapolating laboratory experiments to 037%4274/95/$09.50

modeling;

Benzene metabolism;

the human situation is whether mechanisms that are operative in laboratory animals are identical to mechanisms operating in humans. The toxicologist’s interest in understanding and quantitatively modeling such mechanisms is based on the assumption that understanding mechanisms permits

0 1995 Elsevier Science Ireland Ltd. All rights reserved

SSDI 0378-4274(95)03369-V

(PBPK/PD)

186

M.A. Medinsky 1 Toxicology Lelrers 79 (1995) 185-191

more rational extrapolation between species, across routes of exposure, or from high to low doses. What is a mechanism of action? There are 2 levels at which toxicologists encounter mechanisms of action. First are the mechanisms resulting in the translation of an exposure concentration to the effective dose at the target site. In the broadest sense, the mechanisms that are operative at this pharmacokinetic level include those that are physiologically driven and those that are metabolically based. Second are mechanisms through which the effective dose at the target site elicits the ultimate adverse response. These processes are pharmacodynamic in nature and can include altered gene expression, cytotoxicity, cell proliferation and chronic disease. A quantitative understanding of the mechanisms involved in progressing from exposure to dose and dose to response can aid in answering the question of whether or not these mechanisms in animals and humans are identical. The objective of this presentation is to briefly review 2 mechanisms at the pharmacokinetic level that are important in the translation of an exposure concentration to the effective dose at the target site: physiological mechanisms and biochemical mechanisms. 2. Target tissue dosimetry Predicting the effective dose at the target site is the ultimate goal of mechanism-based dosimetry models. Dosimetry can be approached from several levels of complexity. A pyramid approach to understanding dosimetry is presented in Fig. 1. A detailed discussion of this concept of dosimetry has been presented by Dahl [l]. Briefly, the most basic, and broadest, level relates to the exposure concentration. To proceed to the next level, information on exposure concentration is essential. Other important, and more detailed, aspects of dosimetry include pharmacokinetic parameters of the parent chemical and flux of the parent chemical through metabolic pathways, in particular, ones that involve production of toxic metabolites. Information on the latter can often be obtained by analysis of urinary metabolites. This

Target Tissue/ Cell

Blood Levels of Metabolites uiinmy Metaboliiea Parent chemical

Fig. 1. Pyramid from Dahl [l].

approach

to understanding

dosimetry

adapted

basic pharmacokinetic information is required before one can integrate data on target tissue and target cell dosimetry. Data on target tissue and target cell dosimetry are more difficult to obtain experimentally, but are more meaningful for risk assessment purposes. At the apex of the pyramid is molecular dosimetry, the area of dosimetry that begins to describe the process by which the ultimate toxicant interacts with and modifies proteins, DNA, or other biological molecules which consequently may cause or contribute to a toxic response.

* h

3. Physiological

mechanisms

The incorporation of physiological mechanisms into quantitative mathematical models is demonstrated in Fig. 2. In this example a physiological

1.0

z 0

m

-g

.l

.

RAT

l

. .

.

.

.

.

.OO I1

0

-

20

40

Hours

After

-b Start

of

Exposure

Fig. 2. Concentration of styrene in the venous blood of rats and humans exposed to 80 ppm styrene for 6 h. Data points represent individual measurements of styrene in blood. Lines are results of simulations using the PBPK model of Ramsey and Andersen [2]. Taken from Medinsky [4] with permission.

