8
A Multilevel Approach to Improving Risk Assessment for Developmental Toxicity

In this chapter, the committee addresses its third charge to evaluate how the new information and opportunities described in Chapters 4, 5, 6, and 7 can be used to improve qualitative and quantitative risk assessment for developmental effects. To make such improvements, the committee envisions exploiting the insights and opportunities from two kinds of research efforts:

  • First, research advances can be made in the understanding of mechanisms of developmental toxicity. As discussed in Chapter 4, this entails an understanding of the toxicokinetics of delivery of the chemical to a target site with an understanding of all the subsequent toxicodynamic steps. These steps include the chemical’s interaction with the target molecule(s), the consequence of altered target molecule activity for one or more developmental processes, and the subsequent emergence of a particular pathogenesis (developmental defect).

  • Second, research advances can be made that would increase our ability to reliably extrapolate from test results from model test animals to humans and to all members of the heterogeneous human population. As discussed in Chapter 3, risk assessors often resort to applying large default corrections to the animal data to estimate allowable human exposure concentrations of chemicals because of uncertainties about the relevance of animal data to humans. These uncertainties underlie the quantitative limitations of current risk-assessment approaches.

In the committee’s judgment, a research agenda should be prepared that addresses these gaps in knowledge about mechanisms of developmental toxicity and in the ability to extrapolate the results of animal tests to humans. The new information about development and genomics should be incorporated into the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



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 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment 8 A Multilevel Approach to Improving Risk Assessment for Developmental Toxicity In this chapter, the committee addresses its third charge to evaluate how the new information and opportunities described in Chapters 4, 5, 6, and 7 can be used to improve qualitative and quantitative risk assessment for developmental effects. To make such improvements, the committee envisions exploiting the insights and opportunities from two kinds of research efforts: First, research advances can be made in the understanding of mechanisms of developmental toxicity. As discussed in Chapter 4, this entails an understanding of the toxicokinetics of delivery of the chemical to a target site with an understanding of all the subsequent toxicodynamic steps. These steps include the chemical’s interaction with the target molecule(s), the consequence of altered target molecule activity for one or more developmental processes, and the subsequent emergence of a particular pathogenesis (developmental defect). Second, research advances can be made that would increase our ability to reliably extrapolate from test results from model test animals to humans and to all members of the heterogeneous human population. As discussed in Chapter 3, risk assessors often resort to applying large default corrections to the animal data to estimate allowable human exposure concentrations of chemicals because of uncertainties about the relevance of animal data to humans. These uncertainties underlie the quantitative limitations of current risk-assessment approaches. In the committee’s judgment, a research agenda should be prepared that addresses these gaps in knowledge about mechanisms of developmental toxicity and in the ability to extrapolate the results of animal tests to humans. The new information about development and genomics should be incorporated into the

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment agenda. The agenda would address basic scientific questions about which there is inadequate information. Specific areas of opportunity include: Intraspecies differences in sensitivity to toxicants. Genetic differences are suspected to be a major factor in intraspecies differences within human populations and within test-animal populations. Do individual differences concern mostly genetic variation in toxicokinetics, particularly in DMEs? Do they also include genetic variation in developmental components, such as those of the 17 intercellular signaling pathways used throughout development? Do multiple genetic differences have additive effects for the individual’s toxicant susceptibility? Is genetic variation in components of molecular-stress pathways also important in the individual’s response to toxicants? What is the contribution to individual susceptibility of nongenetic differences, such as those of age, history of disease, nutrition, and exposure to other chemicals and pharmaceuticals? Cross-species extrapolation. What are the toxicokinetic differences, particularly in the activities of drug-metabolizing enzymes (DMEs), between test animals and humans? Are differences in susceptibility to developmental toxicants due to differences in developmental molecular components and processes? Can some of the differences between test animals and humans be reduced or eliminated? Extrapolation of high-dose exposure of small populations of test animals to low-dose, long-term exposure of a large human population. Do chemicals have different toxicokinetic and toxicodynamic effects at high doses versus low doses? Does the organism rely on different protective responses to chemicals at different doses, such as enzymatic detoxification at low doses, molecular-stress reactions at intermediate doses, and the apoptotic response at high doses? Expanded test information for numerous chemicals and, especially, mixtures of chemicals. Can structure-activity relationships be obtained for a larger variety of chemicals by using in vitro tests with purified components (e.g., DMEs and developmental components of signaling pathways and transcriptional regulation)? Can model animals be genetically modified so that more mechanistic information can be obtained from them than from standard animal tests? As mechanisms are better understood for certain chemicals, can the effects of related chemicals be better predicted? THE MULTIDISCIPLINARY, MULTILEVEL, INTERACTIVE APPROACH The committee will outline in the remainder of this chapter a multidisciplinary, multilevel, interactive approach in which recent and future advances in developmental biology and genomics can be integrated with developmental toxicology to improve risk assessment for human developmental defects. This approach is not simply an alternative to current practices, but represents a novel

