Toxicological Sciences 59, 17-36 (2001)
Copyright © 2001 by the Society of Toxicology
Forum |
In Vitro Human Tissue Models in Risk Assessment: Report of a Consensus-Building Workshop
a Center for Drug Evaluation and Research, Food and Drug Administration, Rockville, Maryland; b Department of Biopharmaceutics, Graduate School of Pharmaceutical Sciences, The University of Tokyo, Tokyo, Japan; c SRI International, Menlo Park, California; d Sanofi Synthelabo, Inc., Malvern, Pennsylvania; e Zeneca Pharmaceuticals, Alderly Park, UK; f Colorado State University, Center for Environmental Toxicology and Technology, Fort Collins, Colorado; g Institute for in Vitro Sciences, Gaithersburg, Maryland; h School of Pharmacy and Pharmaceutical Sciences, University of Manchester, Manchester, UK; i National Center for Toxicological Research, Food and Drug Administration, Jefferson, Arkansas; j Chemical Industry Institute of Toxicology, Research Triangle Park, North Carolina; k Université de Montréal, Santé Environnementale et Santé au Travail, Montréal, Québec, Canada; l In vitro Technologies, Inc., Baltimore, Maryland; m Karmanos Cancer Institute, Wayne State University, Detroit, Michigan; n University of Washington, Department of Pharmaceutics, Seattle, Washington; o National Cancer Institute, Toxicology and Pharmacology Branch, Rockville, Maryland; p Abbott Laboratories, Abbott Park, Illinois; q Novartis Institute for Biomedical Research, Department of Preclinical Safety, East Hanover, New Jersey; and r Eli Lilly & Company, Indianapolis, Indiana
Received May 30, 2000; accepted September 26, 2000
 |
ABSTRACT |
|---|
 TOP ABSTRACT IntroductionREFERENCES |
|---|
Advances in the technology of human cell and tissue culture and the increasing availability of human tissue for laboratory studies have led to the increased use of in vitro human tissue models in toxicology and pharmacodynamics studies and in quantitative modeling of metabolism, pharmacokinetic behavior, and transport. In recognition of the potential importance of such models in toxicological risk assessment, the Society of Toxicology sponsored a workshop to evaluate the current status of human cell and tissue models and to develop consensus recommendations on the use of such models to improve the scientific basis of risk assessment. This report summarizes the evaluation by invited experts and workshop attendees of the current status of such models for prediction of human metabolism and identification of drug-drug interactions, prediction of human toxicities, and quantitative modeling of pharmacokinetic and pharmaco-toxicodynamic behavior. Consensus recommendations for the application and improvement of current models are presented.
Key Words: risk assessment; human tissues; pharmacologically based pharmacokinetic (PBPK) modeling; hepatocytes; in vitro cell models.
|
Introduction |
|---|
|
TOPABSTRACTIntroductionREFERENCES |
|---|
Improved cell culture technologies and the recognition that studies with cultured human cells and tissues can play an important role in studies of toxicity, metabolism, and transport have resulted in an increasing utilization of these models in toxicological risk assessment and in quantitative modeling of metabolism and transport processes. Human tissue models are now used routinely in the pharmaceutical industry to predict human metabolism, select appropriate animal models, and identify potential drug-drug interactions. Human cell and tissue models are being used increasingly for prediction of organ-specific toxicity. Importantly, quantitative parameters derived from in vitro human cell and tissue models are being used in conjunction with physiological models to predict quantitative metabolism, transport, clearance, and pharmacodynamic outcome with increasing sophistication and success. Some of the ways that in vitro human tissue models are being used in the risk-assessment process are illustrated in Figure 1
.
|
In recognition of the potential value of the expanded use of such models for assessment of product and environmental risks, the Society of Toxicology's Task Force to Improve the Scientific Basis of Risk Assessment organized a workshop to bring together a multidisciplinary group of experts with experience in risk assessment, pharmacokinetic and pharmacodynamic modeling, and the use of human tissue models, with the objective of developing consensus recommendations on the use of human tissues to improve the scientific basis of risk assessment. The two-day workshop was held September 20–21, 1999 at the Turf Valley Conference Center in Ellicott City, MD. An invited group of experts was asked to summarize the scientific basis of the utilization of human tissue models to derive data to support risk-based decision-making and quantitative risk assessment, and to present specific recommendations for discussion and consensus-building by the multidisciplinary experts present at the workshop.
The workshop was attended by 117 individuals from government, academia, and industry, and included many experienced users of human tissue models. The attendance of experts beyond the invited core group was facilitated by the Hepatocyte Users Group of North America, which held its annual symposium at the same site in conjunction with the workshop.
The workshop was organized into three principal topic areas:
- Prediction of human metabolism and of drug-drug interactions
- Prediction of human toxicity
- Quantitative modeling of pharmacokinetic and pharmaco-toxicodynamic behavior.
The presentations and recommendations in each of these topic areas are summarized below.
Session 1: Prediction of Human Metabolism and Drug-Drug Interactions
Human Tissue Models in Vitro for Prediction of Metabolism and Drug Interactions
Dr. Jerry Collins of the FDA Center for Drug Evaluation and Research introduced the session on prediction of metabolism and drug interactions, summarized the current state-of-the-art in this area, and presented proposed recommendations for discussion and consensus-building. Based on the rapidly expanding database of human tissue metabolism studies in vitro, an overall consensus has emerged that in vitro studies with human cell and tissue models reliably predict human metabolism and drug-drug interactions. The use of primary human liver cellular preparations to study drug metabolism and to identify potential metabolic interactions has become a routine part of pharmaceutical development. Although not every drug-drug interaction is metabolism-based, understanding metabolism is the key to identifying and avoiding the majority of interactions. The occurrence of idiosyncratic reactions has been minimized by systematic utilization of such models for prospective predictions and intervention.
The overall goal of in vitro drug interaction studies with human tissue is to provide guidance to patients and prescribers. In extreme cases, it may be recommended that patients should not take a particular drug at all, or should not take the drug in combination with other drugs. In other cases, a modified dose would be more appropriate. Although the most reliable way to determine if there are metabolic drug-drug interactions is a direct clinical investigation, the number of potential combinations of drugs is beyond the capacity of resources for clinical investigations. Thus, the ability to substitute rapid, low-cost evaluation in vitro is an enormous advantage and has come into widespread use. In current practice, employing pivotal clinical studies that "bridge" to the metabolic data in vitro could reduce the number of clinical studies.
Over the last few years, a broad-based consensus has been reached on the types of metabolic information that are expected for approval of new drugs. The CYP-mediated metabolic pathways, in particular, are now extremely well understood. In many cases, data in vitro reliably predict the situation in vivo, and in these cases there is no need for clinical studies. Although positive findings obviously receive the most attention, it is noteworthy that the most common result is that there is no interaction. More consistent publication of results, especially by large drug-development programs, would enable all investigators to assess the strengths and weaknesses of contemporary tools. The outstanding progress in our understanding of the CYP family needs to be generalized to other metabolic pathways.
Although most experimental work in vitro is excellent, and experimental models have already demonstrated considerable value, some areas of apparent discord remain between studies in vitro and in vivo. Discordance is often traced to differences in protein binding in vivo and in vitro. Also, some researchers continue to use concentrations of drug so high that high-affinity, low-capacity pathways cannot be determined. Thus, the routine use of unbound drug concentrations within the range of physiological relevance, for comparisons in vitro with in vivo results, would avoid some elements of uncertainty.