M.A. Medinsky / Toxicology Letters 79 (1995) 185-191

model, developed by Ramsey and Andersen [2], was used to predict the concentration of styrene in blood of both rats and humans based on the inhaled exposure Iconcentration. Both rats and humans were exposed to an identical concentration of styrene (80 ppm), for 6 h by inhalation. In both species, blood concentrations of styrene were measured during and after exposure. Inspection of the data revealed that the terminal half-life for styrene is much sh.orter in rats than that in humans. In fact, using data-based compartmental analyses, one might conclude that the mechanisms responsible for styrene disposition are different for the 2 species. However, this does not appear to be the case since the structure of the physiologically based pharmacokinetic (PBPK) model used to predict styrene kinetics in both rats and humans is identical. Only the values for model parameters, such as organ volumes, blood flows to organs, and ventilation and perfusion rates vary; and these vary based on measurable parameters in the 2 species. The power of PBPK models is that a single model structure is able to describe what appears to be very different pharmacokinetic behaviour between species. Simply by changing measurable physiological parameters, such as blood flows and organ volumes, specific for rats or for humans, a single model structure simulates both data sets. How is a physiological model able to do that? The answer is baseId on the observation that some processes, such as blood flow and clearance, occur significantly faster in small animals than compared to large ones. Mordenti [3] described this difference among species quantitatively by examining half-lives of the drugs cephalosporin and monobactum in various mammals. Measured in conventional time units, such as minutes, the halflives were much longer for humans than dogs and monkeys, which were longer than for rats, or mice. However, after converting conventional time to a physiological unit, such as heart beats, all 5 species actually had similar drug half-lives. Since the heart beats faster in small animals than large ones, expressing drug clearance as a function of a physiological variable, such as heart beat, can take into account physiological processes being faster in small animals compared to large ones.

187

Returning to the example of styrene, the terminal half-life for styrene is proportional to the ratio of the volume of distribution to clearance (I’,/ CI). The terminal half-life is shorter for rats compared to humans because the volume of distribution decreases faster than clearance does when extrapolating from a large body weight to a small body weight [4]. This change in ratio exists because the volume of distribution across species is related to body weight raised to an exponential power approximately equal to 1 (e.g., BW 2 I). In contrast, clearance across species is related to body weight raised to a fractional power (e.g., BW
188

M.A. Me&sky

/ Toxicology Letters 79 (1995) 185-191

as styrene. In the hepatic compartment, the model assumed that benzene could be metabolized via saturable kinetics to benzene oxide, a reactive intermediate. This intermediate could in turn be metabolized by 1 of 4 competing pathways: formation of glucuronide and sulfate conjugates of phenol, formation of glucuronide and sulfate conjugates of hydroquinone, oxidation to muconic acid, and conjugation of benzene oxide with glutathione with the eventual production of Sphenyl-iV-acetyl cysteine (Fig. 3A). Each of these pathways was also assumed to be saturable. Embedded in this model was a number of simplifications related to the mechanism of benzene metabolism. For example, the formation of the hydroquinone conjugates, which involves oxidation of benzene to benzene oxide, rearrangement to phenol, oxidation of phenol to hydroquinone, and conjugation of hydroquinone with either glucuronic acid or sulfate, was simulated using a the simple single V,,, and K,. Nonetheless, model was useful in predicting a number of important aspects of benzene metabolism. Since the model incorporated saturable kinetics, a realistic biochemical mechanism for benzene metabolism, the uptake and metabolism of benzene could be simulated over a range of oral and

e .

? . 200

4w

BM

BENZENE CONCENTRATICW @pm)

Fig, 3. (A) Flow-diagram for benzene metabolism in the hepatic compartment. Metabolic scheme taken from Medinsky et al. [7]. All processes are assumed to be saturable. (B) Micromoles hydroquinone metabolites excreted in urine per kg body weight for B6C3Fl mice exposed to various concentrations of benzene. The data points represent experimentally determined values for hydroquinone excretion in mice exposed to benzene concentrations of 5, 50, or 580 ppm for 6 h as reported by Sabourin et al. [6]. The lines represent simulations using a physiological model and the metabolic scheme presented in panel A. Benzene’s inhibition of phenol metabolism to hydroquinone is not included in the mode1 structure.

