Conclusions and Recommendations
The general conclusions regarding the four charges to the committee are given in the first part of this chapter. In response to the fourth charge, specific recommendations are given in the second part. Several of these recommendations concern a multilevel, multidisciplinary approach for obtaining data relevant to risk assessment for developmental toxicity, as described in Chapter 8.
CONCLUSIONS IN RELATION TO THE CHARGE
Charge 1: Evaluate the evidence supporting hypothesized mechanisms of developmental toxicity.
Issues in developmental toxicology have been clarified incisively in the past 30 years, and many experimental advances have been made. Still, there are only a few compounds for which developmental toxicity is partially explained and no compound for which it is fully explained in an inclusive hypothesis supported by strong evidence. Several reasons for this limited understanding should be cited:
Development is complicated. Only recently have developmental processes and molecular components been elucidated. Many steps are likely to be involved between the toxicant’s initial interaction and the ultimate developmental defect.
The etiology of developmental toxicity is complex. Many developmental defects of individuals might be the outcome of a multifactorial impact with overlap of genetic variation, exposure variation, and variation in other factors, such as nutrition and disease. Genotype-environment interactions, especially those involving several genes and environmental factors, are difficult to study, and have usually been avoided in basic laboratory research. The complexity of develop-
ment is manifest by both temporal- and tissue-specific sensitivities; thus, assessing a toxicant’s potential effects for development requires a dynamic and multilevel assessment strategy.
Environmental toxicants represent a broad spectrum of agents, probably working by a variety of mechanisms. Some toxicants probably have one or a few targets in the conceptus (the embryo or fetus, plus the embryo-derived extraembryonic tissues). Others probably have numerous targets (“broad specificity”) in the mother and conceptus, and others probably affect the mother, whose altered health secondarily affects the conceptus.
Without a thorough understanding of basic mechanisms of development and knowledge about variability in responses across species to toxicants, insights from animal studies have largely been only of assumed validity for human mechanisms.
The analysis of mechanisms of toxicity requires advanced interdisciplinary information and approaches of the kind that have only recently become available.
In considering hypothesized mechanisms, the committee discussed the different scopes and levels of understanding implied by the term “mechanism,” as used by different researchers in biochemistry, molecular biology, genetics, developmental biology, toxicology, and epidemiology. If the emphasis on toxicant action is exclusively molecular, some members felt that the mechanism misses the scope of potential linked impacts of a toxicant on overall development and morphogenesis. Additionally, most felt that a mechanism lacking molecular detail is inadequate for explaining the action of toxicants. Realizing this complexity, the U.S. Environmental Protection Agency (EPA 1996b) and International Programme on Chemical Safety (IPCS Workshop on Developing a Conceptual Framework for Cancer Risk Assessment, 16-18 February 1999, Lyon, France) have defined chemical “modes of action” in addition to “mechanisms of action.” In these definitions, “mechanism of toxicant action” is taken to refer to a detailed understanding of the overall toxic response. In contrast, “mode of action” usually refers to a more limited description of the overall process of toxicity that focuses on defining possible cascades of biological events that can occur following exposure to a toxic agent. To preserve the full range of causes and effects relevant to risk assessment of human developmental toxicity, the committee sought to designate “levels of information” obtainable from various model systems (including in vitro assays and mammalian and nonmammalian assays) to illuminate mechanisms of action (see Chapter 8). Hypotheses about toxicant action in humans, based on the information from animal models, can then be strengthened or dismissed by using information obtained from various types of human data. Chapter 8 provides suggestions on how these different types of data can specifically improve our ability to predict potential developmental toxicity in humans.
The committee believes that it is impossible to provide the most scientifically defensible risk assessments without understanding mechanisms of action.
The committee generally agreed that a complete description of the mechanism of action of a developmental toxicant should include the following types of mechanistic information:
The toxicant’s kinetics and means of absorption, distribution, metabolism, and excretion throughout the mother and conceptus.
The toxicant’s interactions (or those of a metabolite(s) derived from it) with specific molecular components of cellular or developmental processes in the conceptus or with maternal or extraembryonic components of processes supporting development.
The consequences of those interactions for the function of components in a cellular or developmental process.
The consequences of an altered process for the developmental outcome, namely, the generation of a defect, functional changes, or altered growth and development.
The committee acknowledges that a complete explanation of mechanism of action is not currently available for any chemical and that having even partial mechanistic information of the kind described above can improve the ability to predict adverse human developmental outcomes.
