Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 92
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment 6 Application to Analyzing Variation in Human Susceptibility As a rule, humans vary in their responses to environmental factors because of variability in their genes and their genes’ epigenetic modification. Consequently, the same level of exposure to a chemical compound may give rise to different biologic effects in different individuals. For example, severe life-threatening toxicities can occur in some individuals treated with irinotecan, an anticancer drug. Although multiple genes play a role in irinotecan activity, polymorphisms in the UDP glycuronosyltransferase 1 family, polypeptide A1 (UGT1A1), enzyme have been strongly associated with irinotecan toxicity. Prospective screening of patients before chemotherapy could reduce the frequency of severe toxicities by alerting physicians to consider an alternative therapy (Marsh and McLeod 2004). With completion of the sequencing effort of the Human Genome Project, new opportunities have arisen to more fully characterize the genetic contributions to variation in human susceptibility to toxic effects of pharmaceuticals and other chemicals. The remarkable advances in our ability to rapidly detect thousands of genetic variations have led to high expectations for the ability to discover and then apply critical new information to understand human susceptibility to disease. More than 6 million single nucleotide polymorphisms (SNPs) have been identified and catalogued in public databases. Research efforts are now under way to identify which SNPs are associated with variation in chemical toxicity as well as drug responsiveness. Animal genome projects also provide opportunities to understand how genetic variation affects toxicity in other animal species. The new knowledge is likely to have scientific, clinical, and policy effects. Toxicogenomic technologies are expected to revolutionize strategies for predicting disease susceptibility and toxic response to environmental agents. Studies of
OCR for page 93
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment gene polymorphisms in the paraoxonase I gene (PON1) and the resulting differential response to organophosphate pesticides typify the type of genetic markers of toxic response that are likely to surface during the next decade (see Box 6-1). Another potential impact arising from toxicogenomics is the development of new classifications of disease subgroups. Most adverse reactions to chemical or therapeutic compounds have been classified by biochemical or clinical markers (frequently based on histopathologies). New molecular classifications of disease are likely to arise as researchers better understand the genomic, transcriptomic, proteomic, and metabonomic characteristics of disease. Finally, new knowledge about genetics and human variability in response is expected to enable greater tailoring of existing pharmaceuticals to patients to reduce toxicities and to better design new pharmaceuticals that produce fewer toxicities. STATE OF THE ART IN ASSESSMENT OF INDIVIDUAL VARIATION Toxicogenomic studies relevant to understanding human variability encompass various technologies and study designs. These range from investigations of variability in human gene expression profiles in responseto chemicals to large population-based cohort studies focused on identifying the genetic variations that influence sensitivity to chemicals. Dynamic modification of gene expression patterns without modification of the sequence, known as epigenetic phenomena, are also becoming better understood and characterized. This chapter reviews the state of the art in these areas and assesses future needs and challenges. Variation in Gene Sequence Gene-environment interactions refer to effects in which human genetic variability governs differential responses to environmental exposures such as the examples already discussed in this chapter. In this section, this concept is explored through a review of recent studies that identify genetic mutations associated with differential response to cigarette smoke and its association with lung cancer (Box 6-2). This study indicates that smoking is protective in some genotypic subgroups, which raises multiple ethical and policy related issues (see Chapter 11) yet typifies how gene-environment interactions may often appear counterintuitive with respect to our current knowledge base. This type of study also demonstrates the increased information provided by jointly examining the effects of multiple mutations on toxicity-related disease. Studies of polymorphisms in genes involved in Phase II metabolism (GSTM1, GSTT1, GSTP1) have also demonstrated the importance of investigating the combined effects of these variants (Miller et al. 2002).