4600

3000

!

my’

2000

k l/

/

/

/’

0

d

1000

*3’

l

;

I

0

0 ORAL

EXPOSURE

(mglkg)

INHALED

400

800

1000

CONCENTRATION Gwm)

Fig. 4. Total pmol benzene metabolized per kg body weight for F344 rats and B6C3Fl mice during and after exposure to various doses of benzene. The data points represent experimentally determined values for metabolite formation in mice exposed to various benzene doses by inhalation (A) or gavage (B) as reported by Sabourin et al. [16]. The lines represent simulations using a physiological model as reported by Medinsky et al. [7]. Taken from Medinsky et al. [7] with permission.

inhalation exposures in both rats and mice (Fig. 4). The simulations for total benzene metabolized as a function of increasing dose reflect the saturable nature of benzene metabolism. One important physiological mechanism inherent in the model structure is that, as benzene metabolism in the liver becomes saturated, benzene is exhaled unchanged [S]. The blood/air partition coefficient for benzene, a measurable parameter, determines the proportion of the benzene exhaled unchanged. The experimental data reflect this competition between metabolism and exhalation. As the oral benzene dose is increased and metabolism becomes saturated, both rats and mice exhale an increasing percent of the administered benzene. In general, this simple physiological model also adequately simulated the formation of individual benzene metabolites, including the conjugates of phenol, muconic acid, and phenyl mercapturic acid. However, the formation of hydroquinone conjugates, especially after inhalation exposure (Fig. 3B), was not simulated well. Whereas the model predicts that the maximum hydroquinone formation occurs at high benzene exposure con-

M.A. Medinsky / Toxicology Letters 79 (1995) 185-191

centrations, the experimental data indicated that more hydroquinone was produced at a low (50 ppm) benzene exposure concentration compared to a high one (600 ppm). This discrepancy between the model predictions and experimental data indicated that some important mechanism, most likely of a biochemical nature, was missing from the model structure. Insight into a potentially important biochemical mechanism was obtained from in vitro studies that examined the metabolism of benzene and phenol by liver microsomes. Koop et al. [9] observed that benzene and phenol are substrates for the same cytochrome P450 isozyme (P450IIEl). Schlosser et al. [lo] developed a simulation model for benzene and phenol metabolism by rat and mouse liver microsomes that incorporated inhibition of phenol metabolism by benzene and benzene metabolism by phenol. The in vitro model suggested that inhibition occurs through competition for a common reaction site. More recently, Hargreaves et al. [l l] investigated the inhibition of p-nitrophenol hydroxylase activity by a series of aromatic molecules, including benzene and phenol. Both benzene and phenol displayed inhibition of a competitive type towards P450IIEl. Taken together, these data suggested that an inhibition term should be incorporated into the model. Based on the in vitro observations, the PBPK model for benzene was modified to include an inhibition term in the equations describing both benzene and phenaol oxidation [12]. In the model, benzene can be oxidized to the reactive intermediate, benzene oxide, a portion of which can rearrange to phenol (Fig. 5A). Phenol itself plays a pivotal role in tb: dose-dependent behavior observed for benzene metabolism. Phenol formed from benzene can be further metabolized through 1 of 2 pathways: conjugation with sulfate and glucuronic acid or oxidation to hydroquinone. Clearance of phenol by conjugation is incorporated into the model (Fig. 5A). Additionally, the model incorporates a very simple expression for competitive inhibition by phenol of benzene oxidation and competitive inhibition by benzene of phenol oxidation. Fig. 5B demonstrates the results of that simple, but mechanistical1.y sound change, in the model.

189

Incorporation of a term for competitive inhibition into the model results in an improved fit to the experimental data. Depending upon the values that are assigned to either V,,,,, or K,,, for phenol oxidation to hydroquinone, the benzene concentration resulting in maximum hydroquinone formation changes. Additional experiments with phenol are needed to provide independent estimates for these values. Nonetheless, this model does reflect the observed pharmacokinetic behaviour in that it predicts that more hydroquinone is produced at low compared to high benzene concentrations. The unusual shape of this curve is due to the inhibition of phenol metabolism to hydroquinone by high concentrations of benzene at the enzyme site. Since the rate of phenol oxidation is reduced, elimination of phenol by conjugation reactions becomes the preferred pathway at these high benzene concentrations. This model prediction is consistent with the experimental data. Are there any data on the toxic effects of benzene that suggest that this mechanism might be relevant? In fact, this unusual dose-response relationship is seen in a number of studies in which mice are exposed to benzene by inhalation, most notably those of Green et al. [13] and Wells and