In addition to intact cells and tissue slices, subcellular preparations of liver continue to be useful for screening. Some important limitations of subcellular preparations need to be appreciated. In the absence of an intact cell membrane, the concentration of drug in extracellular fluid (plasma) is not readily estimated from the concentration present in microsomal incubates. This is a particular problem for estimating Km and Ki values. The use of intact hepatocytes would seem to be an ideal approach to overcoming this problem, but further research is needed. Other participants at this workshop have emphasized the use of a combination of tools for metabolic studies. In particular, it is appropriate to use intact hepatocytes to confirm any pivotal subcellular findings, e.g., Km and Ki values, or the relative contributions of microsomal and non-microsomal pathways including those involving unknown enzymes.
Certain artifacts arise from the processing of subcellular preparations. Care must be taken to select preparation techniques that yield parameters comparable to those in vivo. For example, detergents can be used with microsomes to increase the catalytic activity of glucuronyl transferases. Our laboratory found that 4% bovine serum albumin produced a 20-fold increase in the glucuronidation of zidovudine (AZT) in human liver microsomes, and that the increased value was more consistent with clinically observed metabolism (Trapnell et al., 1998). For phenytoin hydroxylation by CYP2C9 in human liver microsomes, our laboratory found that the Km value from hepatocytes (in slices), but not from microsomes, was predictive of the clinical value (Ludden et al., 1997). For phenytoin, the use of 4% BSA, in combination with microsomes, also improved the Km value (with correction for protein binding).
Although knowledge obtained from metabolic studies of the human liver is being used very effectively, challenges remain. For some drugs, there is substantial extrahepatic metabolism, particularly in the gastrointestinal mucosa (Fitzsimmons et al., 1997). An increased appreciation has also been gained for non-metabolic processes that transport drugs into or out of tissues. Transport mediated by carriers such as P-glycoprotein mimics many of the properties of metabolism, and is another mechanism of drug-drug interaction. Our understanding today of even the most complex drug metabolism cases (e.g., HIV protease inhibitors) stands in sharp contrast to our inability to predict and manage even the simplest cases less than a decade ago. Thus, increased attention is needed to the identification of non-metabolic drug-drug interactions using appropriate human cell or tissue models.
Intact Hepatocytes in Drug Metabolism and Drug-Drug Interactions
Dr. Albert Li of In Vitro Technologies, Inc. presented a synopsis of the use of cultured intact hepatocytes for studies of drug metabolism and prediction of drug-drug interactions. As cells with an intact plasma membrane and complete metabolic pathways, isolated hepatocytes represent a physiologically relevant model of the liver. The use of human hepatocytes allows human drug properties to be evaluated in vitro. Applications include metabolite identification, metabolic-stability determination, and drug-drug interactions.
Hepatocytes are isolated from freshly procured human livers via collagenase digestion (Li et al., 1992). In general, the viability is high (over 70%) and the yield is approximately 10 x 106 cells per gram of liver. A rat liver yields 100–200 x 106 hepatocytes. Perfusion of large segments of the human liver can yield over 10 billion hepatocytes per isolation. Hepatocytes are used either as suspension cultures on the day of isolation or as monolayer cultures for experiments that require a longer duration (days to weeks). During longer-term hepatocyte culture, drug metabolizing enzyme activities are known to change (usually decrease) with time.
Because of the limited availability of fresh human livers for research, successful cryopreservation of human hepatocytes is a goal that will represent a major technological advance that can greatly enhance the general availability of this important experimental system (Li et al., 1999a). Cryopreservation of hepatocytes immediately after isolation allows one to store cells for an extended period of time without any significant loss of drug CYP-mediated metabolizing enzyme activities. Cryopreserved human hepatocytes have been reported to maintain the major cytochrome P450 isoform activities, as well as umbelliferone UDP-dependent glucuronosyltransferase and sulfotransferase activities (Li et al., 1999b). Quantitative and specific inhibition of CYP isoform activities by known inhibitors have been reported for cryopreserved human hepatocytes (Li et al., 1999b).
Because of their complete metabolic pathways and cofactors, hepatocytes are an appropriate system for metabolite identification and metabolic stability evaluation. These applications are supported by numerous reports on the similarity between metabolite profiles and intrinsic clearance values generated by hepatocytes in vitro and the animal in vivo (Bayliss et al., 1999; Iwatsubo et al., 1997; Kane et al., 1995; Zomorodi et al., 1995).
One of the major types of drug-drug interactions is the interference of metabolic clearance of a drug by a co-administered drug. One approach to estimating the drug-drug interaction potential of a new chemical entity is to evaluate its ability to inhibit the major drug-metabolizing enzyme activity such as that of cytochrome P450 isoforms. In the past, microsomes were routinely used for this area of research. Recently, intact human hepatocytes have also been investigated. While microsomes permit evaluation of the direct interaction of inhibitors with the microsomal enzymes, hepatocytes allow one to include all factors that would lead to inhibition of enzyme activities. Further, intact hepatocytes model the partitioning of drugs between the plasma and the liver. Results with intact hepatocytes, therefore, allow direct extrapolation from the extracellular concentration used in vitro to the extracellular concentration (i.e., plasma concentration) in vivo. For example, when human hepatocytes were applied to study ketoconazole inhibition of terfenadine metabolism, results were similar to those observed in humans in vivo (Li and Jurima-Romet, 1997).
Another important mechanism of drug-drug interactions is the induction of drug-metabolizing enzymes. A drug that induces a specific drug-metabolizing pathway would accelerate the metabolic clearance of a co-administered drug that is a substrate of the induced pathway. While the consequence of inhibitory drug-drug interactions is generally toxicity (due to an unexpected higher level of the affected drug), inductive drug-drug interactions may lead to loss of efficacy (due to lower than expected drug levels). Human hepatocytes have been found to be responsive to known human inducers in vivo and are now used routinely for the evaluation of P450 induction potential of drugs and drug candidates (Li et al., 1997; Pichard et al; 1991; Silva et al., 1998; Strom et al., 1996).
Human Tissue Utilization in Pharmaceutical Development
Over the past 10 years there has been an explosion in the use of human tissue and enzyme models in drug metabolism studies. The major driving force for the incorporation of human models of drug metabolism into drug development was the recognition that animal models were not always predictive of what was observed with patients (Wrighton et al., 1995). Dr. Steven Wrighton of Eli Lilly and Company summarized the application of these models in pharmaceutical development. Our understanding of the various enzymes involved in drug metabolism has increased greatly (Guengerich, 1995), and we now have a good understanding of the endogenous and exogenous factors that influence the catalytic activity of the drug metabolizing enzymes, in particular the cytochromes P450 (CYPs). Therefore, in vitro studies can be used to predict the factors that may affect the metabolic clearance of a new drug entity (NDE) and how an NDE affects the clearance of other drugs.
Four questions are often addressed with human tissue or enzyme preparations:
- Potential human metabolism of an NDE can be investigated. In these studies, the metabolic stability, metabolite profile, and species comparisons in the metabolites of an NDE are investigated.
- The ability of an NDE to interact with the drug-metabolizing enzymes, particularly the CYPs, can be determined. Thus, the ability of an NDE to inhibit or in some cases activate the catalytic activity of the CYPs is investigated. These results are used to predict the ability of the NDE to affect the activity of the CYPs in vivo.