Toxicokinetics describes the steps of toxicant entry and absorption, distribution, metabolism, and excretion throughout an organism or, in this case, throughout mother and conceptus. Toxicodynamics, in the context of this report, describes the steps of the toxicant’s effects and interaction with the developmental processes. Both are important. Toxicokinetics explains whether, when, and how much of a potential toxicant reaches the embryo or fetus. The understanding of the toxicokinetic steps of detoxification or metabolic potentiation of a chemical holds great promise for safe drug design and preclusion of toxicant effects in humans from environmental agents. Furthermore, human individual differences in susceptibility to toxicants might in large measure result from differences in toxicant uptake and metabolism, and some of the problems of extrapolating toxicity test results from animals to humans can certainly be attributed to differences of laboratory animals and humans in the metabolism of chemicals.
In toxicokinetics, researchers have identified routes and rates of exposure of the conceptus to certain toxicants, and the recent information on drug-metabolizing enzymes (DMEs; the numerous P450 heme oxidases and conjugating enzymes) is very significant. Researchers have been able to verify the presence of parent compounds and metabolites in the mother and the conceptus during development. They have successfully explained some species differences in toxicity responses based on metabolism differences and have explained some human variations in drug responsiveness. Nevertheless, such knowledge about critical metabolites and their reactivity with specific target tissues is lacking for most agents, and much remains to be done in this promising area.
In toxicodynamics, the mechanistic picture of toxicity is less complete. This limitation has been inherent to the field, because so little has been known until recently about the identity and activity of specific molecular components of the developmental processes or about the roles of the processes in the development of the embryo. Hence, little could be said about the developmental consequences of a toxicant’s reduction or exaggeration of a component’s activity. In the absence of such information, hypothesized mechanisms of toxicant action (and evidence for these mechanisms) have had limited ability to ascribe developmental defects to failures of specific components and processes. Furthermore, species differences in developmental components have been poorly discerned, as has human variation in these components.
In a few cases, toxicodynamic hypotheses of mechanism emphasize molecular components and activities. Multiple retinoic acid (RA) receptors have been identified for retinoids in animal models. The molecular function of these receptors in regulating gene expression has been ascribed, and the altered time and place of gene expression has been detected in the presence of the toxicant. The availability of multiple structurally related analogs of RA has helped these investigations. Diethylstilbestrol is also known to bind to other nuclear hormone receptors and alter gene expression, and cyclopamine (a plant alkaloid) is known to bind to signaling components of the Hedgehog signaling pathway and alter inductive responses. Such information, coming from recent molecular studies, has greatly furthered the understanding of the toxicity of those agents. Still lacking, however, is full understanding of the developmental processes affected by this altered gene expression or altered signaling and, hence, the generation of the developmental defect.
Of relevance to the committee’s later proposals, these three examples of advanced toxicodynamic hypotheses concern signaling proteins and transcriptional regulators, the kinds of molecular components the committee recommends for greater attention in future analysis and testing of toxicants. In the near future, developmental toxicology will likely provide more comprehensive explanations of toxicity, but at present, mechanistic information is available for only a small number of toxicants and these have had limited application for risk assessment.
Charge 2: Evaluate the state of the science on testing for mechanisms of developmental effects.
The state of the science has improved greatly in the past decade, indeed even since this committee was first formed. Relevant advances have occurred in developmental biology and genomics (gene sequencing and gene identification), built upon advances in genetics, cell biology, molecular biology, and biochemistry. Developmental processes have been illuminated for the first time in a number of animals at the level of identification of molecular components and their activities, especially of the signaling pathways and genetic regulatory circuits of these processes. The same molecular components are used repeatedly at different
times and places in an animal’s development, and these same components are used across widely different animal phyla. Species differences of development seem to be largely differences in the combinations and sequences of use of conserved molecular components and differences in the final target genes of the conserved genetic regulatory circuits. The committee believes that the newly accessed level of molecular components and processes of development is the level that will provide incisive understanding of mechanisms of toxicity and improved predictability of toxicant effects. Developmental processes are increasingly understood in terms of the activities and ordered interactions of molecular components. Because of the recent advances in developmental biology and genomics, the committee is optimistic about the ability to improve testing procedures and the interpretation of data in the future.