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment approach to assess risk for developmental defects. In Chapter 9, the interface of risk assessment and developmental toxicology is further explored within the fourth charge to the committee to develop recommendations for research in developmental toxicology and developmental biology to assist in risk assessment. The approach is multidisciplinary and multilevel because it invokes a wide variety of sources of information for risk assessment, including not only the assessment of toxicity and mechanism of action of chemicals in a variety of model systems (in vitro assays, nonmammalian models, and mammalian models), but also the assessment of toxicity, susceptibility, and exposure in human populations. The approach is appropriate to the risk assessment task, because assessing developmental toxicity is a broad and difficult area. The understanding of toxicity mechanisms entails both toxicokinetics and toxicodynamics, ranging from molecules to pathogenesis. Furthermore, the analysis of the differences between model systems (e.g., test organisms) and humans, as needed to improve extrapolation of test results, will require extensive comparative work at a variety of levels. These advances will depend on knowledge from chemistry, biochemistry, molecular biology, cell biology, developmental biology, genetics, ecogenetics, anatomy and organ physiology, genomics, and even systematics and evolutionary biology (e.g., finding conserved and nonconserved processes of development). A multidisciplinary, multilevel approach is needed to “bridge the gap” between the emerging scientific information and the assessment of human risk. At the same time, an interactive approach is needed for the dynamic interplay between the sources of new information and the needs of risk assessors. The committee’s multilevel approach should not be mistaken for a multi-tiered approach, where a specific order of evaluation and types of testing are specified. In a strict tiered approach, screening data at a low tier (low-cost, high-throughput tests, low assurance of relevance to humans) are first used to estimate risks of a large number of agents, and an agent with a high potential risk estimate at this tier triggers more rigorous and relevant tests, with higher associated costs, at successively higher tiers. For example, the Endocrine Disruptor Screening Program (EDSP) will use a tiered approach (EPA 1998g). Although the committee’s multilevel approach also involves tests ranging from inexpensive, high-throughput tests to slow, rigorous, and expensive tests, the approach differs in several respects. First, there is no uni-directional triggering of higher level testing by results obtained from a lower level test. Testing can be done independently at any level or at several levels, depending on the particular compound, the risk assessment questions, the anticipated human exposures, and the commercial uses considered for that compound. Results at one level could lead to tests at lower levels rather than higher levels. Second, the different testing levels yield different kinds of information, all with the potential to contribute to the knowledge of toxicity mechanisms, from molecular interactions to pathogenesis, and the understanding of the basis for extrapolation. All levels are designed to provide information useful for human developmental toxicity risk assessments. For

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment example, if a chemical (parent compound or metabolite) shows toxicity in a mammalian test, with hints of interference with a particular cell signaling pathway, the next step might be to analyze the mechanism of toxicity in a genetically sensitized invertebrate model system specifically designed to evaluate that cell signaling pathway. Likewise, observations from receptor-binding assays might prompt a re-examination of the overall impacts of a compound on organ development in a mammal. Thus, assessments are anticipated to have implications for analyses in both directions in the information levels. As new scientific observations are made, this approach allows the incorporation of new data into the risk-assessment framework. Ultimately, risk assessment has much to gain from the multidirectional flow of information across these information levels. In Figure 3-1, the committee introduced two-way arrows to indicate the importance of the responsiveness of the whole process to issues and ideas raised not only by science but also by risk-assessment needs. In light of the gaps in knowledge of toxicant effects, risk assessment is most likely to improve when research and risk assessment reinforce each other. Understanding mechanisms of toxicity can then be useful for predicting which other potential toxicants might act by the same mechanism, improving the ability to develop structure-activity relationships. Understanding the basis for extrapolations between test animals or in vitro assays and humans will give risk assessment greater validity. An iterative and interactive process for risk assessment was first defined in the National Research Council (NRC) report Science and Judgment (1994). However, such a process has yet to be fully implemented in risk assessment for developmental toxicology. Table 8-1 summarizes the committee’s multilevel approach. There are two components, each with four sources of information: (A) assessment in model systems of toxicity and mechanism of action of developmental toxicants (Table 8-1A), and (B) assessment in human populations of toxicity, susceptibility, and exposure to toxicants (Table 8-1B). The left column in both tables lists for each information level the experimental description of the tests, the application of the tests, the number of tests that can be done per year, and the value of the test information for risk assessment. Toxicity Assessment in Model Systems Information levels 1 and 2 of model systems in Table 8-1A generally involve relatively inexpensive and fast characterizations of chemicals and developmental effects. They should provide valuable information about which developmental pathways (signaling pathways and transcriptional regulatory circuits) are affected by which toxicants. Although extrapolations to human risk would be very limited without additional toxicokinetic and toxicodynamic information, the testing capacity should be available to characterize a large number of chemicals, and indeed, most of the several million chemicals in the environment, including chemi-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment TABLE 8-1A Information Levels for Assessment of Toxicity and Mode of Action of Chemicals in Model Systems Information Level Information Level 1 Information Level 2 Information Level 3 Information Level 4 Experimental description High-throughput biochemical and cellular assays. Identify potentially vulnerable pathways in development, e.g., the 17 signaling pathways, molecular stress pathways, and the main toxicokinetic pathways (DMEs). Identify molecular interactions of toxicants and target molecules, and activity changes Developmental assays with genetically optimized and sensitized nonmammalian animals. Identify potentially vulnerable developmental pathways, target organs, and times of susceptibility in development. In vivo mammalian developmental toxicity testing. Use genetically optimized animals for relevant sensitive tests. Identify potentially vulnerable pathways, target organs, and times of susceptibility. Link dose-response relationships to both developmental and functional (e.g., behavioral) changes. Detailed mechanistic studies to understand mode of action for selected developmental toxicants. Toxicokinetic and dynamic processes of chemicals would be identified and quantitatively assessed. Animal tests of toxicant sensitivities of polymorphisms. Analysis of differences of test animals and humans, regarding DMEs and signaling components. Application Use to assess most chemicals (>100,000) used by humans or released into the environment. Assess combinations of chemicals. Collect dose-response data. Use to assess many chemicals, doses, and combinations. Use to assess relative toxicity of related compounds. Use for hazard assessment of chemicals of likely high exposure or concern to humans. Can generate useful dose-response data relevant for mammalian evaluation. Gain information on toxicant sensitivity of polymorphisms. Use for a few chemicals where significant human exposure is likely or where knowledge about toxicant mechanism or about animal vs. human differences would be useful for human risk assessments. Assays per year 105-106 103-104 102 10 studies

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment Information Level Information Level 1 Information Level 2 Information Level 3 Information Level 4 Database output Linked databases of chemical information and effects on signaling pathways, molecular-stress pathways, and conversions by DMEs. SAR (structure-activity relationship) database for chemical effects on developmental and signaling pathways, end points, and susceptibility factors. Developmental toxicity database with both qualitative and quantitative information. Mode-of-action database for developmental toxicants. Risk assessment information available from each assay type Identification of potentially vulnerable signaling pathways. Identification of chemical properties associated with potential to alter signaling and stress pathways. Improvement of SAR data for developmentally relevant impacts. Identification of potentially sensitive target organs. Information on periods of sensitivity. Dose-response characterization. Mammalian relevancy information. Special toxicant sensitivity of genetic variants. Improved mechanistic understanding. Quantitative information for predicting human susceptibility. Better validation of animal-to-human extrapolation.