- The enzyme(s) responsible for the formation of a metabolite of an NDE can be determined. By gaining an understanding of the enzyme catalyzing the formation of a particular metabolite, and by knowing the exogenous and endogenous factors that regulate the activity of that enzyme, it is possible to predict the effect of those factors on the metabolism of the NDE.
- The ability of an NDE to induce the catalytic activity of the CYPs can also be investigated.
Such studies are of interest because the animal models yield information that is often not predictive of what is observed in patients.
Three types of metabolism studies are performed:
- Metabolic stability, usually determined as loss of parent compound, yields information that the discovery scientist can use to order, by rank, the potential metabolic clearance of a series of compounds. Models for metabolic stability range from examining oxidative metabolism only, with hepatic microsomes, to more complete metabolism systems like hepatocytes or liver slices.
- The risk-assessment information obtained by metabolic-stability studies is measured in terms of relative metabolic lability of a series of compounds.
- Additional metabolism studies yield information on the profile of potential human metabolites of an NDE.
These studies use all types of human models from recombinant enzymes to tissue slices. From these types of studies, the identification of potential human metabolites and the relationship of these to metabolites obtained with similar models from animals is discovered. Thus, all these studies are important in assessing risk, in that they provide the toxicologist with such relationships. There are limitations on the information that these metabolism studies yield: metabolite formation (both type of metabolite and rate of formation) can vary with concentration and time in the incubations; often only liver systems are used, and the detection of metabolites may vary depending on the analytical system utilized.
Studies examining the ability of an NDE to interact with the drug-metabolizing enzymes have focused mainly on the CYPs. These studies yield IC50 or Ki values for the potent inhibitors of the catalytic activity of the CYPs. The CYP of interest is usually studied by examining the inhibition of a substrate that is selectively metabolized by a specific CYP in human liver microsomes. Specificity is gained by using recombinant systems that express only a single CYP. Recent work suggests that inhibitory constants may also be obtained by using selective substrates and human hepatocytes. However, the database for such hepatocyte studies is limited, and therefore the relationship of the constants generated with hepatocytes to those from other systems is not currently understood. The ability of an NDE to inactivate a CYP by mechanism-based inhibition or the formation of a metabolic-intermediate complex can also be determined by any of the above systems. The interaction studies yield risk-assessment information by providing guidance for, or prioritizing, the appropriate clinical studies. The limitations for these predictions include the difficulty of determining the appropriate in vivo concentration of the NDE to model, the clinical relevance of activation of the catalytic activity of the CYPs in the in vitro systems, and the participation of metabolites of the NDE in the potential interaction.
A great deal of information is known about the various factors that alter the activity of the CYPs. Therefore, the way that these factors influence the formation of a metabolite of an NDE in vivo can be predicted once the enzyme is identified. Thus, this information is used in risk assessment to guide clinical studies examining the pharmacokinetic variability of an NDE. The limitation of such studies is that, in some cases, the relationship between formation of a metabolite and the clearance of an NDE may not be known, and that an appropriate concentration of the NDE to assure the determination of the pharmacologically important CYP must be examined.
The ability of an NDE to induce human drug-metabolizing enzymes can also be predicted using human hepatocytes in primary culture. Many studies have shown human hepatocytes to respond to known in vivo inducers of the CYPs appropriately (e.g., Strom et al., 1996). Thus the results of studies with human hepatocytes are useful in risk assessment for guiding the appropriate clinical studies. Recently, reporter systems that are constructed with the regulatory elements of the inducible CYPs have shown promise as high-volume screens for induction. However, these systems have not been fully characterized and are not available for all CYPs. Thus, reporter systems to study induction need further evaluation and development. The major limitation to models of induction is that the relationships between concentration and time of treatment with the NDE in culture, and how these variables relate to the in vivo situation, are often not understood.
Role of Genetic Polymorphisms
Dr. Fred Kadlubar of the FDA National Center for Toxicological Research led the discussion of the application of human tissue models to the assessment of genetic polymorphisms. The human genome project has already provided a nearly complete sequence of the human genome and, within a few years, should identify single nucleotide polymorphisms (SNPs) that account for the genetic diversity of the human population. For the application of in vitro human tissue models for risk assessment, polymorphisms in drug and carcinogen metabolism enzymes have received the most attention. When using human tissues, the genotype of the donor may be of paramount importance in interpreting the data obtained. The current development of high throughput phenotyping and genotyping methods should soon be able to provide the necessary genetic information for population-based risk assessment.
The acetyltransferases, NAT1 and NAT2, represent some of the most studied and best-understood metabolic polymorphisms (Hein et al., 2000). They are regarded as both phase-I and phase-II enzymes in that they catalyze both N-acetylation of arylamines (detoxification) and the O-acetylation of their N-hydroxy metabolites (activation). NAT2 exhibits 5 major alleles that show wide ethnic variation: a wild-type allele and 4 others that involve point mutations in the coding region and result in a protein with low activity or stability. Thus far, another 21 minor alleles have been identified. NAT1 similarly exhibits 4 major alleles, with some 18 minor alleles identified. An updated listing of both NAT2 and NAT1 alleles can be found at the Web address http://www.louisville.edu/medschool/pharmacology/NAT.html.
The most widely studied of the phase-I enzymes are the cytochromes P450 (CYPs). Within this superfamily of about 30 different enzymes, only CYP1A1, CYP1A2, CYP1B1, CYP2A6, CYP2B6, CYP2C9, CYP2E1, CYP3A4, CYP3A5, and CYP3A7 are involved to a significant extent in carcinogen metabolism, while CYP1A2, CYP2A6, CYP2C8, CYP2C9, CYP2C18, CYP2C19, CYP2D6, and CYP3A4 are critical to the biotransformation of most commonly used drugs. The number of functionally different (i.e., high vs. low activity) genetic variants range from only one known variant for CYP2B6 to from 2 to 10 known for most of the CYPs, to 19 for CYP1B1 and >25 for CYP2D6. An updated listing of these CYPs can be found at http://www.imm.ki.se/CYPalleles/ (Ingelman-Sundberg et al., 2000; see also Table 1). Another non-CYP oxidative pathway of considerable importance in drug metabolism involves the flavin monooxygenases, and FMO3 has been found to exhibit functionally active alleles (Kang et al., 2000).
|
For the phase-II enzymes, the functional genetic diversity is rapidly being discovered and understood. These enzymes include the phenol sulfotransferases SULT1A1, SULT1A2, SULT1A3, and SULT1C1, the triiodothyronine enzyme SULT1B2, and the steroid sulfotransferases, SULT2A1 and SULT1E1 (Glatt et al., 2000
). In addition, the UDP-glucuronsyltransferases (UGTs) of the UGT1A and UGT2B families are now known to contain a variety of polymorphic variants (Burchell et al., 2000). The cytosolic glutathione S-transferases (GSTs), which include the alpha, mu, pi, theta, and zeta families, are each polymorphic with 2 to 4 variants known (Eaton and Bammler, 1999) that involve either a gene deletion (GSTM1, GSTT1) or a point mutation that affects regulation (GSTM3) or catalytic function (GSTP1). Each of these GSTs has also been associated with differences in cancer risk in a variety of studies. Finally, polymorphisms in the microsomal epoxide hydrolase (HYL1) and an NAD(P)H quinone reductase (DIA4 or NQO1) have now been identified (Lin et al., 1999; Omiecinski et al., 2000) and are further implicated in cancer risk or toxicity.