OCR for page 94
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment BOX 6-1 Paraoxonase 1 Gene Polymorphisms and Occupational Exposure to Pesticide The paraoxonase 1 gene (PON1) encodes an enzyme involved in the metabolism of chlorpyrifos, an organophosphate pesticide, widely used in agricultural settings to protect crops from insects. Organophosphate pesticides affect the nervous system of animals and acute toxicity is characterized by nausea, diarrhea, impaired muscle function, and respiratory illness due to the effects on respiratory muscles (Blondell 1999). Studies of agricultural workers indicate that overexposure to organophosphates is relatively common (Ames et al. 1989; Ciesielski et al. 1994). Over the past two decades there has been extensive research on the relationship between pesticide toxicity and variations in the PON1 gene. Battuello et al. (2002) recently reviewed implications for workplace screening of these PON1 polymorphisms to assess whether identifying farmers who are more genetically susceptible to the adverse outcomes of pesticide use could reduce the disease burden in this population. Variations in the PON1 gene have been associated with both the amount and the type of paraoxonase 1 enzyme produced as part of the body’s normal detoxification system. Specifically, a mutation in the amino acid sequence of the gene at position 192 changes a glutamine to arginine (Q192R) and is associated with variable enzyme activity in the population at large—low activity, intermediate activity, and high activity. In addition, at least one mutation in the promoter region of the gene has been associated with a twofold increase in gene expression and, consequently, enzyme production. The combined effect of the Q192R polymorphism and the promoter mutation effect on enzyme production has been referred to as the PON1 status. Numerous animal studies have documented the relationships between these polymorphisms and pesticide toxicity, and variability in PON1 status is associated with a greater than 60-fold interindividual difference in the chlorpyrifos detoxification rate in humans (Li et al. 1993). Moreover, studies of the effects of the Q192R polymorphism in the human PON1 gene introduced into transgenic mice illustrate the power of using animal models to understand the effects of human genetic variations on differential toxicity to pesticides in genetic subgroups of the human population (Cole et al. 2005). In humans, the R192 alloform has a higher catalytic efficiency of hydrolysis than does the Q192 alloform for chlorpyrifos oxon (CPO), the oxon form of the pesticide chlorpyrifos (CPS). Transgenic mice expressing PON1 Q192 were significantly more sensitive to CPO, and to a lesser extent CPS, than were mice expressing PON1 R192. Dose-response and time course of inhibition studies in these transgenic mice suggest that humans expressing only the PON1 Q192 allele would be more sensitive to the adverse effects of CPO or CPS exposure, especially if they are expressing a low level of plasma PON1 Q192 (Cole et al. 2005). Although there is ample biochemical evidence of a significant genetic contribution affecting workers’ health, there has not been substantial epidemiologic research in populations of exposed workers to quantify genetically influenced risk for this group.
OCR for page 95
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment BOX 6-2 Multiple Genetic Factors Influence Response to Cigarette Smoke Tobacco smoke contains a broad array of chemical carcinogens that may cause DNA damage. Several DNA repair pathways operate to repair this damage and the genes within this pathway are prime biologic candidates for understanding why some smokers develop lung cancers, but others do not. Zhou et al. (2003) examined variations in two genes responsible for DNA repair for their potential association with lung cancer. Briefly, one putatively functional mutation in the XRCC1 gene and two putatively functional mutations in the ERCC2 gene were genotyped in 1,091 patients with lung cancer and in 1,240 controls. When the patients and controls were stratified into heavy smokers and nonsmokers, Zhou et al. (2003) found that nonsmokers with the less frequent mutant genotype had a 2.4 times greater risk of lung cancer than those with the more prevalent genotype. In contrast, heavy smokers with the less frequent mutant genotype had a 50% reduction in lung cancer risk compared with their counterparts with the more frequent genotype. When the three mutations from these two genes were examined together, the extreme genotype combination (individuals with five or six mutations present in their genotypes) was associated with a 5.2-fold greater risk of lung cancer in nonsmokers and a 70% reduction of risk in the heavy smokers compared with individuals with no putatively functional mutations. The protective effect of these genetic variations in heavy smokers may be caused by the differential increase in activity of these protective genes stimulated by heavy smoking. Similar types of gene-smoking interactions have also been found for other genes in this pathway, such as ERCC1. The findings from this study