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment TABLE 8-1B Databases for Assessing Toxicity, Susceptibility, and Chemical Exposure in Human Populations Database Human Developmental Outcomes Database Human Genome and Genetic Polymorphism Database Biomarker Database for Exposure, Effect, and Susceptibility Human Gene-Environment Interactions Database Experimental description Human epidemiology and surveillance databases relevant for birth-defects research will be linked for morphological and functional impacts. Profiling of human populations for polymorphisms of developmentally relevant genes, such as those encoding DMEs, signaling components, and stress-response components. Link human biomarkers of exposure, susceptibility, and effect for development. Detailed investigation of genotype-environment interactions for toxicant effects on development. Application Identify, characterize, and link human databases: case reports, active and passive birth defects surveillance databases, post-market surveillance databases, link with chemical exposure and toxicity databases, link with known genetic birth defects syndrome databases, link with human genome project. Prioritize human genomic profiling for genes encoding components of pathways relevant to developmental toxicity, i.e., DMEs, molecular-stress pathways, cell-cell signaling pathways and developmentally relevant signal pathways. Improved understanding of critical signaling pathways should allow for improved linkage of chemically induced early cell biological effects with impacts on development (organogenesis and behavior). Knowledge about susceptibility biomarkers such as DMEs, signaling components, and developmentally critical systems will allow for linkage of susceptibility biomarkers with birth defects. Identification of critical interactions between genes and environment, e.g., additivity of defects of several signaling components in determining toxicant sensitivity. Risk assessment information Linkage of relevant databases, improved hazard identification, improved surveillance. Understanding of basic genetic variability across human populations. Identification of potential susceptibility genes relevant for development. Improved biomarkers and improved understanding of biomarker data in risk assessment. Characterization of susceptibility profiles for human developmental toxicants. Identification of variability in the susceptibility of human populations for birth defects from environmental factors. Improved quantitative information.