Although these polymorphisms are not usually, in themselves, strong risk factors for a disease (low penetrance), their high prevalence in the population can result in a highly attributable risk, particularly in combination with exposure to a drug, carcinogen, or toxicant. Accordingly, there is a clear and present need for high throughput (>1000 samples/day) genotyping, so that one can characterize human tissue samples for 10–100 genes or alleles concurrently, using small sample volumes and carried out at low cost. DNA microarray technology would appear to be ideally suited for these needs (Ames, 1999; Evans and Relling, 1999). One can design oligonucleotide probes fixed on glass surfaces, or "chips," to conduct allele-specific hybridization of PCR-amplified DNA samples. After washing and activation of a fluorescent or other suitable signal, the information can be readily captured by image analysis and sorted by bioinformatics systems. Moreover, the rapid screening of these genes should allow the selection of genetically heterogeneous groups for inclusion in clinical trials and should thereby increase our ability to protect public health.
However, it is strongly recommended that further consideration be given to the following: legal issues (discrimination by employers, insurance companies, or licensing agencies involving susceptible individuals; rights of individuals to litigate such discrimination); and ethical and moral issues (e.g., rights of individuals to choose whether or not to accept administration of a test for susceptibility, to know the results, to decide to accept a hazardous situation, to avoid particular drugs or exposures, or to take preventive measures; and knowledge of personal health risks that provoke anxiety and confusion, damage family relationships, and compromise the quality of life or lifestyle for the individual [Gostin and Hodge, Jr., 1999; Vineis and Schulte, 1995]).
Many benefits are expected to follow from our increased knowledge about individual genetic variation, resulting in a new paradigm in "individual risk assessment." Patient choice based on genetics regarding lifestyle decisions, including reproductive, environmental, and occupational choices, will become possible. Clinical benefits will include empowerment of individuals to prevent, treat, or cure disease by controlling their diet and lifestyle through knowledge of metabolic differences, exposure avoidance, own susceptibility, or other symptomatology. Finally, the development of individualized drugs may even become feasible.
Session 1 Recommendations
- While most work can continue to rely upon subcellular preparations such as microsomes, additional research should be conducted to determine the role of intact human hepatocytes in confirming key findings such as the relative importance of parallel pathways and the determination of the Km and Ki values.
- Binding of substrates and inhibitors to components in the incubation system must be considered in extrapolations to situations in vivo.
- Hepatocytes can be used to predict the in vivo metabolite profile. Comparison of profiles generated in human and animal hepatocytes can be used for the selection of the most appropriate animal species as a human surrogate for pharmacological and toxicological studies.
- Hepatocytes can be used to estimate in vivo metabolic clearance (e.g., intrinsic clearance), inhibitory potential of a xenobiotic on P450 activities, and induction potential of a xenobiotic on P450 isoforms.
- Cryopreserved hepatocytes can be used for short-term studies (hours) in suspension cultures for CYP-dependent metabolite profiling, in metabolic stability, and in enzyme inhibition.
- Cryopreserved hepatocytes have been characterized mainly for CYP isoform activities. Other drug metabolizing enzymes such as FMO, NAT, and GST should also be evaluated for a better understanding of the strengths and limitations of the system.
- Research should continue to optimize cryopreservation procedures. Although some cryopreserved hepatocytes can attach and form a confluent monolayer that is responsive to CYP1A and CYP3A inducers, application of cryopreserved hepatocytes in induction studies is hampered by the poor attachment efficiency of most cryopreserved cells.
- The use of human enzymes, subcellular fractions, and/or hepatocytes provides information on the metabolic stability, metabolite identification, and metabolite-formation rates, which may be compared with that observed with similar preparations from other species.
- In interaction studies, results primarily from enzymes and microsomal systems can be used to order, by rank, the potential of an agent to interact with the various CYPs.
- The weight of evidence from several complementary techniques, potentially using a combination of human enzymes, hepatic microsomes, and hepatocytes, yields the identification of the CYP(s) responsible for the metabolism of an NDE.
- Human hepatocytes in primary monolayer culture provide a viable model to examine the ability of an NDE to induce CYPs.
- High throughput genotyping using DNA microarray technology, or "pharmacogenomics," has the potential to improve and protect public health by allowing facile genetic screening of human samples for studies of adverse drug reactions, cancer susceptibility, and individualized drug dosing and prescribing.
- Unlike the majority of mammalian data that are derived from homogeneous species and strains, the heterogeneity of human populations creates special issues in the interpretation of data. While it is not feasible for each laboratory to have its own bank of human specimens from a fully diverse population, it is possible to generalize from what is already known about allelic frequencies and to use this information for the selection criteria for subjects in clinical trials.
- Further consideration should be given to legal and ethical issues involved in genetic screening to protect patient rights and confidentiality.
Session 2: Prediction of Toxicity
Primary Cell-Based Systems for Predicting in Vivo Toxicity
Dr. Charles A. Tyson of SRI International introduced and chaired the session on the utility of human tissue models for predicting in vivo toxicity. Human in vitro cell systems have been developed for a variety of research problems, including prediction of human toxicities. A review of the literature and society proceedings reveals approximately 100 citations involving cytotoxicity studies during the last 6.5 years. The cell-based systems most commonly used were from liver, kidney, bone marrow, and skin—in large part because these are common targets of adverse xenobiotic action, and because fresh donor organs or tissues and isolation methods are available. The principal applications were studies of mechanisms of action (for model validation, toxicity pathways, and antidote development), interspecies comparisons (for assessing which animal species respond most like man), and estimation of safe exposure margins (e.g., therapeutic indices) and levels in man.
This section focuses on results of studies with systems derived primarily from those organs and tissues that support these applications. A large number of studies involving drug- and chemical-induced effects on isolated cells and tissues from animals have been conducted over the last 2 decades. Based on this body of work, it appears that perfect in vitro models for replicating in vivo toxicological manifestations do not exist, although useful models do exist, and can provide valuable information for the toxicologist. Interspecies differences in response must be considered in all human-risk or -safety assessments, and studies in primary human cell or tissue cultures can aid greatly in interpreting these differences.
It is recognized that in vitro toxicology studies are more complex than metabolism studies, and that there are more variables to consider and optimize in designing the study. These variables include the optimal cell system, culture conditions and incubation time, toxicity indicators that best model the effect being studied, and the drug concentration range and interaction time with the tissue to approximate drug/chemical exposure time in vivo. For comparative studies, optimization is usually performed with animal-derived cell systems under conditions that replicate observed effects in those species; the inclusion of positive and negative reference compounds can help provide additional confidence in the results.