have not been replicated. It is an illustration of the counterintuitive findings that are likely to emerge in studies of gene-environment interactions. Gene-gene interactions are another important area of research for understanding human susceptibility to chemical sensitivity. This refers to situations in which one gene modifies the effect of another gene on disease or other adverse effect. In a recent study, McKeown-Eyssen et al. (2004) associated a gene-gene interaction between NAT2 and CYP2D6 enzymes with multiple chemical sensitivity. These results suggest that individuals with the rapid-metabolizing forms of both enzymes were 18 times more likely to have chemical hypersensitivity than individuals with normal metabolizing forms of these enzymes. Gene-gene interactions between CYP2D6 and another P450 enzyme (CYP3A4) have also been found to influence the metabolism of commonly used pharmaceutical agents (Le Corre et al. 2004). If we are to adequately understand the continuum of genomic susceptibility to toxicologic agents that influences public health, more studies of the combined effects of multiple mutations are needed. The current emphasis on identifying single gene mutations associated with differential response to environmental exposures only delays understanding the distribution of genetic and genomic risks in human populations. Advances in bioinformatics can play a
OCR for page 96
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment key role in understanding combined effects of multiple mutations. For example, methods to screen SNP databases for mutations in transcriptional regulatory regions can be used for both discovery and functional validation of polymorphic regulatory elements, such as the antioxidant regulatory element found in the promoter regions of many genes encoding antioxidative and Phase II detoxification enzymes (X. Wang et al. 2005). Comparative sequence analysis methods are also becoming increasingly valuable to human genetic studies because they provide a way to rank-order SNPs based on their potential deleterious effects on protein function or gene regulation (Z. Wang et al. 2004). In addition, methods of performing large-scale analysis of nonsynonymous SNPs to predict whether a particular mutation impairs protein function (Clifford et al. 2004) can help in SNP selection for genetic epidemiologic studies and can be used to streamline functional analysis of mutations statistically associated with response to toxicologic agents. The use of bioinformatics in identifying and analyzing the biochemical and physiologic pathways (for example, systems analysis) by which gene-environment interactions occur is another key role toxicogenomics can play in helping genetic epidemiologic studies move beyond simple statistical association. From a public health point of view, the impact of gene-environment studies on our understanding of the distribution of environmentally induced disease could have major ramifications for public policy. For example, a recent study of drinking water contaminants (commonly associated with trihalomethanes from chlorination) and CYP2E1 gene mutations found a significant gene-environment interaction that affects fetal growth (Infante-Rivard 2004). Chlorination by-products in drinking water come from reactions between chlorine and organic material in the source water. Studies of the putative mechanisms underlying such an association are essential to establishing the biologic plausibility of epidemiologic information, with integrated use of transcriptomic, metabonomic, or proteomic technologies to understand environmentally induced disease. These types of studies are likely to play a major role in translating basic genetic epidemiologic science into public health policies and practices. Epigenetic Variability Variations in susceptibility are due not only to polymorphisms in DNA sequence. Epigenetics refers to the study of reversible heritable changes in gene function that occur without a change in the sequence of nuclear DNA (see Chapter 2). Differences in gene expression due to epigenetic factors are increasingly recognized as an important basis for individual variation in susceptibility and disease (Scarano et al. 2005). The best known mechanism for epigenetic regulation of cell phenotypes is DNA methylation, which turns off a gene or gene region by changing the chemical structure of the DNA (Jaenisch and Bird 2003). For example, as a normal part of human development, genes are turned on and off by methylation processes stimulated by other gene products in the embryo,
OCR for page 97