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment cals from natural products such as plants, have never been characterized for developmental toxicity. (It must be noted, however, that the EDSP plans to test 87,000 chemicals.) Of course, it will take years to develop, standardize, and validate these assays; however, the potential utility in screening with validated tests is high and justifies the effort. At information levels 3 and 4, the cost is higher and the test times are longer, but extrapolation to human risk potential can be more direct. Information levels 1 and 2 make use of model systems of far less complexity than humans. The results from these specialized cell assays and model organisms would be useful to organize chemicals according to their effects (e.g., to reveal chemicals that bind to the same protein—e.g., a nuclear-hormone receptor—or interfere with the same conserved cell signaling pathway). Assays in information levels 3 and 4 are likely to improve in their relevance to humans in the near future as differences between rodents and humans are better understood and as genetically modified model animals become available. These models will more closely resemble humans with respect to toxicant uptake, metabolism, and developmental response. The information level approach integrates risk assessment information from a variety of sources, both model systems and humans, and incorporates steadily improving methods into these sources of information. The recent advances in developmental biology that reveal the conservation of cell signaling pathways and genetic regulatory circuits across species, even phyla, gives a new demonstration to the toxicological principle that chemical impacts in humans can be predicted from animal systems. Further research will clarify the similarities and differences between model animals and humans and will improve the ability to extrapolate risk across species. How the test results will inform risk assessment will depend in part on the questions asked. The bottom row of Table 8-1A describes the information available to risk assessment from each assay type. Until scientists gain a better understanding of embryonic development and the mechanisms of toxicity, especially the effects of chemicals on the highly conserved signaling pathways, the approach in risk assessment should be to use combined information about predicted chemical activity, bioassays on model animals, and identification of individuals with susceptible genotypes to predict potential risk for developmental defects in humans. For example, knowing that a chemical disrupts the activity of a component of the Hedgehog signaling pathway in a high-throughput cell assay (a level 1 result) has limited value for direct human risk assessment. However, from level 1, it might be useful to know that four structurally related compounds all cause inhibition of a specific kinase involved in the phosphorylation of an intermediate of the Hedgehog pathway but with widely varying potency. If the most potent compound is the one proposed for widespread use and release into the environment, the level 1 information would prioritize testing of that compound for effects in mammals in vivo. Thus, information on molecular and biochemical ac-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment tions of chemicals at a cellular level, when considered in light of specific risk questions (human use and exposure paths), can be informative to human risk assessment. The four information levels for toxicity assessment in model systems (Table 8-1A) are presented below in detail. Model Systems Information Level 1 This level includes molecular, biochemical, and cell-based assays. Assays should be designed for a high throughput of chemicals, perhaps 105-106 assays per year, to provide basic information on the types of chemicals that disrupt signaling pathways and activate molecular-stress pathways and on the conversion of chemicals by DMEs. Such information will inform hazard identification and the evaluation of modes of action in risk assessment. For hazard identification, such assays will provide structure-activity information, relative potency information for chemicals evaluated in the same assays, information about the activity of chemical mixtures, and some quantitative information across assay end points for estimating relative potency across chemical classes. When coupled with estimates of actual or impending human exposure, such assay information would be useful in prioritizing chemicals for in vivo assessments at information levels 2 and 3. With the recent advances in molecular techniques, results from 105-106 tests of chemicals or chemical mixtures are feasible within 1 year. Assays already exist for estrogen receptors (EPA 1998g), and they could be readily modified to include other nuclear-hormone receptors, including the orphan members. Related assays could be devised for the transmembrane receptors, the various transduction intermediates, and the other 16 signaling pathways and their genetic regulatory proteins. Several pharmaceutical companies have active programs to evaluate chemicals in relationship to retinoic acid receptor binding and their pharmaceutical-versus-developmental-toxicity activities. Biochemical assays could involve purified human proteins expressed in bacteria or insect cells. Cell assays could involve mammalian cell lines or yeast into which human receptors (e.g., various receptor tyrosine kinases (RTK)) have been introduced. Molecular-stress pathways, cell-cycle checkpoint pathways, and the apoptosis pathway have already been shown to be especially relevant for environmental toxicants because of their roles in the cell’s response to chemically induced deoxyribonucleic acid (DNA) damage, impaired DNA synthesis, spindle damage, and kinetochore malfunction. The genes encoding these pathways could be introduced into yeast or cell lines equipped with reporter genes for easy assessment. At this high-throughput level, many chemicals should be assayed as sub-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment strates or inhibitors of human DMEs, both the oxidases and the converting enzymes, in biochemical assays or in cell lines carrying a variety of human DMEs. Such approaches have begun (for a review, see Crespi and Miller 1999). Thus far, this information has found limited use in risk assessment for several reasons. First, many assays have not utilized consistently stabilized transfected cell lines, and therefore the responses have been variable. (This variability has also been problematic for the EDSP.) Second, gene-induction profiles for specific compounds have shown tremendous variability when inducers are compared across established human cell lines. Thus, although comparisons can be made across compounds within some cell assays, comparisons across cell lines and with animals remain problematic. Ongoing research efforts in these areas should help clarify and resolve these issues. However, risk assessors need to understand what types of information these assays can provide for assessment. Recent conferences have summarized key information available from the use of human cell- and tissue-based assays (Society of Toxicology (SOT) Workshop on In Vitro Human Tissue Models in Risk Assessment, September 1999). Additional work on such systems would prove especially useful, as such recent conferences attest. As data are obtained, they would be entered in a widely accessible database (e.g., the recent Science magazine Web site for cell signaling pathways, www.stke.org). Compounds with high activity in such tests could be prioritized for higher levels of testing, especially if human exposure is current or pending. Compounds that do not show effects in these assays would still need testing at other levels of assessment if human exposure or environmental release is likely. At this information level, false positives are preferable to false negatives, and high sensitivity is preferable to low sensitivity (see discussions on how to use such information to strategize test applications by Lave and Omenn 1986). It is expected that comprehensive gene expression assays will soon become routinely available with DNA microarrays. Some arrays now include 6,000-10,000 DNA sequences in order to detect changes in messenger ribonucleic acid (mRNA) levels (after conversion to complementary (c) DNAs) in cells or tissues exposed to a chemical. Libraries of yeast strains, each carrying one of the 6,000 genes on a plasmid with a reporter gene, will soon be available. These libraries will represent a full-spectrum profile of the effects of chemicals on gene expression. The committee expects that the assays at this level will be better and broader in the near future and emphasizes the need to expeditiously put the results from these assays in the context of temporal patterns, and dose and downstream responses. Model Systems Information Level 2 At this level, nonmammalian animals should be used to assess the potential for chemicals to affect developmental processes. The animals are small, inexpen-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment sive, and fast developing. Information about their development is abundant, and genetic manipulations are easy. They can be genetically optimized to contain various sensitized signaling pathways and molecular-stress pathways and can often be coupled to reporter genes for enhanced observation of effects. Also, the animals could be genetically modified to reduce their differences from humans in various ways, such as their array of drug-metabolizing enzymes. (The committee acknowledges, however, that unrecognized differences between humans and these test animals may exist and may invalidate comparison. For example, humans and test animals may differ in unknown proteins of trans-epithelial transport of the toxicant or in unknown serum proteins that bind the toxicant. Therefore, validation studies would have to be done with a set of toxicants to establish cross-species concordance.) Assays would be designed for a medium throughput of chemicals, perhaps 103-104 assays per year. Some combinations of chemicals could be tested, and various doses could be examined in some cases to discern low-concentration and threshold effects for specific developmental pathways. The fruit fly and the nematode are currently the most favorable organisms for use. The zebrafish will probably be the most favorable vertebrate for use. Genetic modifications can include the following: Sensitization of animals (e.g., individual signaling pathways are made rate-limiting for some aspect of development, such as eye or wing formation in Drosophila, so that a slight increase or decrease in the pathway’s function due to a chemical would give an altered phenotype). The signaling pathways would be those used repeatedly in early development and conserved across many phyla, namely, the RTK, transforming growth factor (TGF) ß, Wnt, Notch, Hedgehog, and nuclear-hormone receptor pathways. Introduction of various reporter genes to enhance the readout of effects. Introduction of human signaling components into the animals to reduce the extrapolation. Introduction of human DME genes into the test organism so that, whenever possible, animals are presented with the same metabolized form of chemicals that humans would produce. General toxicity caused by a chemical can be distinguished from specific effects on development in the animals by evaluating general lethality, growth, and developmental effects versus specific effects on the particular locally sensitized pathway of development. An argument against the use of these model organisms is that the amount of information relevant to chemical effects on human organogenesis will be small, because the organs of model organisms, such as the fruit fly and nematode, differ substantially from those of humans. However, it is important to note that the choice of model organisms reflects the new insights about conserved processes of development, and emerging opportunities to directly assay for processes that are