Fig. 5. (A) Flow diagram for benzene metabolism in the hepatic compartment that includes competition between benzene and phenol for the active enzyme site. (B) Micromoles hydroquinone metabolites excreted in urine per kg body weight for B6C3Fl mice exposed to various concentrations of benzene. The data points represent experimentally determined values for hydroquinone excretion in mice exposed to benzene concentrations of 5, 50, or 580 ppm for 6 h as reported by Sabourin et al. [6]. The lines represent simulations using a physiological model and the metabolic scheme presented in panel A. Benzene’s inhibition of phenol metabolism to hydroquinone is included in the model structure.

190

M.A. Medinsky 1 Toxicology Letters 79 (1995) 185-191

Nerland [14]. In these studies, toxicity to blood cells, either circulating cells or those in the bone marrow, was maximized at lower exposure concentrations compared to higher ones. For example, Wells and Nerland [14] examined the effect of benzene exposure on the number of circulating leukocytes in mice exposed to up to 2200 ppm benzene for 6 h/day for 5 days. Maximum depression in peripheral blood leukocyte count as a percent of controls was observed at 200 ppm benzene (Fig. 6A) and did not get worse even though the exposure concentration increased lofold. The observed dose-response function is similar in overall shape to that observed for hydroquinone formation in Fig. 5B. Wells and Nerland also measured blood concentrations of phenol at the end of the benzene exposure. The concentration of phenol in end-exposure blood samples increased almost linearly as a function of benzene concentration (Fig. 6B). The shape of the curve for phenol was not similar to either the hydroquinone dose-response function or the leukocyte dose-response function, suggesting that if phenol is involved in benzene toxicity, its role is not straightforward. In contrast, the similarities in the shape of the curves for hydroquinone formation and leukocyte depression as a function of benzene exposure suggest that the role for this metabolite in benzene-induced toxicity should be explored in more detail. In vitro studies with human lymphocytes have demonstrated that hydroquinone is a more potent genotoxicant than benzene or phenol [15]. There is no information on the nature of the metabolite responsible for leukemia following benzene exposure. While the results of the PBPK model described here are not conclusive, they are consistent with the observed potency of hydroquinone as a genotoxicant.

0

500

1000

1500 2000

2500

Benzene Concentration (ppm) Fig. 6. (A) Depression in circulating peripheral blood leukocytes expressed as a percent of control. (B) Concentration of phenol in blood of mice following exposure to various concentrations of benzene. Data taken from experiments conducted by Wells and Nerland [I41 in which Swiss Webster mice were exposed to various benzene concentrations for 6 h/day for 5 days. At the end of the last exposure, blood samples were taken and analyzed for number of white blood cells (A) or analyzed for phenol using mass spectrometry.

5. Conclusions To make the most effective use of PBPKjPD models for understanding mechanisms of action, a close association between experimental data collection and model development is required. The possible nature of this association is illustrated in the paradigm in Fig. 7. The 4 elements in this

paradigm include: (1) formulation of a hypothesis regarding proposed mechanisms of action, (2) development of a mathematical model that incorporates the hypothesis in quantitative terms, (3) conduction of experiments to verify model predictions, and (4) re-evaluation of the original hy-

M.A. Medinsky 1 Toxicology Letters 79 (1995) 184-191

Fig. 7. Paradigm for integrating experimental data and quantitative hypothesis testing using PBPK models.

pothesis in light of the experimental findings. The iterative application of experimentation and quantitative hypothesis formulation should result in a continued increase in understanding of mechanisms of action of hazardous substances. When appropriate, this mechanistic data should be incorporated into human quantitative risk assessments which in turn will provide more scientific and better estimates of risk. Acknowledgements