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment fetus, newly born infant, adolescent, and aging adult. Environmental factors such as infection, diet, and chemical exposures are known to affect gene methylation (Sutherland and Costa 2003). Anway et al. (2005) investigated the impact on rats of transient in utero exposures to two endocrine disruptors, vinclozolin (a fungicide commonly used on crops) and methoxychlor (a pesticide used as a replacement to dichlorodiphenyltrichloroethane [DDT]). Mothers were treated at a critical time during gonadal sex determination or a later embryonic period. The adult male offspring developed reduced spermatogenic capacity (decreasing sperm count and spermatogenic cell viability) and decreased fertility in this and two previous studies (Cupp et al. 2003; Uzumcu et al. 2004). In the latest study, although only the original gestating mother for the first generation was treated with vinclozolin, diminished male fertility was transmitted to the subsequent four generations (F1 to F4) when offspring males were crossed with offspring females from mothers that were exposed only once. Methoxychlor had similar effects but they extended only to the F1 and F2 generations. Further analysis indicated that these were male germ line effects associated with altered DNA methylation patterns. The study thus suggests that environmental factors can induce an epigenetic transgenerational phenotype through an apparent genetic reprogramming of the male germ line. (However, for this to be truly considered to be epigenetic, germline DNA mutations must be ruled out, which would require sequencing the entire genome). Nickel, cadmium, and xenobiotics (such as diethylstilbestrol) have been shown to affect gene methylation (Sutherland and Costa 2003; Bombail et al. 2004). As this field progresses, it will be important to integrate epigenetic and genetic approaches to better model the risk of disease caused by environmental toxicants. Models of how to merge epigenotype and genotype information are now starting to emerge (Bjornsson et al. 2004) and more theoretical, as well as applied, work is needed in this area of toxicogenomics. Furthermore, work on integrating epigenetic data, both the causes and consequences of epigenetic modification, into dynamic system biology models of the regulation of gene expression, proteomic, and metabonomic profiles is also needed. Gene Expression Variability The sections above describe individual variability as assessed by studies that look at variations in gene sequence or epigenetic modification among individuals. Another way to assess human variability is to look downstream of the gene sequence or its epigenetic modification to the amount of mRNA expressed by the genes, examining differences in amount expressed rather than just differences in what is expressed. The variability in gene expression can reflect individual variability due to mutations in the gene, its promoter or other regulatory regions, and other modifications of expression such as epigenetic effects.
OCR for page 98
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment Several landmark studies have shown that gene expression may profoundly vary due to gene sequence variation. Lo et al. (2003) investigated allele-specific expression of 602 transcribed SNPs and found that 54% showed preferential expression of one allele over another, frequently greater than a fourfold difference in expression between the two alleles. Similarly, Chueng et al. (2002, 2003) demonstrated that the expression level of genes is highly heritable in humans, with one-third of the genes with heritable expression patterns showing evidence of mutations that directly affected transcription levels. With transcriptomic profiles and genomic data simultaneously provided, new insights into the causes of variability in gene expression are being discovered. This type of research could explain variation in toxic responses to chemical agents that is not due to underlying differences in gene sequence. A study of SNP variation in human carboxylesterases illustrates how research on both gene expression and genetic sequence together could be used to study human variation in drug responsiveness. Human carboxylesterases 1 and 2 (CES1 and CES2) catalyze the hydrolysis of many exogenous compounds and play an important role in the metabolism of toxic chemicals in the body. Alterations in carboxylesterase sequences could lead to variability in both the activation and inactivation of drugs. Marsh et al. (2004) sequenced the CES1 and CES2 genes in individuals in European and African populations, identifying novel SNPs in CES1 and CES2. At least one SNP in the CES2 gene was associated with reduced CES2 mRNA expression. In summary, functional analysis of novel mutations found to affect gene expression patterns could provide important insight into variation in drug responsiveness. Using Animal Models to Identify and Evaluate Susceptibility Genes Animal models offer important experimental research opportunities to understand how genetic factors influence differential response to toxicologic agents. Animal models are advantageous as a first line of research because they are less expensive, less difficult, and less time-consuming than human studies. In addition, animal studies can address questions that are almost insurmountable in human studies, such as questions about sporadic effects or effects that cannot be adequately examined for sex linkage because of sex bias in employment. Because response is most often quantitative, theoretical models of the cumulative action of mutations in multiple genes and multiple gene-environment interactions have been used to identify which regions of animal genomes are related to response. Genetic analysis of these complex quantitative traits by classic Mendelian methods is not possible. Because of advances in statistical approaches capable of analyzing extensive genetic data, rapid quantitative trait mapping in animal and human genomes has become more feasible (Lander and Botstein 1989; Silver 1995; Manly and Olson 1999). When combined with selective breeding designs in model species, this approach identifies genes in the