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment ways for oxidative stress, protein unfolding, and checkpoints (including those involving p21 or p53) have been used for developmental toxicity assessment (Wilson et al. 1999; Wubah et al 1997). Test animals might express a reporter gene in response to a chemical without suffering overt developmental damage. This result could be taken as an indicator of the need to retest the chemical at other doses and times, or in combinations with certain other chemicals, for synergistic detrimental effects. Mouse strains are available with sensitized signaling pathways (see Tables 6-4 and 6-5 on targeted disruption of genes encoding signaling components) in which null alleles are used in heterozygous or homozygous states and in various combinations. Animals could also be made to possess a variety of reporter genes by which the responses through specific signaling pathways could be assessed easily. Similar programs are already under way with sensitized mice for carcinogen assays (Eastin et al. 1998). Until recently, the lack of pathways with clearly identified developmental relevance has limited similar programs for developmental toxicity assessment. Level 3 assessments provide the highest level of information routinely available to a regulatory agency for evaluation of risk—that is, information requiring the least default correction for estimating human risk, although some default corrections are still likely to be needed. The new information on development and genomics implies that different genetically modified mammals would be optimal for testing different chemicals. The choices could be guided by the results of levels 1 and 2 tests. The following kinds of information available from mammalian systems would include structure-activity information, with activity carried to the level of effects on mammalian organogenesis, relative in vivo potency information, some information about activity of chemical mixtures, mechanistic information from sensitized animals, and quantitative information on shape of dose-response relationships in in vivo organ systems. Model Systems Information Level 4 Perhaps only 10 chemicals per year can be studied at this level. Those studied should be those for which further research would give important information about (1) chemical effects on development, (2) basic mechanisms of developmental toxicity, and (3) significant clues for human risk assessment (e.g., analyzing differences between test animals and humans). Such chemicals might be prototype members of families of chemicals for which other derivatives would deserve testing at lower levels for relative toxicity, where mechanisms of action can be elucidated or where the effects are difficult to score, as in behavioral and other functional assessments. These chemicals might also represent compounds

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment for which widespread environmental exposure is anticipated or is occurring. Presumably, this information would feed back into improvements of the level 1, 2, and 3 assays and would provide valuable input for the chemical databases. The animals used in these studies would probably be mammals, especially genetically optimized rodents. The committee recognizes that the action of some toxicants might fall outside the realm of current understanding of development (i.e., of areas emphasized in this report). Studies of such mechanisms would necessarily fall into level 4 assays (i.e., basic research). Two approaches of note are (1) DNA microarray surveys of changes in gene expression of cells or test organisms treated with potential toxicants, which then require the researcher to interpret the changes and substantiate the interpretation; and (2) phage display methods to screen and isolate cellular proteins that bind particular toxicants, which then require the researcher to identify the role of the protein in the cell and its relevance to a toxicity mechanism. Assessment of Toxicity, Susceptibility, and Chemical Exposure in Human Populations The committee believes that the quality and the accessibility of human epidemiological information need re-examination, in light of its present and increasing relevance for developmental toxicity risk assessment. The committee considered ways to link data from human surveillance studies with data from in vitro studies and in vivo animal studies and discussed how new biomarkers of exposure and susceptibility in humans could be linked more effectively with new biomarkers of effect, in order to improve the assessment of human risk for developmental toxicity. The committee defined four informational databases as domains of information about humans. These databases were not referred to as “levels,” because they provide different kinds of information and cannot be ranked, as the model systems can, by remoteness or immediacy of human relevance. All the databases contain information of use to risk assessors. Database of Human Developmental Outcomes This database is the domain of information from epidemiology and surveillance. The quality of various case reports of birth defects and possible toxicant exposure varies widely. Many are incomplete and of unknown accuracy. However, most known human developmental toxicants were first identified by case reports, including thalidomide, diphenylhydantoin, diethylstilbestrol (DES), and valproic acid. At the outset, case report information is of unknown value to risk assessment and cannot be used without extensive follow-up, a situation similar to information from the level 1 assays of the model systems. When correlations are strong enough or the developmental effects are incontrovertible, the prospective

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment toxicant should be brought forward for further investigation, such as more rigorous epidemiological characterization of the correlation of exposure and birth defect and more incisive animal testing to ascertain dose-response relationships. In addition to case reports, there are both active and passive birth defects surveillance systems in place, both nationally and internationally (NBDPN 2000). They vary in quality and completeness. A number of established non-governmental databases are used to monitor drug exposures and developmental defects, and these are frequently used in pharmacoepidemiology studies (Strom 1994). A recent study entitled “Healthy from the Start: Why America Needs a Better System to Track and Understand Birth Defects and the Environment” and conducted by the Environmental Health Commission (Goldman et al. 1999) evaluated the quality of state tracking systems for birth defects. The study analyzed existing data from those systems and looked at the connection between environmental agents and birth defects. The authors concluded that the majority of states either do not have a tracking system or have one that is inadequate. They concluded that the data are inadequate to draw conclusions about the role of environmental exposures in causing birth defects and recommended that a national effort for tracking birth defects and a national approach for monitoring environmental exposures be established. Those state tracking systems currently in place are usually either active or passive systems. In an active surveillance system, such as that administered by the Centers for Disease Control and Prevention, trained personnel actively seek data from sources such as vital records and hospital reports. Passive case identification involves relying on patients and health care providers to voluntarily report exposures (usually drug exposures) and outcomes (Strom 1994); follow-up is minimal or nonexistent. Passive, voluntary reporting of systems have several limitations, including under-reporting of adverse events, incomplete information on cases, and the retrospective nature of most adverse event reports (i.e., the reports are made after an adverse pregnancy outcome has occurred), and the true denominator for exposed pregnancies is not known. There are a number of pregnancy registries currently being conducted (Weiss et al. 1999). For example, the North American Registry for Epilepsy and Pregnancy is a surveillance program to monitor pregnancy outcomes in women taking antiepileptic drugs (NAREP 1998). Some prospective post-surveillance follow-up studies have been directed by pharmaceutical companies evaluating post-marketing impacts of drugs. The committee suggests that the data from pharmaceutical company studies be made available and that the existing efforts to track human developmental outcomes be better characterized and recognized. Improvements could include making various information from epidemiology databases accessible via the Internet. There are databases for which known genetically based malformations are linked with databases for specific clinically described syndromes. One example of such a database is the On Line Mendelian Inheritance in Man (OMIM) data-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment base (see Appendix B for Internet address). As of January 2000, the results from OMIM for genetically based malformations and pediatric diseases (out of 11,080 entries) are 10,362 (93%) with autosomal transmission; 620 (5%) with X-linked transmission; 38 (0.3%) with Y-linked transmission; and 60 (0.5%) with mitochondrial transmission. Information from levels 1 and 2 of the model systems (e.g., the numerous targeted gene disruptions in the mouse) has already pointed to molecular and cellular processes that might be affected in numerous human clinical cases with similar organ dysmorphogenesis. Linking human databases on clinical syndromes with information from model system levels 1 and 2 would be especially valuable, and earlier examples in this report reveal the importance of establishing such linked databases. Database of the Human Genome and Genomic Polymorphisms The primary focus of this database is to provide information about the frequency and distribution of genetic polymorphisms in humans. Such characterization would allow developmental toxicologists and risk assessors to include human variation in their definition of a human response to a developmental toxicant. One approach to address this would be to study offspring with developmental defects for alterations in the genes of interest (e.g., DMEs and morphogenetic pathways). Such populations (infants with malformations and surgical candidates for correction of malformations) and tissues from collections around the world are available, and have been well defined for the anomalies of interest. Determining the frequency, distribution, and correlation for various human polymorphisms of consequence for susceptibility to toxicants (e.g., polymorphisms of the DMEs and of developmental targets, such as components of signaling pathways and genetic regulatory circuits) with observed developmental defects would be especially useful in defining human responses. The new information on human polymorphisms is seen as providing entrance to the realm of gene-environment causes of developmental defects. The technologies for approaching this are now available for the first time. Such approaches can be useful for determining the correlation between genetic variants in human populations with those observed in model systems for developmental defects. There are probably large, undefined differences and understanding these differences will be invaluable. Information from the Human Genome Project and the National Cancer Institute (NCI) will be relevant for understanding human polymorphisms of disease-susceptibility traits and cancer-susceptibility traits. Most of these traits will probably be complex and have several genes and modifiers. Human populations are expected to contain a large number of sequence polymorphisms (e.g., single nucleotide polymorphisms occur at least at 1 in 500 bases, recent estimates being as high as 1 in 25 bases); therefore, population samples cannot be selected only on the basis of a shared sequence difference. Indeed, there are too many, and some polymorphisms probably have no effect on phenotype (so-called synonymous