The verification that in vitro models respond in like manner to xenobiotic-induced toxicity in vivo is fundamental to their use for risk/safety-assessment data. Because acetaminophen (APAP) has been studied so extensively, it is worth evaluating how faithfully cell-based models have replicated its target-organ effects. Aspects of APAP-induced hepatotoxicity reproduced in vitro include: the requirement for metabolic activation, species differences in susceptibility and resistance, principal metabolic pathways and entities, modulation by P450 inhibitors and inducers (by manipulation of liver GSH levels and by various hepatoprotective agents), and salient features of the toxicity mechanism. In vitro studies clarified that the mechanism is more complex than originally thought—prominent features are protein binding, redox cycling, protein-sulfhydryl oxidation, lipid peroxidation, inhibition of mitochondrial respiration, DNA fragmentation and/or macrophage activation, depending on species, strain, APAP concentration/exposure, and incubation time, among other factors. These studies have yielded other important findings, including the following evidence. (1) The same covalent binding patterns occur in human and mouse hepatocytes as occur in vivo after APAP exposures, indicating clearly that hepatocyte cultures can faithfully reproduce key hepatotoxic events in vivo (Birge et al., 1990). (2) The toxicodynamic aspects of liver injury can be simulated in vitro (Carfagna et al., 1993; Harman and Fischer, 1983). (3) A previously unknown metabolism-independent pathway also exists that does not produce liver failure but also has potential life-threatening consequences (Esterline and Ji, 1989). In vitro studies with renal tubular fragments likewise indicate significant interspecies differences in response to APAP (Tyson et al., 1991).
Studies by a number of investigators support the conclusion that in vitro data can aid in deciding whether or not to continue development of a drug and can support regulatory submissions. Mechanistic and other study reports using human in vitro systems are relatively few compared with those performed with animal systems, primarily because of technical or supply limitations and the absence of verified in vitro models for some important toxicological endpoints. Consequently, there is considerable room for technological innovation to improve existing models and develop even more useful ones. A recent example is the development of a Kupffer cell/target cell co-culture system to assess the significance of interspecies differences in drug-induced liver macrophage activation for humans (Hamano et al., 1999). Adaptation of the sandwich method of co-culturing rat non-parenchymal and parenchymal cells (Bader et al., 1993) to human cells offers a potentially elegant model for unequivocally defining the key cell types responsible for toxicity (and designing more effective antidotal intervention) and for developing more sophisticated models of complex disease processes and injury.
Prediction of Eye and Skin Toxicity Using Specialized Tissue Models
It is critical that toxicologists ascertain the potential for new industrial chemicals, household products, personal care products, cosmetics, medical devices, etc. to cause local toxicity to the eyes or skin of humans after accidental or intentional exposure. This hazard-identification may be specifically required by a government agency, or it may be needed to determine labeling and packing categories, in order to determine proper handling procedures for plant workers, or to make internal safety decisions prior to product release. Historically, most of these data have been obtained from animal experiments. Although the animal data have provided a reasonable indication of human hazard, they suffer from the variability (Weil and Scala, 1971) associated with animal experiments and the fact that the animal response to irritants is not always the same as the human response. It is hoped that new in vitro human models will provide toxicologists with more reproducible and more relevant information, especially when the in vitro information is part of a hazard assessment process that uses information from a variety of sources (Curren et al., 1998). Dr. Rodger Curren of the Institute for In Vitro Sciences, assisted by Dr. John Harbell, discussed the specialized tissue models currently available for this purpose.
Recent advances in the bioengineering of reconstructed human tissue have allowed basic tissue culture systems to be supplemented with more sophisticated artificial organ models for both the cornea and the epidermis. Validation of these models for both in-house use and regulatory acceptance is now in progress (Balls et al., 1995, 1999; Kruszewski et al., 1997; Osborne et al., 1995).
Ocular Irritation
The animal model currently used most often to assess ocular irritation is the Draize rabbit eye test (Draize et al., 1944); thus, results from the Draize test are generally the standard against which new in vitro ocular tests are measured. Since injury to the cornea is weighted much more heavily than injury to other ocular tissues in determining the final eye irritation score in this test, most of the in vitro human models being developed address injury to that tissue. The models range from monolayer culture of human keratinocytes (Harbell et al., 1997) to 3-dimensional, stratified systems (Curren et al., 1997). Monolayer systems attempt to measure damage to the outer layer of the cornea using two different endpoints. One involves determination of viability after a short term exposure to materials that are likely to exert a rapid cytolytic action on cell membranes (e.g. surfactants), and the other used where the test material's toxicity is expected to require longer exposure to manifest its action (e.g. preservatives). Both of the assays have performed the best when used with relatively mild, water-soluble materials (Gettings et al., 1996), and less well when used with a broader range of chemicals.
To accommodate testing of non-water-soluble materials and to more closely approximate the exposure conditions that occur in vivo, 3-dimensional human cell models of the corneal epithelium have been developed. These models are constructed by growing human epithelial cells on a permeable membrane and allowing the cells to differentiate vertically to form a 4-5-cell thick layer of non-keratinized tissue. The surface of this tissue can be exposed directly to test materials, including powders and insoluble materials. A time-dependent evaluation of cytotoxicity (Osborne et al., 1995) or loss of barrier function (Kruszewski et al., 1997) is then made. Prediction models have been developed (Southee et al., 1999) that relate these endpoints to an eye-irritation score. This type of tissue can be obtained commercially from both U.S. and European sources (MatTek Corporation, Ashland, MA, U.S., and SkinEthic® Laboratories, Nice, France).
Recently, entire artificial corneas have been produced using only human cells (Griffith, 1998). These tissues—containing an epithelial, stromal, and endothelial layer—are initially transparent but become opaque after injury, similar to normal human corneas. As these models become more developed, they will likely provide relevant data on the depth of damage associated with the degree and duration of injury (Maurer et al., 1997).
Skin Irritation
Human cell models for skin irritation also vary in complexity from monolayer cultures of human keratinocytes (Botham et al., 1998) or fibroblasts to 3-dimensional models of the human epidermis. Monolayer cultures are usually used to test the potential action of individual ingredients on specific cells of the skin. Changes in proliferative capacity, differentiation (e.g., collagen synthesis), and cytokine expression are common end points. The artificial human tissues are most commonly used to test mixtures and final formulations. These 3-dimensional constructs can either be fabricated directly by the researcher or purchased commercially, as noted above. They are constructed similarly to the corneal models; however, the tissues are allowed to further differentiate so that a mature stratum corneum is formed. The stratum corneum provides a barrier function that is very important in determining the potential for an irritation reaction in humans.
Because the cultured tissue has its upper surface exposed to the air, test materials can be applied directly to the surface as is done in the intact animal or human. In general, subsequent measurements are made of time to 50% cytotoxicity and/or time course of cytokine release. These results are then correlated to either animal-skin irritation scores or the results of human clinical patch tests (ethical considerations allow a much wider range of materials to be tested in human dermal clinical trials than in human ocular trials).
Prediction of Hepatic Toxicity Using in Vitro Models
In the pharmaceutical industry, drug safety evaluation laboratories are charged with 2 distinct functions: aiding in discovery and selection of lead drug candidates and conducting human risk assessment. As part of the drug discovery process, assessment of toxic potential is used to prioritize compounds for further evaluation, thus saving resources by focusing on compounds that are more likely to succeed. Risk assessment is used both to establish a dose that is safe to administer to humans during clinical development and to evaluate risk due to prolonged exposure. Drug toxicity is conventionally determined by conducting animal studies in at least 2 species and examining changes in clinical signs, growth, serum chemistry parameters, and histopathology; human risk is then estimated by extrapolation from animal-study results. Dr. Roger G. Ulrich of Abbott Laboratories discussed how cultured primary hepatocytes are being used to predict the hepatotoxicity of developmental pharmaceuticals.
