Roles of Quantitative In Vitro ADME in Toxicology and Risk or Safety Assessment
Published: December 6, 2007
Dr. Hugh A. Barton
Office of Research and Development
National Center for Computational Toxicology
Research Triangle Park, NC 27711
Uses of in vitro ADME
There are at least two major contexts in which in vitro ADME needs to be developed. The first is to predict pharmacokinetics in vivo and the second is to provide and/or characterize ADME processes for in vitro assays of effects. Predicting in vivo pharmacokinetics requires obtaining data on selected ADME processes to provide qualitative understanding and then quantitation for use in computer simulation models for intact organisms. Characterizing and supplementing ADME processes in effects assays can be essential to evaluating the forms of the chemical that may be responsible for the effects in the intact organism, notably metabolites.
It is useful to consider the range of chemicals of interest before discussing approaches to evaluating ADME processes for them. It is generally more feasible to develop in vitro assays and computer modeling methods that are successful for chemicals that are more closely related and more difficult to develop methods applicable to wider ranges of compounds. Sometimes, it is reasonable to define classes of compounds and measure or describe their characteristics separately, as is sometimes done for acidic, basic, and neutral pharmaceuticals. For some methods, such as the development of quantitative structure activity or property relationship (QSAR or QSPR) methods, the chemicals used to develop and test these models provide substantial, though not necessarily explicit, bounds on the range of chemicals to which they can be securely applied.
Whether it is the conditions for in vitro assays or the range of applicability of a computer model, it is essential to consider whether the method is robust enough to work for the chemicals under consideration that were not directly used in developing the assay. It may be reasonable to make initial predictions to test further using chemicals outside the range of compounds for which the assay or computer model was developed, but it will be essential to demonstrate that the predictions are reasonable. It is not always easy to characterize the appropriate range of chemicals for which a method works because subtle changes at a molecular level in three dimensional structures, such as mirror image enantiomers, or substituting hydrogens with fluorine, chlorine, or bromine atoms, can lead to substantial changes in their physicochemical properties and similarly in handling by biological systems. Therefore, characterizing the bounds of “chemical space” is probably as important as developing improved in vitro assays and computer models.
Predicting in vivo pharmacokinetics
Predicting in vivo pharmacokinetics for both animals and humans is a very active area of research driven largely by the pharmaceutical industry. A portion of the high attrition rates in pharmaceutical development is a direct result of poor pharmacokinetics, so methods to predict pharmacokinetics for a large number of compounds with minimal data would be extremely valuable. In addition, efforts to utilize the results of some in vitro assays for toxicity has led to consideration of what is required to predict equivalent in vivo exposures (Coecke et al., 2005). Predicting in vivo pharmacokinetics is also integral to the recently proposed vision for a new toxicity testing paradigm relying on characterizing and modeling perturbations of toxicity pathways at environmentally relevant concentrations (NRC, 2007).
One major challenge in predicting in vivo pharmacokinetics is characterizing the correspondence between the in vitro conditions and the in vivo situation. For instance, measurements of glucuronyl transferase enzymatic activity using microsomes, a preparation of subcellular organelles, are highly dependent upon the experimental conditions, in part due to inaccessibility of the enzyme without solubilizing or pore-forming agents (Miners et al., 2006). Accurate quantitation of the microsomal content of liver tissue has been another concern (Barter et al., 2007). Similarly, there have been comparisons of the metabolic capabilities of genetically expressed enzymes, microsomes, isolated liver cells, and other systems for studying metabolism. Similar issues exist for characterizing transporter activities, which are often essential to absorption, tissue distribution, and excretion. In vitro methods to assess some processes, such as biliary excretion and reabsorption or urinary excretion, do not currently exist and need to be developed.
Another major challenge for predicting in vivo pharmacokinetics is developing and implementing computer simulation and modeling. A range of modeling approaches can be useful, with the long range goal being to make predictions for the intact organism from the formula and structure of the chemicals. At least two general objectives are required to link in vitro to in vivo – 1) mathematically describing selected ADME processes, and 2) integrating multiple pharmacokinetic processes within physiologically based models to track chemical disposition within the “in vivo” context.