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment and neutral mutations). Instead, the population sample must be selected on the basis of a shared sequence difference known to result in the altered function of a gene product of suspected relevance to toxicant exposure and susceptibility. Information from model systems (e.g., mice) will be useful to establish altered function and suspected susceptibility. Toxicant susceptibility genes are not well identified, but the current favorites for attention are those encoding products involved in the toxicokinetic aspects of exogenous chemicals (uptake, distribution, metabolic conversion, and clearance), particularly the DMEs. There are at least one thousand of these gene products, although several dozen probably metabolize 90% of chemicals. Individuals and ethnic groups are already known to have substantial differences in these genes, and a few such polymorphisms are associated with altered developmental toxicity (e.g., a polymorphism in the epoxide hydrolase gene that might result in fetal hydantoin syndrome in susceptible individuals; see Chapter 5 for details). However, most DME polymorphisms have not yet been associated with increased (or decreased) susceptibility to a chemical. Polymorphisms of genes involved in toxicodynamics need to be investigated. As evident from the extensive discussions of conserved cell signaling pathways and genetic regulatory circuits, the committee suggests that polymorphisms in components of these pathways and circuits be tracked for the following reasons: The pathways and circuits are used widely in embryonic development. Polymorphisms of the genes encoding some components correlate with particular kinds of cancers (e.g., Patched (a component of the Hedgehog pathway) heterozygosity and basal-cell carcinoma; adenomatous polyposis coli (a component of the Wnt pathway) loss and colon carcinoma). A few correlates already exist, such as higher frequency of cleft palate in humans who smoke cigarettes and have TGF variants. Identification of Hox A1 polymorphisms in autistic populations is also progressing, as described in Chapter 4. Information from level 2 model-system studies and from basic developmental biology on model organisms will become ever more important for human evaluations because many aspects of embryonic development are conserved across phyla. It would be useful for epidemiologists interested in studying developmental defects to interact with their counterparts involved in NCI projects for profiling cancer-susceptibility polymorphisms (several of which concern signaling components). Some information from the epidemiological analysis of polymorphisms might be directly relevant to understanding birth defects that have mainly a genetic rather than a genotype-environment basis. Information from the database on the human genome and human polymorphisms would benefit risk assessment by providing information on human diversity in relation to sensitivity to potential toxicants. This information would be

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment used in level 3 of the model systems, because test animals can be made more similar to humans in their DMEs and signaling-pathway components, in order to improve the extrapolation from test animals to humans. The rat and mouse genome projects are expected to reach completion not long after the time of completion of the Human Genome Project, and useful information about gene identity, location, and mutants should flow both ways between researchers focusing on humans and those focusing on rodents. Database of Human Biomarkers This domain would provide the best information on human exposure to chemicals and on human susceptibility to developmental effects from chemical exposure. Biomarkers of exposure indicate the actual level of a chemical in the individual (e.g., lead concentrations in blood or dentine in children; organophosphate metabolites in urine). Biomarkers of effect allow researchers to examine dose-response relationships at environmentally relevant exposures in humans, and biomarkers of susceptibility allow researchers to identify sensitive subpopulations of humans. Used in combination, such biomarkers are essential for improving human risk assessment. There remains a need to improve biomarkers based on incisive new information about toxicokinetics and toxicodynamics. Thus, a linked database on the Internet for biomarkers of exposure, effect, and susceptibility in humans should be developed. Advances in DNA microarray technology discussed in information level 1 of the model systems will help immensely in the development of human biomarkers. Eventually, all the gene expression profiles of all organs and embryonic parts at different developmental stages will be catalogued; this will provide the control condition for detecting chemical-induced departures in gene expression. The potential to monitor thousands of gene expression changes simultaneously in regard to dose-response relationships and temporal patterns will provide a critical link of exposure biomarkers with early effect biomarkers. Applications in human birth defects research could be immense. However, the relevant developing databases must be linked so that the numerous changes are connected to functional effects in a developmental framework—that is, a framework organized around the temporal and spatial changes of the embryo and fetus. Given the sensitivity of reverse transcription polymerase chain reaction (RT-PCR) techniques, small maternal and embryonic or fetal samples could be simultaneously monitored for gene-expression changes. The gene expression changes could serve as biomarkers of effect, and responses in utero could be compared with maternal responses to evaluate differential sensitivity. If successful, the amount of information in such databases will challenge the organizational abilities of scientists. When temporal changes in developmental patterns of gene expression are combined with changes at different doses for thousands of genes, the data set grows to immense size. Obviously, the storage of data in a retrievable