Hepatic toxicity, and in particular idiosyncratic hepatotoxicity, remains a major concern for the industry. To help avoid or minimize hepatic toxicity issues, compounds are often studied using primary cultured hepatocytes (Ulrich et al., 1995). These are typically isolated from rodents and other laboratory species. Over the past several years, considerable progress in human tissue availability and hepatocyte isolation and culture have been made—to the point that such studies with human hepatocytes or tissue slices are essentially routine in many companies. At present, primary cell assays have their greatest utility in studying mechanisms, in screening out undesirable compound characteristics, in comparative toxicity, and in identifying drug-drug interactions (applications in drug metabolism and transport have been addressed elsewhere in this report).
The reliability of hepatic toxicity assays for risk assessment, which requires extrapolation between species, depends on the relationship of the parameters measured in vitro to the actual mechanisms of toxicity for drug candidates in vivo. Determining toxic mechanisms requires a considerable commitment of resources, hence most investigations of this type are not conducted until after problems have been encountered (such as in clinical development or post-marketing). Nonetheless, several mechanism-based assays have been developed by a number of investigators. These assays can be divided roughly into 2 groups: functional assays that measure some aspect of cell physiology or function and gene expressional assays that measure fluctuations in gene products (mRNA or protein). In vitro hepatocellular physiological assays, based on defined mechanisms such as mitochondrial dysfunction (Ulrich et al., 1998a) or lysosomal dysfunction (Ulrich et al., 1991), have been described and can be used to predict toxicity in vivo. Physiological assays are typically low-throughput and, when using human hepatocytes, cost or availability may be restrictive. Strict attention must also be paid to exposure levels, which should be based on in vivo exposure parameters, including Cmax, AUC, and intended therapeutic levels. Exposure time is also a consideration, and improvement in long-term culture techniques (for example, Adams et al., 1995) may allow for chronic in addition to acute exposure.
The development and application of genomic tools hold the promise of both accelerating and refining toxicity evaluations (Todd and Ulrich, 1999). Information-intensive techniques such as hybridization microarrays can point towards a mechanistic basis for toxicity and provide markers for subsequent screening assay design or clinical monitoring. The potential utility can be seen by analogy to recent developments in cancer classification by gene expression monitoring (Golub et al., 1999); similar success in toxicology would provide for rapid and accurate diagnosis of mechanisms. Development of databases of "signature responses," or "response spectra," may help in hazard identification, and it should be possible to better predict human in vivo responses by comparing responses in vitro with those obtained for other species, as is presently done for physiological assays. Assays using techniques such as branched DNA, real-time PCR, or 33P-antisense hybridization (reviewed in Todd and Ulrich, 1999) relying on single or signature gene read-outs can provide for detailed examination of individual genes.
The expanded use of human hepatocytes in toxicity evaluation requires considerable improvements to be made in tissue procurement, isolation yields, and cryopreservation. Since availability is limited by access to donor tissue, several laboratories have developed methods for improved isolation and cryopreservation of human hepatocytes (reviewed in Ulrich et al., 1998b) extending the utility of this model well beyond the initial isolation time period. Many commercial sources of cryopreserved cells can be found (see Internet sites at www.invitrotech.com, www.cedracorp.com, www.hccc.com, www.clonetics.com, www.iiam.org). However, the cost of cryopreserved cells may be prohibitive for screening applications and, particularly to academic laboratories, for much-needed studies on the basic biology of the human liver. Further, recovery after cryopreservation is not optimal, generally less than 50% plating efficiency, with the majority of preparations showing poor recovery.
Improvements in assay and endpoint design are also needed. Establishing a set of well-characterized metabolic indicators for hepatocellular injury that are compared across species and across chemical classes would add greatly to our ability to predict potential human toxicity; these should include the major liability of mitochondrial dysfunction.
Finally, while toxicogenomics shows promise, characterizing human hepatocytes for gene expressional signature responses to reference compounds, comparison with other species, and providing wide access to this reference database are essential to determining the utility of this emerging concept. For practical applications in drug safety evaluation, the cost of this technology must be reduced, databases need to be established for gene responses to proprietary and public domain compounds, and verification/validation must be provided.
Molecular and Biochemical Endpoints to Define Species Differences in Organ Slices
During drug development, specific questions concerning compound-induced organ damage may arise from animal studies. The extrapolation of animal findings to predict compound safety in man can be bridged through the application of in vitro human tissue models. Organ slices represent an in vitro model in which tissue organization and functional heterogeneity are retained, allowing the researcher to investigate both the role of regional differences and the contribution of cell-cell interactions to the toxic outcome. Dr. Alison Vickers of the Novartis Institute for Biomedical Research discussed these models. The methodology of organ-slice preparation is similar regardless of the species, thereby facilitating cross-species and cross-organ comparisons, and overcoming the hurdle of isolating single-cell populations from various organs and species. Additionally, the methodology maximizes the use of the available tissue, which is critical when working with larger animals and humans.
The application of organ slices for assessing species susceptibility has primarily involved the use of liver tissue, and has paralleled the types of studies performed with hepatocytes. Such liver-slice studies include comparisons of animal to human responses, including the areas of genotoxic potential, adduct formation, cytochrome P450 induction, and compound-induced cytotoxicity (Bach et al., 1996; Hasal et al., 1999; Lake et al., 1997, 1999; Olinga et al., 1997; Vickers, 1997). Studies are now emerging that demonstrate the potential of slices to identify responses that require the presence of different cell types such as cytokine responses and cell-specific markers present in human tissue (Kayama et al., 1995; Luster et al., 1994). Additionally, clinical markers that are indicative of compound-induced organ damage can be identified. For example, the increased production of interleukin-6 and lipoprotein (a) by cyclosporin A was demonstrated in human liver slices, corresponding to elevated levels that have been found in transplant recipients (Vickers et al., 1998). In vitro human tissue studies can also reveal individual differences in responses to compound-induced organ damage, which is analogous to individual patients exhibiting side effects in clinical studies.
Extra-hepatic studies involving human tissue include the kidney, lung, heart, intestine, and lymphoid tissues (Bach et al., 1996). With non-hepatic tissues, both the preparation and culture conditions may require further improvements and may be accomplished through the measurement of organ-specific functions. Future challenges for assessing risk may also require performing co-cultures of liver slices with a nonhepatic tissue. To facilitate this application of human tissue, progress in the area of organ procurement as well as in long-term storage by preservation at both cold and cryogenic temperatures is needed.
Mechanistic studies that integrate genomics and proteomics, in conjunction with biochemical endpoints and histopathology, have the potential of identifying more sensitive and discriminating markers of organ damage. For example, gene expression changes induced by cisplatin can be demonstrated to precede decreases in liver- and kidney-slice viability and altered morphology (Hasal et al., 2000; Rose et al., 1999). Defining the molecular pathways leading to compound-induced changes in cell and tissue function and altered morphology will be the basis for predicting mechanisms of organ damage in man.
Human organ slices of both hepatic and non-hepatic tissues will contribute significantly to our understanding of the molecular and biochemical pathways leading to altered morphology, for the purpose of bridging the animal findings and defining species susceptibilities to organ damage, and for predicting safety in man.