A wide range of approaches exist for modeling selected ADME processes. Often termed predictive ADME or in silico ADME, QSAR/chemoinformatic approaches are currently available to predict, a priori, specific physicochemical properties, such as octanol-water partitioning, that can be important determinants of pharmacokinetics in vivo. Some software predict metabolism of chemicals, but additional development is needed to make quantitative predictions of the rates and amounts of metabolites formed by different species of animals including humans. Software predicting oral absorption exists, but further development is needed to address chemicals with wider ranges of properties or to more explicitly describe molecular level processes such as active transport (Liu & Pang, 2006). Some of these models have been developed by using data from specific in vitro systems, which may have substantial limitations for predicting occurrences in vivo.
Biologically based modeling approaches, notably physiologically based pharmacokinetic (PBPK) modeling, provide methods for combining multiple ADME descriptors coupled to physiological compartment parameters (i.e. tissue volume, flow, tissue partition constants) to make predictions of what would occur in vivo. There are often competing processes, such as elimination by metabolism and urinary excretion, which need to be accounted for to make good predictions for intact organisms of different species and strains, exposed by different routes of exposure under different exposure conditions. PBPK models have largely been developed as tools to retrospectively analyze multiple different datasets for a single chemical sometimes in several species, but increasingly there are efforts to develop models to make predictions for uncharacterized or less well characterized chemicals. Clearly, integrating models for individual processes into more comprehensive physiologically based models would be one avenue for development. In addition, biologically based models provide the opportunity to link pharmacokinetic or in vivo chemical disposition processes with effect processes (i.e. tissue-specific receptors) leading to changes in the dynamics of target organs. This linkage relies upon relating estimates of concentrations of specific forms of chemicals in specific organs (tissue dosimetry) or cell types of the intact organism with measurements or estimates of concentrations derived from causal chemical-effect relationships from the in vitro assays.
Combining ADME processes with in vitro assays for effects
The development and use of in vitro assays for effects requires consideration of the appropriate ways to characterize what happens to the chemical in that system and whether augmentation to reflect ADME processes in vivo is needed, for example transporters that influence distribution or enzymes that mediate metabolism. Beyond the specific endpoint for which the assay was developed, robust assays consider factors such as chemical stability for experimental lifetimes or unanticipated but confounding effects such as whether a chemical tends to stick to glass or plastics tubes rendering it unavailable to interact with the system for measuring effects. However, it is relatively common for in vitro assay methods to assume that the chemical added is soluble and homogeneously distributed. As methods to predict in vivo ADME processes evolve, it will also be essential to improve the measurement or prediction of chemical concentrations in the effects assays to make the linkages described above. Biologically based models linking the “pharmacokinetics” in the effects assays with description of the effects and the processes leading to them are needed, just as ultimately we will link predictions of in vivo chemical concentrations with predictions of in vivo effects (Teeguarden et al., 2007).
Characterizing and augmenting the ADME processes for in vitro assays is a well established need, though methods for implementation are not always available. It has long been recognized that bacterial assays for mutagenicity did not account for metabolism that would be expected in other organisms, particularly mammals. Thus, a compound could appear inactive in measures of DNA damage, but only the parent compound was being tested. This has frequently been addressed by adding the S9 fraction of rat liver homogenate to provide metabolic capability. Other approaches can be used and may be necessary, as was the case for studies with brominated di- and trihalomethanes, drinking water disinfection byproducts. They require conjugation with glutathione to initiate the metabolic pathway leading to mutagenic activity, but the conjugate is unable to cross the bacterial cell membrane so even if it was formed in the media, it would not get into the cell to cause damage (Landi et al., 1999; Thier et al., 1993). In contrast, bacteria genetically engineered to express the relevant enzyme inside the cell showed the mutagenic activity (Pegram et al., 1997, Their et al., 1993). Demonstration of the estrogenic activity of methoxychlor in vitro was similarly dependent upon addition of a metabolizing system (Bulger et al., 1985).