OCR for page 99
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment model species that can then be mapped onto human chromosomes by using comparative genomic and bioinformatic methods. The mouse offers several advantages in the initial determination of genetic traits that control human conditions. First, inbred and wild-derived inbred mice allow research to focus on the mechanisms of resistance and clear distinctions in susceptibility among inbred strains of mice. Study of inbred mouse strains can also be advantageous because, unlike humans, their polymorphisms often become “fixed” in a population (carried by all the mice) by inbreeding of the strain. Moreover, resistance to one disease may lead to susceptibility to another. Because the mouse genome has been mapped and is largely (>97%) identical to the human genome, studies to identify new genes for susceptibility can be efficiently accomplished in mice, significantly accelerating research on homologous genes in humans. A number of studies illustrate how these advantages have enabled the mouse to be a powerful model for the dissection of genetic factors contributing to a number of complex diseases, ranging from immune disorders and cancer predisposition (Todd et al. 1991; MacPhee et al. 1995; De Sanctis and Drazen 1997) to coagulation disorders (Mohlke et al. 1996). Genomic animal approaches also increased the ability to uncover genes not previously associated with susceptibility to adverse effects from ozone (Kleeberger 1991; Kleeberger and Hudak 1992; Kleeberger et al. 1993a,b, 1997), lipopolysaccharide-associated lung injury (Arbour et al. 2000; Kiechl et al. 2002; Cook et al. 2004), and acute lung injury (Prows et al. 1997, 1999; Prows and Leikauf 2001; Wesselkamper et al. 2005). These discoveries, although requiring considerable time and effort, yield new information about the biology of the disease process underlying environmental injury and could lead to further detection of human mutations and their functional significance. Understanding the role of a genetic association in mice can lead to identification of analogous human mutations or analogous alterations in human protein function. Studies of the SLC11A1 gene for proton-coupled divalent metal ion transporters best illustrates the concepts of using inbred mice and the effects of a single mutation on multiple traits (Liu et al. 1995; Fortier et al. 2005). Nucleotide sequence analyses of the SLC11A1 cDNA in 27 inbred mouse strains that were either resistant or susceptible to intracellular parasite infection demonstrated that susceptibility was associated with a mutation that caused a glycine-to-aspartic acid amino acid substitution in the corresponding protein product (Malo et al. 1994). The human SLC11A1 gene encodes a 550-amino acid protein showing 85% identity (92% similarity) with mouse SLC11A1. Although the mouse susceptibility polymorphism was not found in the human gene (Blackwell et al. 1995), other human polymorphisms associated with disease resistance were found. Bellamy et al. (1998) examined the role of SLC11A1 in tuberculosis. In a case-control study in Africa, four SLC11A1 polymorphisms were significantly associated with tuberculosis susceptibility. Searle and Blackwell (1999) found a polymorphism that confers resistance to infection and was also associated with
OCR for page 100
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment chronic hyperactivation of macrophages. They hypothesized that this polymorphism was functionally associated with susceptibility to autoimmune disease. Analysis of these polymorphisms in patients with rheumatoid arthritis found that increased susceptibility to arthritis was associated with the mutation that conferred resistance to tuberculosis (Shaw et al. 1996; Bellamy et al. 2000). In these examples, understanding the role of a genetic association in mice led to a hand-in-hand assessment of associations in humans. Although the same mutations were not identical across species, knowledge of mutations that can alter protein function (in mice) or gene expression (in humans) were linked by the functional role this gene played in infection and arthritis. In another study with inbred mouse strains, Arbour and coworkers (2000) compared the susceptibility of 40 strains to bacterial lipopolysaccharide administration. Genetic linkage analysis and transcriptional profiling identified the TLR4 gene, which encodes the toll-like receptor 4 as the gene primarily responsible for variation in susceptibility. Moreover, the toll receptor showed variation not only among differentially susceptible mouse strains, but variants were also shown to determine differential susceptibility in humans. This