Interspecies Comparisons Using a Bone Marrow Assay
Drs. Joseph Tomaszewski of the National Cancer Institute and Ralph Parchment of Wayne State University summarized the utility of human bone-marrow cultures for interspecies toxicity predictions. The development of drugs to treat cancer is inherently difficult because these compounds are typically the most toxic agents that are purposely administered to man. Selection of the starting dose for the Phase-1 trial, which is typically based on the maximally tolerated dose (MTD) of the most sensitive species, is critical in that the dose selected must not be toxic, while at the same time it must be high enough to give the patients therapeutic benefit. Even with the advent of the use of molecular targeting to develop new therapeutic approaches, the majority (>50% in the last 15 years) of cancer drugs still produce myelosuppression as the dose-limiting toxicity (DLT) in man. With this in mind, the National Cancer Institute, through the Small Business Innovative Research (SBIR) Program, initiated an effort a number of years ago to develop an assay to evaluate myelotoxicity in vitro, using progenitor cells from different species, including man. An assay developed by scientists at the Hipple Cancer Research Center, under this program, based on a clonogenic assay using the neutrophil progenitor termed CFU-GM, is considered particularly useful, because neutropenia is the most common form of myelosuppression produced in the clinic by cancer drugs (Parchment, 1998; Parchment et al., 1993).
A correlation has been found between the severity of neutropenia in the clinic with pyrazoloacridine and the inhibition of CFU-GM in vitro (Parchment et al., 1994). Subsequently, in vitro/in vivo correlations were found for the camptothecins (Erickson-Miller et al., 1997) and anguidine (Parent-Massin et al., 1998). A key finding in these studies was that the concentrations that inhibited CFU-GM by 90% (IC90) was a more predictive endpoint than the IC50 for MTD in animals and man. As a result, the NCI and its collaborators set out to determine whether this correlation could be expanded using 10 myelosuppressive drugs in a retrospective study. Each drug was evaluated in vitro using an exposure time that most closely approximated the schedule used in man and a comparison across the various species was made using the IC90. The results from the study indicated that the rank order of in vitro CFU-GM sensitivity from mouse, dog, and human correctly predicted the rank order of in vivo sensitivity (MTD) for all 3 species in 8/10 drugs where myelosuppression was dose-limiting in the species compared. The rank order of mouse vs. human was correctly predicted for 10/10 drugs. More importantly, the assay was able to detect the extreme sensitivity of human marrow for both fazarabine and fludarabine, drugs that unfortunately entered clinical trials at toxic doses. If the bone marrow data had been available prior to the initiation of the clinical trials with these agents and with topotecan, safe starting doses could have been predicted (Erickson-Miller et al., 1997; Schweikart et al., in press). Not surprisingly, the relative tolerance of drugs with target tissues other than bone marrow (renal, gastrointestinal, and cardiac) was not reliably predicted by the bone marrow assay.
Several pharmacological variables can cause in vitro/in vivo discrepancies (Parchment, 1998; Parchment et al., 1998), including protein binding, drug clearance from the system, and bioactivation or inactivation. In addition, drug solubility limits under in vitro testing conditions can cause large artifactual differences between species that require substantially different concentration ranges for testing, notably when the higher range is only partially soluble (Parchment, 1998). Published clinical-prediction models (Parchment, 1998 and Parchment et al., 1998) take those sources into account, and require chemical analysis of drug levels under in vitro conditions when the predictions may be used to adjust human dosages.
Due to these successes, the NCI has started integrating the CFU-GM test into its drug-discovery pathway. Specifically, the in vitro sensitivity of human CFU-GM (huCFU-GM) will be used to set an upper limit on human drug tolerance. Therefore, requiring efficacy at predicted human tolerated levels, either in vitro or in vivo (xenograft models), could become an additional criterion for distinguishing promising clinical candidates, i.e., those with efficacy against human targets at huCFU-GM-tolerated levels. This is a new characteristic upon which to base decision-making, and basically a therapeutic index-driven strategy to complement existing considerations of novelty and potency (Parchment et al., 1993). Future studies will determine whether the in vitro therapeutic index in the human system is predictive of clinical effectiveness. Some drugs are already being evaluated in the light of this concept (LoRusso et al., 1999).
Session 2 Recommendations
For applications of currently available human in vitro systems, the following were agreed upon:
- Mechanistic studies can provide essential information towards understanding the consequences of exposure for man. Knowledge of mechanisms is currently used to verify in vitro models, to facilitate antidote development, and to assist in rational product design.
- Comparative toxicity data from known or putative target organ systems in vitro and in vivo can be used to aid in the selection of appropriate animal models for toxicological studies or weighting of toxicological results from different species.
- Improved hazard and safety assessments are possible whenever the in vitro human database is sufficiently well developed to justify this application.
The following are needed to increase the value of human in vitro systems for risk assessment:
- Refine existing technologies and models to improve the isolation and preservation of human cells and tissues for research and increase availability of human tissue through the organ-donor network and tissue banks. In conjunction with this effort, conduct research into the understanding and characterization of cellular functions and processes leading to toxic outcomes.
- Develop criteria for assessing functional integrity of cells and tissues, including development and use of reference compounds for the intended toxicity application.
- Evaluate emerging technologies, including genomics and proteomics, and their applications where appropriate.
- Encourage research collaborations among government, industry, and academia to share knowledge, improve models, refine and speed acceptance and implementation of models and assays, develop databases, and provide independent verification. Databases should be made broadly accessible, for example, through Internet resources (toxicants need not be identified with a product name for their data to be useful for validation studies).
- Those developing new methods designed to replace methods currently required by a United States regulatory agency should be aware of and make use of the guidelines published by the U.S. Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM) and/or the European Center for the Validation of Alternative Methods (ECVAM) (NIEHS, 1997NIEHS, 1999; http://iccvam.niehs.nih.gov/iccvam.htm; http://www.ei.jrc.it/report/ecvam.html).
In order to expand human in vitro models for toxicology research, we must continue efforts to develop, verify, and implement new models for understanding and predicting a greater variety of toxicities in multiple organs from humans that are poorly predicted from animal studies.
Session 3: Quantitative Modeling
Prediction of in Vivo Pharmacokinetic and Pharmacodynamic Parameters
The session on quantitative modeling was introduced and led by Dr. Yuichi Sugiyama of the University of Tokyo Department of Biopharmaceutics. He pointed out that the area under the blood concentration-time curve (AUC) and the steady-state blood concentration (Css) are pharmacokinetic parameters directly related to the pharmacological and/or side effects of a drug. Therefore, it is very important to have information on the total body clearance (Cltot) and hepatic availability (Fh) that govern these values. Cltot is expressed as the sum of the clearances of tissues such as hepatic (Clh) and renal (Clr). AUC and Css, after the oral administration of a drug, can be governed by Cltot, Fh, and fraction absorbed via intestinal epithelium.
The prediction of in vivo drug disposition in the circulating blood and tissues from in vitro data has become possible by applying the "physiologically based pharmacokinetic modeling (PBPK)" or "clearance concept" (Fig.1
). The parameters that should be obtained in PBPK modeling include biochemical parameters such as the intrinsic metabolic activities (Clint, Km, Vmax), plasma membrane permeabilities, renal and biliary excretion clearances, plasma protein binding, tissue-to-plasma partition coefficients (Kp), and physiological parameters such as tissue mass and blood-flow rate. In most cases, the physiological parameters and their changes in pathophysiological conditions can be obtained from the literature. The degree of effectiveness that can be achieved through quantitative modeling of in vivo parameters based on rate constants derived from in vitro cell and tissue systems is illustrated in Figure 2.
|
The biochemical parameters can be obtained either from the in vitro experiments or so-called animal scale-up technique (Iwatsubo et al 1996
, 1997) (Fig. 1):
- Intrinsic metabolic clearance (Clint), Km and Vmax values for drug-metabolizing activities can be obtained by using human hepatocytes and/or subcellular fractions such as microsomes.