The author gratefully acknowledges financial support, in part, through a grant from the American Petroleum Institute. Valuable discussions with a number of colleagues, including Drs. Elaina Kenyon and Mark Seaton, were also helpful. References [l] Dahl, A.R. (1990) Dose concepts for inhaled vapors and gases. Toxicol. Appl. Pharmacol. 103, 185 197. [2] Ramsey and Andersen (1984) A physiologically based description of the inhalation pharmacokinetics of styrene monomer in rats and humans. Toxicol. Appl. Pharmacol. 73, 1599175. [3] Mordenti, J. (1985) Forecasting cephalosporin and monobactam antibiotic half-lives in humans from data collected in laboratory animals. Antimicrob. Agents Chemother. 27, 887-891. [4] Medinsky, M.A. 1[1990)Critical determinants in the systemic availability and dosimetry of volatile organic chemicals. In: Principles of Route-to-Route Extrapolation for

191

Risk Assessment, Elsevier Science Publishers, N.Y., pp. 1555172. [51 Snyder, R., Witz, G. and Goldstein, B.D. (1993) The toxicology of benzene. Environ. Health Perspect. 100, 293-306. 161Sabourin, P.J., Bechtold, W.E., Griffith, W.C., Birnbaum, L.S., Lucier, G. and Henderson, R.F. (1989) Effect of concentration, exposure rate, and route of administration on metabolism of benzene by F344 rats and B6C3Fl mice. Toxicol. Appl. Pharmacol. 99, 421-444. 171Medinsky, M.A., Sabourin, P.J., Lucier, G., Birnbaum, L.S. and Henderson, R.F. (1989a) A physiological model for simulation of benzene metabolism by rats and mice. Toxicol. Appl. Pharmacol. 99, 193-206. 181 Medinsky, M.A., Sabourin, P.J., Henderson, R.F., Lucier, G. and Birnbaum, L.S. (1989b) Differences in the pathways for metabolism of benzene in rats and mice simulated by a physiological model. Environ. Health Perspect. 82, 43-49. 191 Koop, D.R., Laethem, C.L. and Schnier, G.G. (1989) Identification of ethanol-inducible P450 isozyme 3a (P450IIEl) as a benzene and phenol hydroxylase. Toxicol. Appl. Pharmacol. 98, 278-288. PO1 Schlosser, P.M., Bond J.A. and Medinsky, M.A. (1993) Benzene and phenol metabolism by mouse and rat liver microsomes. Carcinogenesis 14, 2477-2486. 1111 Hargreaves, M.B., Jones, B.C., Smith, D.A. and Gescher, A. (1994) Inhibition of p-nitrophenol hydroxylase in rat liver microsomes by small aromatic and heterocyclic molecules. Drug Metab. Dispos. 22, 806-810. f121IMedinsky, M.A. (1992) Critical issues in benzene toxicity. In: Toxic Air Pollutants from Mobile Sources: Emissions and Health Effects, A&WMA VIP-23, Air and Waste Management Association, Pittsburgh, pp. 244-253. [13] Green, J.D., Snyder, C.A., Lobue, J., Goldstein, B.D. and Albert, R.E. (1981) Acute and chronic dose/response effect of benzene inhalation on the peripheral blood, bone marrow, and spleen cells of CD-I male mice. Toxicol. Appl. Pharmacol. 59, 204-214. u41 Wells, M.S. and Nerland, D.E. (1991) Hematotoxicity and concentration-dependent conjugation of phenol in mice following inhalation exposure to benzene. Toxicol. Lett. 56, 159-166. [I51 Erexson, G.L., Wilmer, J.L. and Kligerman, A.D. (1985) Sister chromatid exchange in human lymphocytes exposed to benzene and its metabolites in vitro. Cancer Res. 45, 2471-2477. P61 Sabourin, P.J., Chen, B.T., Lucier, G., Birnbaum, L.S., Fisher, E. and Henderson, R.F. (1987) Effect of dose on the absorption and excretion of “‘C-benzene administered orally or by inhalation in rats and mice. Toxicol. Appl. Pharmacol. 87, 325-336.