is an excellent example of how toxicogenomic investigation of interstrain response variability can be used to study the effects of human variability. Of particular utility to this approach is the set of recombinant inbred mouse panels that the National Institutes of Health has generated (Churchill et al. 2004). Because each strain represents a random assortment of susceptibility loci, the use of these panels will be particularly helpful in elucidating the effects of quantitative trait loci of susceptibility. Recently, Churchill et al. (2004) proposed an initiative entitled the Collaborative Cross to promote the development of a genetically diverse set of mouse resources that can be used to understand pervasive human diseases. The goal is to breed current inbred mouse strains to create a more genetically heterogeneous, yet stable, resource for examining polygenic networks and interactions among genes, environment, and other factors. Existing resources optimized to study the actions of isolated genetic loci on a fixed background are less effective for studying the complex interactions among genetic factors that are likely to give rise to a substantial proportion of human susceptibility. The Collaborative Cross will provide a common reference panel specifically designed for the integrative analysis of complex systems and has the potential to change the way animal models can be used to understand human health and disease. New strategies for using animal models of toxicity are likely to yield valuable information for assessing therapeutic strategies and genetic differences that alter susceptibility. RISK ASSESSMENT Integrating genetics into the risk assessment process, including protecting sensitive populations, will require more directed research to support estimates of
OCR for page 101
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment key parameters such as uncertainty factors and on physiologically based pharmacokinetic (PBPK) models associated with genetically influenced human variability. Risk assessment methodologies currently assume a 10-fold range in sensitivity to toxics in the human population and use a 10-fold uncertainty factor to account for this variability. Developing literature clearly indicates that the range in human sensitivity to toxic exposures has a genetic component for at least some classes of compounds.1 For example, the Glu-69 polymorphism in the HLA-DP6 gene has been shown to lead to unusual sensitivity to beryllium, the PON1 gene appears to be important to the metabolism and detoxification of organophosphate pesticides, and the NAT2 gene is associated with slow acetylation of arylamine compounds. A review of human data on therapeutic drugs indicates that the metabolism and elimination of most drugs are also subject to wide variation (Renwick and Lazarus 1998). More recently, a review of human variability in different routes of metabolism of environmental chemicals suggests a range greater than 10-fold in individual susceptibility to some chemicals (Dorne et al. 2005). In general, improvements in risk assessment are expected as more research is done to determine the range of allele frequencies and the impact of genetic variability associated with different ethnic groups as well as the elderly, children, and neonates. Currently, the default kinetic uncertainty factor of 3.16 would not be conservative enough to cover the variability observed in all subgroups of the population for compounds handled by monomorphic pathways versus polymorphic pathways (for example, CYP2C19 and CYP3A4 metabolism in Asian populations; CYP2D6, CYP2C19, NAT, and CYP3A4 in the elderly; and CYP2D6 and CYP2C19 in children). Kinetic data available in neonates compared with healthy adults for four pathways (CYP1A2, CYP3A4, glucuronidation, and glycine conjugation) demonstrated that the default value of 3.16 would be adequate for adults, whereas uncertainty factors greater than 12 would be required to cover up to 99% of neonates (Dorne et al. 2005). In a recent paper, Haber et al. (2002) analyzed the potential contribution of mutations in enzymes influencing the disposition of four different types of compounds—methylene chloride, warfarin, parathion, and dichloroacetic acid—by PBPK modeling. They identified several key uncertainties regarding whether genetic mutations are an important source of variability in human susceptibility to environmental toxicants. The key issues they identified include the following: (1) the relative contribution of multiple enzyme systems, (2) the extent of enzyme induction/inhibition through coexposure, (3) differences in mutation frequencies across ethnic groups, (4) the lack of chemical-specific kinetic data for different genetic forms of the enzymes, (5) the large number of low-frequency mutations with significant effects, and (6) the uncertainty caused by differences 1 Genetic variations in susceptibility are due not only to polymorphisms in DNA sequence. As discussed above, differences in gene expression due to epigenetic factors are increasingly recognized as an important basis for individual variation in susceptibility and disease (Scarano et al. 2005).