- Renal clearance can be predicted in most cases by use of animal scale-up methods based on the allometric relationship with body weight after correction for plasma protein binding.
- Tissue-to-plasma partition coefficient (Kp value) can be predicted based on the assumption that the tissue-to-plasma unbound concentration ratio (Kp,u) is the same between humans and animals. Tissue slices may be also useful to assess the Kp values.
- For drugs/chemicals whose membrane permeability is not rapid enough to allow the rapid equilibrium assumption (e.g., blood-brain barrier transport), estimation of membrane permeability is necessary in addition to the Kp values.
The method of predicting in vivo hepatic metabolic clearance from in vitro data will be described below. Kinetic parameters for enzymes [such as Km, Vmax and Clint (in vitro = Vmax/Km)] have been estimated from in vitro studies using isolated hepatocytes or subcellular fractions such as microsomes. These parameters can be converted into values for the whole organ with knowledge of the enzyme mass recovery for the preparations used. A combination of these Clint, in vitro values for unbound drugs, unbound fraction in the blood (fB), and hepatic blood flow (Qh) can give hepatic clearances (Clh), hepatic extraction ratio (Eh), and hepatic availability (Fh) based on appropriate organ perfusion models. Several kinetic models to describe the hepatic elimination of drugs have been used to relate Clh, Eh, Fh to Qh, fB, and Clint (Ito et al., 1998; Iwatsubo et al., 1996, 1997). The predicted values depend on the models selected, particularly for highly cleared drugs (Eh > 
0.8). For low-clearance drugs, a simple, well-stirred model can be used for the prediction. Either the distributed or the dispersion model gives better predictions of in vivo metabolism from in vitro experimental data, even for high-clearance drugs (Iwatsubo et al., 1996). Because of the relative ease of mathematical handling, the dispersion model with the dispersion number (DN) of 0.17 has been used frequently to predict the values of Clh, Eh, and Fh of high-clearance drugs (Ito et al., 1998; Iwatsubo et al., 1996, 1997).
Although it may thus be possible to predict hepatic metabolism in vivo from appropriate in vitro data, it is not necessarily easy to determine the appropriate experimental conditions in terms of adjustment of concentrations of co-factor(s), metal ions, and oxygen, or the validity of the initial velocity with the use of limited amounts of such valuable material as human liver. In this case, the scaling factor may be successfully used (Iwatsubo et al., 1996). In this approach, reference compounds are used. The reference compound should be metabolized by the same isozyme as the drug of interest, and the in vivo disposition of the reference compound should be already known. CLh for some reference compounds are calculated from in vivo pharmacokinetic data, and the value can be converted to a Clint, in vivo, for each reference compound, using the appropriate mathematical model (e.g., dispersion model). The ratio of Clint, in vivo, to Clint, in vitro is defined as the scaling factor. If the scaling factor obtained for each reference compound is comparable, we can then use this factor to the in vitro/in vivo scaling for the intrinsic clearance of the drug of interest.
Hepatic clearance is governed not only by intrinsic metabolic activities, blood protein binding, and hepatic blood flow rate, but also by membrane-transport activities such as the hepatic uptake and biliary excretion processes. These membrane permeabilities may also be estimated from in vitro experiments by using cell or membrane-vesicle systems (Kusuhara et al., 1998; Yamazaki et al., 1996);
Hepatic uptake clearance can be obtained by use of freshly isolated or short-term (<5 h) cultured hepatocytes. Prolongation of culture time sometimes causes the down-regulation of some transporters.
Biliary excretion clearance via bile canalicular membranes can be estimated by use of isolated bile canalicular membranes. The membrane transport activities can be measured in the presence of the appropriate driving force (such as ATP).
The validity of such predictions, based on in vitro data, can be confirmed by using experimental animals. The predicted time profiles of drugs in the target tissues or cells can be integrated with the pharmacodynamic information to obtain the degree of pharmacological effects and/or toxic effects. The pharmacodynamic parameters can be sometimes obtained from in vitro experiments with the use of cultured/isolated cells, immortalized cell lines, and/or receptor gene-expressed cell lines. Methods have been presented for predicting the clinical efficacy of anticancer drugs and the human maximal tolerated dose (MTD) using pharmacokinetic and pharmacodynamic considerations based on in vitro experimental data (Fuse et al., 1995).
Prediction of Pharmacokinetics and Toxicity from in Vitro Studies
Dr. Gregory Kedderis of the Chemical Industry Institute of Toxicology discussed the role of kinetic studies with isolated hepatocytes from rodents and humans in predicting the pharmacokinetics of xenobiotics that are primarily metabolized by the liver (Houston, 1994; Kedderis et al., 1993). Extrapolation of kinetic data determined in isolated hepatocytes in vitro requires knowledge of the kinetic mechanism of the enzyme involved in metabolism and a compartmental pharmacokinetic model of the intact animal or human. Most cytochrome P450-dependent reactions exhibit Michaelis-Menten saturation kinetics, with the kinetic parameters Vmax (the maximal rate of metabolism at infinite substrate concentration) and Km (the substrate concentration that gives one-half Vmax) (Kedderis, 1997a). Vmax values are extrapolated to intact animals based on the known cellularity of the liver, approximately 130 x 106 hepatocytes/g liver in mammals (Seglen, 1976). Km values can be used directly but xenobiotic solubility must be accounted for with partition coefficients.
Analysis of the kinetic data using PBPK models allows extrapolation of xenobiotic biotransformation across dose routes in a biologically realistic context (Andersen, 1991). In this compartmental pharmacokinetic approach, compartments represent organs or groups of organs with physiological volumes and blood flows between the compartments. PBPK models can account for physiological influences and internal tissue dose, allowing interspecies extrapolation. Human PBPK models can simulate a wide variety of exposure scenarios, allowing comparisons of tissue dosimetry between humans and animals. The bioactivation of a variety of rapidly metabolized xenobiotics (e.g., CYP2E1 substrates) is limited by hepatic-blood-flow delivery of the substrate (Kedderis, 1997b). The hepatic blood flow limitation dampens or eliminates the effects of interindividual variability and enzyme induction on xenobiotic bioactivation. For example, simulation of a 10-fold induction of CYP2E1 did not significantly increase the concentration of toxic furan metabolites in the liver through exposures to 300 ppm furan (Kedderis and Held, 1996). The limiting effect of hepatic blood flow on xenobiotic bioactivation needs to be considered before making conclusions about the significance of interindividual variations in enzyme activity or enzyme induction for toxicity.
Isolated hepatocytes can also be used to predict the hepatotoxicity of xenobiotics when appropriate concentrations and exposure times are used. The hepatocyte suspension-culture approach was developed from pharmacokinetic and pharmacodynamic considerations to mimic exposure of the liver in vivo to toxicants (Carfagna et al., 1993). Hepatocytes are incubated in suspension with the appropriate concentrations of xenobiotics for the appropriate times (determined from PBPK models), then placed in monolayer culture to express toxicity at 24 h. The suspension-culture approach gives concentration- and time-dependent toxicity similar to in vivo observations (Ammann et al., 1998