Recent and ongoing studies of mammalian development have benefitted greatly from the study of nonmammalian model organisms that are genetically tractable and suited for rapid systematic analysis, such as Drosophila and Caenorhabditis elegans. Although not anticipated at the outset, the information from these model organisms proved to be extensively transferable to mammals, because molecular components and many developmental processes are deeply conserved across animal phyla. In particular, development in diverse animals, including mammals, depends on cell-cell signaling at all stages before cytodifferentiation, and this signaling involves repeated use of the same 10 intercellular signaling pathways. (Seven additional conserved pathways are used in the functions of cytodifferentiated cells.) Genetic regulatory circuits, involving certain transcription factors, are also conserved from Drosophila to mammals, as are components of most basic cellular processes, such as the cell cycle, secretion, and motility. The success in using nonmammalian organisms to illuminate mammalian development suggests that the same organisms could be useful in illuminating toxicodynamic mechanisms in mammals and useful in testing certain kinds of toxicants. Based on the past decade of progress in cellular and developmental biology research, it seems likely that this information will be relevant to mammals, including humans.
Charge 3: Evaluate how that information can be used to improve qualitative and quantitative risk assessment for developmental effects.
Throughout its deliberations, the committee kept in mind that the decisions of risk assessors about a chemical’s potential for developmental toxicity are ideally derived from mechanism-based, quantitative data on test animals for which the validity of extrapolation to humans is known. Unfortunately, such data are rarely available. As currently designed, rodent tests for developmental toxicity are limited in their capacity to provide both qualitative and quantitative mechanistic information for human health risk assessment. They are costly in time and resources, and therefore only a small fraction of the more than 80,000 chemicals in commercial use (and the much larger number, about 6 million, of natural prod-
ucts) have been tested. Chemicals are usually assessed for effects on growth, morphology, and viability of the newborn rodent but not for functional (e.g., behavioral), molecular (e.g., toxicant metabolism or transcriptional changes), or cellular (e.g., mitosis defects and apoptosis) effects. Because the validity of extrapolation from animal results to humans is often assumed, as is the relevance of some of the animal exposure conditions, risk assessment often includes large default corrections in the extrapolation to “safe” human exposure concentrations.
To improve qualitative risk assessment, a better understanding of the mechanisms of toxicity of a reference set of compounds would provide predictions for related, but less well-tested, compounds. Toxicity information on a wider range of compounds would also help qualitatively. Results from current developmental toxicity assessments reveal both qualitative and quantitative differences in animal responses to toxicants. A complete understanding of these species differences is not known, particularly, the proportion of these differences that is due to toxicokinetic versus toxicodynamic differences. Thus, to improve quantitative risk assessment, a better understanding of the differences and similarities between test animals and humans with regard to toxicokinetics and toxicodynamics would allow better extrapolations to be made. Use of default corrections could be limited or be made on the basis of knowledge rather than assumptions. An understanding of genetic differences (polymorphisms) among test animals, and a deliberate effort to control the differences, would reduce and rationalize the wide range of responses of test animals to a toxicant. An understanding of human polymorphisms of molecular components of toxicokinetic and toxicodynamic importance would allow better estimation of safe exposures of individuals over the whole range of human susceptibility to a toxicant. Finally, an understanding of toxicity mechanisms, including toxicant effects on cell division and cell function, as reflected in molecular-stress and cell-cycle checkpoint pathways, and an understanding of the polymorphisms of toxicokinetic and toxicodynamic components, would allow better estimation of low-dose exposure effects, as extrapolated from high-dose results. The committee believes that the possibilities for improvements in data for risk assessment are great.
To date, little of the new information has been brought to bear on the tests and interpretations needed for qualitative and quantitative risk assessment, simply because the problems are difficult and it takes time and resources to do so. The committee believes the new information can be incorporated extensively in the next decade.
Because molecular components and processes of development are best understood in genetically modifiable model organisms, such as Drosophila, C. elegans, the zebrafish, and the mouse, and because the conservation of components is so pervasive that it extends to humans, the committee recommends that these organisms be used more effectively for analyzing mechanisms of developmental toxicity at the molecular level and for assaying the developmental toxicity of the numerous never-tested chemicals and chemical combinations. These or-
ganisms can be used inexpensively and rapidly. Through straightforward genetic manipulation, they can be sensitized to toxicant effects and provided with reporter genes so that the impacts of chemicals on development can be easily scored. The organisms can be analyzed to define their toxicokinetic and toxicodynamic differences from humans, hence improving extrapolation, and they can be genetically modified to reduce some of the differences. Finally, the suspected genotype-environment source of many human developmental defects can initially be explored in model organisms for which the genotype can be controlled.