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment form, as well as analyzing such data, pose serious issues in bioinformatics. Current efforts are ongoing in cancer research to store and retrieve vast amounts of data; however, there is no equivalent example for birth defects and developmental toxicology. Biomarkers of susceptibility would include polymorphism sequence data for which susceptibility has been correlated with protein function, as in the case of decreased DME activity. The use of data on DME genetic polymorphisms has lagged in carcinogen risk assessment, and in developmental toxicity risk assessment, their use is even more delayed. The committee believes that DME polymorphisms can serve as excellent biomarkers of susceptibility and encourages development of programs to ensure these applications. Biomarkers of toxicodynamic (developmental) susceptibility are less advanced but might include polymorphisms of genes encoding components of signaling pathways, genetic regulatory circuits, or molecular-stress pathways. Biomarkers of effect might include indicators of early activation of molecular-stress pathways or signaling pathway inhibition (e.g., due to exposure of a person to environmental chemicals and pharmaceuticals). Biomarkers would, in principle, provide risk assessors with better information than that available from the human developmental outcome and human genome database to link human exposure, developmental effect, and heritable susceptibility. To conclude, the benefits of biomarkers for risk assessment include (1) identification of susceptible populations (addressing intraspecies variability); (2) improved dose-response information where subtle changes of biomarkers of effect can be linked with biomarkers of exposure (addressing issues of extrapolation from high to low doses); and (3) improved linkage between biological effects in humans and mechanisms of toxic action as developed from human and animal studies. Database of Human Gene-Environment Interactions Similar to level 4 of the model systems, this database would be a domain of research inquiry in a few cases where detailed epidemiological information might yield widespread value, and where assessment of gene-environment interactions in birth defects is feasible given the available resources. Investigative epidemiological inquiries, such as those on endocrine disruptors, methylmercury, lead, and organophosphates, might provide adequately robust data for linkage with adverse birth outcomes. Several of these databases have provided information supporting interesting proposals about genetic polymorphisms (e.g., on lead and organophosphates) or exposure conditions (e.g., on lead and mercury). In contrast to pharmaceuticals, many environmental agents interact with a wide range of targets and with different targets at different doses. Yet, the identification of critical early biological effects that dominate or initiate subsequent disease states is essential for risk assessment. Thus, some of these environmental agents are candi-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment dates for an exhaustive research analysis. The multifactorial view of developmental defects accepts the possibility of complex and subtle combinations of circumstances leading to developmental defects, and except for a few special cases, some agents undoubtedly need to be illuminated by research, and databases will be needed to store and cull the large amount of information. The Importance of Linking Databases The committee’s multidisciplinary, multilevel, interactive approach to improving risk assessment assumes that the recent research advances in development and genomics have the potential, not yet realized, to improve cross-species extrapolations and cross-assay extrapolations and to ascertain the developmental targets of toxicants. Although more relevant information will become available for human risk assessment, a significant challenge facing risk assessors who want to use this information is the informatics problem. Most of the information relevant for human risk assessment will exist in the separate databases that the committee has described and will be organized according to discipline-focused applications. For example, the field of medical genetics has databases containing information on birth defect syndromes, but lacks databases containing epidemiological information on chemical impacts on development. It was the consensus opinion of this committee that efforts are now needed to link these diverse databases. In this section, the committee describes the need for integrated databases that link information from model systems and human populations. Relevant databases for such purposes would include the following: Chemical databases with metabolic pathways, structure-activity correlates, and bioassay results. Genome databases of humans, mouse, rat, zebrafish, Drosophila, C. elegans, and yeast. Developmental databases containing information about components of developmental processes and their functions and interactions in model organisms (e.g., in Drosophila, C. elegans, zebrafish, frog, and chick). Functional genomics databases on expressed genes, including their time and place of expression in the embryo, and on the function of the encoded proteins (specific function or categorization of function by motif). Databases on human polymorphisms and disease associations and on human and mouse mutants, including all the targeted disruption mutants of mice and their phenotypes. Databases recording DNA microarray results of the simultaneous changes of expression of thousands of genes (currently as many as 10,000 simultaneously) following the exposure of cells, tissues, or organisms to various conditions. For toxicologists, databases must be searchable by chemical names, chemi-