OCR for page 102
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment between in vitro and in vivo kinetic data. There are critical gaps in the data required to assess and integrate genetic information into PBPK modeling to quantitatively assess its impact on population variability. An example of how genotype-specific PBPK data could be integrated into risk assessment for a population is illustrated by the work of El-Masri et al. (1999), who modeled the effects of GSTT1 mutations on the risk estimates for dichloromethane toxicity in humans. Dichloromethane is used in many industrial settings, including agriculture and food processing. By modeling the effect of genetic variability in the physiologic and biochemical processes underlying risk estimates, they (El-Marsi et al. 1999) and others (Andersen et al. 1987) concluded that the intrapopulation variability caused by the protective effect of the mutation can significantly increase the variability in the safe dose estimate of dichloromethane in a population. Other work also illustrates how understanding human kinetic variability could influence risk assessment (Dorne et al. 2002; Timchalk et al. 2002; Meek et al. 2003b). CHALLENGES There are several significant challenges to using toxicogenomic technologies to understand variation in individual or population susceptibility to chemical and pharmacologic compounds. First, the genetic architecture of human chemical sensitivity is complex. There are likely to be only a few rare instances when single gene mutations convey significant sensitivity to normal levels of exposures regardless of other contexts (e.g., Weber 1997). Much more frequently, there will be many genes with moderate or small effects on susceptibility, which in combination define susceptibility to a toxic agent. Interactions between gene variations, as well as additional gene-environment interactions and epigenetic processes, are likely to play a significant role in determining sensitivity to particular environmental exposures. This etiologic heterogeneity poses substantial challenges from both a methodologic and a risk assessment point of view. Second, the understanding of the distribution of SNPs in the human gene pool is only beginning, and accurately typing large numbers of SNPs remains a work in progress. Multistaged research strategies (for example, linkage analysis to identify potential genomic regions followed by positional candidate gene studies or genome scans using tag SNPs followed by fine SNP mapping; see Chapter 2) are used to identify the set of genes and their variations that are most significantly associated with differential toxicity to chemical and pharmaceutical compounds. These multistaged research approaches have at their core an assumption that single mutations will have statistically significant, context-independent effects (that is, they will have the same effect in many different populations or contexts). True multigene models of susceptibility have not been practically obtainable to date and need to be a major focus of the next generation of toxicogenomic studies.
OCR for page 103
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment Third, most large-scale environmental epidemiologic studies have not embraced genomic questions and toxicogenomic technologies as a part of their investigative approach. For example, clinical drug trials do not systematically collect and store blood for toxicogenomic or pharmacogenomic analyses. In some cases, biologic samples are available, but the funds and expertise for conducting the genomic studies in these population resources are limited or difficult to coordinate. Epidemiologic research in toxicogenomics is difficult because it requires multidisciplinary state-of-the-art teams of experts to measure genetic or toxicogenomic-derived markers, to measure environmental exposures, and to conduct clinical assessments, which take coordinated efforts among many different scientific disciplines. Unlike the toxicogenomic studies being carried out in animal models, which often rely on inbred strains, humans have a much higher level of genetic variability. This natural human variability makes large-scale epidemiologic studies imperative to scientific and policy development and it makes the understanding of disease risk incredibly complex. Fourth, many researchers are finding that results from genetic association studies are not consistent from study to study (Hirschorn et al. 2002). There are several reasons for this lack of replication across studies, ranging from the statistical issues that arise from small studies (stemming from the expense of these technologies) to differences across studies in the distributions of underlying genetic variations, exposure, and accumulated genomic changes that occur at the epigenetic level. Studies of cohorts large enough to offset the small-sample random sources of variation from the important biologic variations will increase the power to identify reliable toxicogenomic predictors of susceptibility. There are also numerous genetic epidemiologic studies that have developed transformed cell lines from human lymphocytes as a way to create inexhaustible supplies of DNA for genomic studies. These biologic samples could provide extremely valuable experimental material to determine the impact of interindividual variation in genes in response to industrial and pharmacologic compounds through in situ studies. Further research in this area is needed to determine how the results from cell line studies are translatable to human health effects. A summary of the research issues and potential applications of genetic studies is listed in Box 6-3. In the following section, we outline the recommendations for immediate, intermediate, and long-term actions. CONCLUSIONS A key stumbling block to applying toxicogenomic information to risk reduction in humans has been the difficulty in conducting large population studies to understand the distribution of gene-environment interactions in the population at large. Without adequate measures of exposure, studies of gene-environment interactions cannot be carried out effectively.