While it is commonly recognized that metabolic activating systems may be required to account for effects from highly reactive metabolites (often the focus of mutagenesis assays), it is less well recognized that the ability of both the parent chemical or metabolic progeny to cross membranes may also need to be accounted for (as described for the glutathione conjugates). Clearly, reliable predictions of effects in vivo will require predicting in vivo pharmacokinetics and then insuring that in vitro effects assays have appropriately mimicked those conditions (e.g., metabolites were tested or that the chemicals could interact with the effects assay system).
Substantial accomplishments over the past half century in both pharmacology and toxicology have dramatically increased our knowledge about ADME processes as well as our ability to quantitatively describe them using mathematical models to anticipate beneficial or adverse outcomes from chemical or molecular level perturbations. Progress will continue, but the challenge is shifting from the traditional reductionist approach of characterizing the pieces of biological systems to an integrative approach of describing the behavior of the biological system as a discrete entity. In vitro assays coupled with computer modeling are two essential components of a more integrated description of the in vivo system. Measurements in intact organisms, representing a range of animal species including humans, done with appropriate ethical considerations, will be critical to demonstrating that the combination of in vitro methods and computer model provides reliable predictions.
As the prediction methods become more commonly used for a wider range of chemicals, it will likely also be important to develop methods to monitor their success in humans or other species in the environment through biomonitoring and environmental sampling. While we should strive to develop the best prediction methods for which there is a need, we should always be aware that limitations on the completeness of our knowledge and the rigor involved in model development require that we continually evaluate those prediction methods and computer models in order to improve them and maintain their credibility. Finally, we can’t forget that a major gift of using these computational approaches coupled to ADME properties to address in vivo scenarios is the insight into complex scientific systems obtained through development of these models, not just the numbers they provide.
Disclaimer: Although this work has been reviewed and approved for publication by the US EPA it does not necessarily reflect Agency policy.
©2007 Hugh Barton
Bulger, W.H., Feil, V.J. & Kupfer, D. (1985). Role of hepatic monooxygenases in generating estrogenic metabolites from methoxychlor and from its identified contaminants. Mol. Pharmacol. 27(1), 115-24.
Coecke, S., Blaauboer, B.J., Elaut, G., et al. (2005). Toxicokinetics and metabolism. Altern. Lab Anim. 33 Suppl 1:147-75.
Landi, S., Hanley, N.M., Warren, S.H., Pegram, R.A. & DeMarini, D.M. (1999). Induction of genetic damage in human lymphocytes and mutations in Salmonella by trihalomethanes: role of red blood cells and GSTT1-1 polymorphism. Mutagenesis. 14(5), 479-82.
Liu, L. & Pang, K.S. (2006). An integrated approach to model hepatic drug clearance. Eur. J. Pharm. Sci. 29(3-4), 215-30.
Miners, J.O., Knights, K.M., Houston, J.B. & Mackenzie, P.I. (2006). In vitro–in vivo correlation for drugs and other compounds eliminated by glucuronidation in humans: pitfalls and promises. Biochem. Pharmacol. 71(11), 1531-9.
NRC (National Research Council). (2007). Committee on Toxicity Testing and Assessment of Environmental Agents. Toxicity Testing in the 21st Century: A Vision and a Strategy. National Academy Press.
Pegram, R.A., Andersen, M.E., Warren, S.H., Ross, T.M. & Claxton, L.D. (1997). Glutathione S-transferase-mediated mutagenicity of trihalomethanes in Salmonella typhimurium: contrasting results with bromodichloromethane off chloroform. Toxicol. Appl. Pharmacol. 144(1), 183-8.
Thier, R., Taylor, J.B., Pemble, S.E., Humphreys, W.G., Persmark, M., Ketterer, B. & Guengerich, F.P. (1993). Expression of mammalian glutathione S-transferase 5-5 in Salmonella typhimurium TA1535 leads to base-pair mutations upon exposure to dihalomethanes. Proc. Natl. Acad. Sci. USA. 90(18), 8576-80.
Teeguarden, J.G., Hinderliter, P.M., Orr, G., Thrall, B.D. & Pounds, J.G. (2007). Particokinetics in vitro: dosimetry considerations for in vitro nanoparticle toxicity assessments. Toxicol. Sci. 95(2), 300-12.