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment cal classes, structural characteristics, and reactivities. Approaches already in place for the structure-activity relationship databases could be used to begin identifying relevant information. Current initiatives are under way to reanalyze archived and frozen tissues from chemical toxicity bioassays for gene expression changes. Any changes are then evaluated specifically for the originally tested chemical and for structurally and functionally related compounds. Such initiatives should provide the types of linked databases the committee believes should be developed. This information should be linked with developmental toxicity bioassay databases as well as with general toxicity information. Several databases are available that contain developmental toxicity and general toxicity information. For example, general toxicity information as well as some developmental toxicity information on selected chemicals are contained in the Agency for Toxic Substances and Disease Registry (ATSDR) Toxicological Profiles, the International Agency for Research (IARC) Monographs on the Evaluation of Carcinogenic Risks to Humans, the Integrated Risk Information System (IRIS), and the Registry of Toxic Effects of Chemical Substances (RTECS). There are additional specific sources for evaluating chemicals for developmental toxicity. These sources include the California EPA Hazard Identification Documents, the Evaluative Process Documents on Lithium and Boric Acid (Moore et al. 1995, 1997), REPROTOX, REPROTEXT, the Teratogen Information Service, the Developmental and Reproductive Toxicological Database, and the National Toxicology Program Teratology Studies. The newly established National Institute of Environmental Health Sciences (NIEHS) Center for the Evaluation of Risks to Human Reproduction will also provide detailed evaluations on developmental effects of chemicals (no completed reports are yet available). Some of the above-mentioned databases and reports contain detailed toxicity evaluations of chemicals; others provide less-detailed summaries or contain only bibliographic information. Some databases also include information on human exposure. It is important to connect these types of data in a way that is ultimately useful for human risk assessments. Genome databases are available on humans, mice, rats, zebrafish, Drosophilia, C. elegans, and yeast with sequence information on open reading frames (ORFs), introns and exons, cis-regulatory sequences, and relatedness to other genes. Of particular use for developmental research would be the ability to search and identify all genes with known developmental relevance, and to link that search with gene expression and temporal and spatial developmental information by organism, as well as organ and tissue development across species. Efforts by the National Institutes of Health through the NCI Cancer Genome Anatomy Project and NIEHS through the Environmental Genome Project are directed toward identifying genes of interest for cancer and toxicology. A similar effort for genes of developmental and toxicological relevance could be initiated and linked. In both cases, profiles of gene expression changes can be determined for tissues with specific developmental defects and for affected tissues after toxicant exposure.

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment These gene expression profiles can be linked with specific tissue changes using microdissection techniques. Again, emphasis on developmental characteristics is important. Such characteristics include temporal aspects of gene expression and multiple organisms, organs, and tissues of interest. Functional genomics databases on expressed genes are particularly relevant for developmental toxicologists. Expressed sequence tags (ESTs), ORFs, and the time and place of expression of each gene in the embryo and on the function of encoded proteins (specific function or categorization by motifs) would be essential information. For example, signal transduction pathways and the interactions of pathways are in the process of being summarized on the Web site www.stke.org. Journal articles, as well as methods protocols, will be posted (announced in Science, April 1999). Likewise, databases on human polymorphisms and disease associations, human and mouse mutants, including all the targeted disruption mutants and phenotypes in mice, would be important. Online access to databases listing all known human genetic syndromes relevant for development would also be essential. Within a decade, most genes encoding components of signaling pathways and genetic regulatory circuits important in development will probably be identified in humans and mice (the extensive synteny among vertebrates will be valuable here), and their times and places of expression will be known. Many human polymorphisms will be identified and correlated with heritable diseases. At levels 1 and 2 of the model system toxicity assays, this database information will be useful for choosing proteins to use in simple biochemical or cell assays or to modify in test animals, such as Drosophila. It is safe to say that almost any human gene and, hence, its encoded protein, can be put into such an animal assay. The question will be which proteins are most relevant to the identification of developmental toxicants in humans. If 100,000 assays can be done per year at level 1 and 10,000 at level 2, the results would have to be preserved in an immense database. Over the same period, all genes for proteins of the major toxicokinetic pathways of chemical uptake, distribution, metabolic conversion, and elimination in humans, mice, and rats will probably be identified, as will polymorphisms of these genes. This information will be preserved in databases, and should be widely available. Catalogs of phenotypes of mouse null mutants for individual genes in the heterozygous and homozygous states, and in combination with other gene disruption, will be useful for comparison with phenotypes of human birth defects in order to gain inferences about what is affected in human development. The level 2 and 3 model system assays for developmental defects will draw from these genomic databases and contribute to them. Several of the databases of epidemiological characterizations will benefit as well. Chemical databases might be more difficult to organize than genome databases. Whereas there are about 140,000 genes in humans, the universe of possible chemicals is unlimited (the number of possible human allele combinations is also unlimited). New compounds can always be synthesized, and their effects on developmental mechanisms are rarely predictable, at least so far. Also, a new

OCR for page 196
Scientific Frontiers in Developmental Toxicology and Risk Assessment generation of specific and complex chemicals will be introduced in the near future, because of new strategies of rational drug design based on knowledge of protein three-dimensional structure and chemical mechanisms, and because of high-throughput function-based screens of huge combinatorial chemical libraries. Still, mechanistic evaluations of the biological effects of chemicals should serve as an organizing principle for grouping compounds in a database and predicting their risk. In the committee’s proposed hierarchy of four information levels for model animal systems, each higher level provides increasingly complex information concerning higher biological complexity and responses to environmental chemicals. Systems of lower complexity might be used to organize chemicals, for example, to reveal those chemicals that bind to the same protein (e.g., a nuclear-hormone receptor) or act in the same signaling pathway. This information would be useful for predicting effects at the next level, for example, on a particular kind of organogenesis in which a particular signaling pathway is used. SUMMARY The committee has developed a multilevel, multidisciplinary, interactive approach for improving risk assessment for developmental toxicity. Model animal systems and human epidemiological studies are shown to be valuable sources of information for risk assessment, and it is emphasized that the multilevel approach is not a tiered approach. To meet the goals of this approach, the committee has described what information is available from model systems of differing complexity, from in vitro assays to whole animals, for the assessment of toxicity and mechanism of action of chemicals. The committee has also described databases and database needs for assessing chemical exposure, susceptibility, and developmental effects in human populations and the continuing value of human epidemiological data for risk assessment. For each database, the committee has identified the type of information provided and how that information answers risk assessment questions. Examples are given of the anticipated interactive aspect of the approach, whereby new data and methods can be incorporated into the risk assessment process. Finally, the committee has described integrated approaches essential for the linkage of numerous relevant databases across chemicals, times of development, descriptions of toxic effects, and applications.