OCR for page 104
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment BOX 6-3 Summary of Research Issues and Potential Applications Questions to be answered How does human genetic variation influence transcriptomic, proteomic, or metabonomic patterns of response to toxic agents? Is genetic susceptibility to different classes of toxicologic agents due to a few key genes or does it represent a continuum of multigenic risk? How many drugs fail clinical trials because of a toxic response in a genetically susceptible subgroup? How do we best use existing environmental cohort studies to identify gene-environment interactions with toxicogenomic approaches? How can toxicogenomic research be translated and tested to reduce health risks for the public? Gaps in knowledge The influence of human genetic and epigenetic variation on transcriptomic, proteomic, and metabonomic studies of toxicologic agents is unknown. Although animal models and established human cell lines offer some insight into human response to toxic agents, questions remain about how well these studies indicate toxicologic risk in free-living humans. Cohort study estimates of the relative risk of disease for most genotype-environment combinations is lacking. Multigenic predictive models of toxicologic risk that integrate pleiotropic and polygenic networks of interactions among genes, environments, and other factors have yet to be developed. How can this technology be applied? Identifying genetic variations associated with environmental susceptibility to toxicity can be used to identify at-risk subgroups of the population through genetic testing. Developing multigenic models of toxicologic risk can be used to better understand the distribution of risk in a population and can be used in risk communication efforts to reduce exposure and disease. Animal models provide an important experimental method for identifying and characterizing genetic factors associated with increased susceptibility to toxicity from chemical exposure. There is substantial evidence that genetic variations in many genes influence individual response to toxic agents. Heterogeneity in the distribution of susceptibility SNPs and environmental exposures, as well as heterogeneity in the relationship to disease of these factors (for example, gene-gene interactions) and how they are affected by other factors (for example, age and sex), needs to be better understood in human populations to identify individuals and subgroups at
OCR for page 105
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment risk. If we are to adequately understand the continuum of genomic susceptibility to toxicologic agents that influences the public’s health, more studies of the joint effects of multiple polymorphisms need to be conducted. Much of the current research emphasizes identifying single gene mutations associated with differential response to environmental exposures. A more holistic approach to the analysis of data, an approach that encompasses gene-gene and gene-environment interactions, is likely to more efficiently advance our understanding of the population distribution of genetic components of risk. The influence of toxic substances on epigenetic modification of an individual’s genome is likely to depend on variation in the type, timing, and duration of exposure as well as the underlying genomic variation. RECOMMENDATIONS Immediate Actions Exposure Assessment Ensure that resources are adequately allocated to exposure monitoring and detection (external and internal to the individual), with approaches that are high speed and high dimensional (that is, can measure multiple chemical compounds simultaneously). Investigate the potential utility of metabonomic technologies to provide quantitative and qualitative assessment of an individual’s exposure. Animal Models Use animal models to identify genes associated with variability in toxicity and to validate causal mechanisms underlying human gene-environment interactions. Use animal models to model the genomic susceptibility (that is, polygenic) that is likely to underlie the continuum of genomic risk found in human populations. Begin developing an animal model resource that mimics the genetic heterogeneity of human populations—a resource that can be used to study the distribution of gene-gene interactions and gene-epigenetic interactions and can serve as a model for understanding population risk. Intermediate Population Studies Use genome-wide association studies, ranging from anonymous dense SNP scans to specialized arrays of putative functional SNP approaches, to iden-
OCR for page 106
Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment tify the full complement of genes and their variations that influence sensitivity to toxicologic agents. Use existing environmental cohort studies and clinical drug trials to investigate the impact of genetic variations on variation in response to a wide range of chemical exposures and pharmaceutical therapies. Context-Dependent Genetic Effects In addition to understanding the influence of single SNP variations on susceptibility, focus more attention on investigating context-dependent genetic effects (that is, gene-gene interactions as well as interactions with other biologic contexts such as developmental age, sex, and life course factors) that reflect the state of biologic networks underlying response to toxicologic agents. Models Develop multigenic and polygenic models of environmental sensitivity to better characterize the continuum of genomic susceptibility to toxicity and to better use genomic information for risk reduction. Long Term Epigenetics Conduct research on the influence of exposure variation, genetic variation, and their interaction in determining epigenetic modification of the human genome. Better characterize the influence of epigenetic modifications on disease processes that are associated with exposure to toxicologic agents to use this information for risk characterization and risk reduction.