At the same time, human individual differences in drug-metabolism components and developmental components are being identified and quantified. As polymorphisms are related to individual susceptibility to toxicants, the difficult domain of human genotype-environment interactions will be entered. The knowledge of human variation will improve quantitative risk assessment.
Charge 4: Develop recommendations for future research in developmental biology and developmental toxicology; focus on those areas most likely to assist in risk assessment for developmental effects.
The committee concludes that recent advances in the fields of developmental biology and genomics provide unprecedented opportunities to understand the molecular mechanisms of action of toxicants, the differences in the developmental responses of test animals and humans to toxicants, the extrapolation of high-dose effects to low-dose effects, and the differences in the individual human susceptibility effects of toxicants on development. These advances can lead to improved animal tests for toxicants, improved extrapolation of animal results to humans, and through the means of better data, improved risk assessment.
A multilevel, multidisciplinary approach to toxicity assessment involves the use of genetically tractable model organisms for which development is well understood and the genome is completely sequenced (or soon will be), namely, C. elegans, Drosophila, zebrafish, and mouse. Because some of these animals are inexpensive to use, testing could be expanded to cover a larger number of chemicals, chemical combinations, and testing conditions, including various genotypic backgrounds of the test organism and scoring of toxic effects at developmental stages before organogenesis. The validity of extrapolation of test results can be tested as well. Risk assessors would benefit, the committee believes, from the use of a variety of toxicity data, combined with human exposure and susceptibility information. The recommendations in the next section are summarized in a text box.
Recommendation 1. To improve the understanding of the mechanisms of action of toxicants, the committee recommends that critical molecular targets of toxicants be identified among the components of developmental processes.
How can progress be made in the analysis of mechanisms of toxicity? What variety of mechanisms is there? Are there certain molecular components and processes of development that seem to be frequent targets of toxicants? The recent advances in developmental biology, cell biology, and genomics provide information about components, functions, and processes for the first time and for a wide range of animals, including mammals. Fortunately, many of the components and processes are conserved across phyla, making it possible to study them initially in experimentally convenient organisms, such as C. elegans and Drosophila, or even yeast and bacteria in some cases, and then carry the information across species to the studies of mammalian development, with great success. Many aspects of mammalian development resemble arthropod and nematode development.
The committee’s recommendation for future studies on mechanisms of developmental toxicity is that greater use be made of model organisms, taking advantage of the shared features of development with mammals. Fruit flies, nematodes, and mice have different organs and body organization, and yet they use many of the same molecular components and basic molecular interactions, although in different combinations and sequences. The analysis of toxicity mechanisms would have to bear the similarities and dissimilarities in mind.
Mechanisms of toxicity can now be understood more fully at the level of affected molecular components and the consequences of the altered activity of this component for developmental processes. Previously, toxicity analysis often ended with defining the time window of sensitivity of the conceptus to the toxicant and the profile of affected organs. This information is still important to have. Because toxicants affect the activity of molecular components, the recommended emphasis would be to identify components with which toxicants interact, determine the toxicant’s effect on the component’s function (the functions of many components are now understood), and fit the altered function into the effects on a developmental process, the abnormal operation of which leads to a structural or functional defect. The committee recommends that model organisms be used extensively to access this level of analysis so that researchers can benefit from genetic tractability, experimental convenience, and a background of information about molecular components and developmental processes.
1.1. In searching for mechanisms of developmental toxicity, the committee recommends research on how toxicants perturb evolutionarily conserved molecular targets and pathways of development.
To learn about mechanisms of developmental toxicity, special attention should be given to research in the areas of (1) signaling pathways and their associated genetic regulatory circuits (a group of key developmental targets); (2) molecular-stress and checkpoint pathways (a group of key cellular targets); and (3) the DMEs (a group of key toxicokinetic components).
Recommendation 1. To improve the understanding of the mechanisms of action of toxicants, the committee recommends that critical molecular targets of toxicants be identified among the components of developmental processes.
1.1.1. Conserved signaling pathways and genetic regulatory circuits as potential targets of toxicants.
Approximately 10 kinds of signaling pathways are repeatedly used at all times and places of development before cytodifferentiation (i.e., during cleavage, gastrulation, neurulation, and organogenesis). Development is built on coupled signaling and genetic regulation. As differentiated cell function begins during cytodifferentiation, seven other pathways are added. Thus, in the mammalian fetus, all 17 pathways are in use. Almost all activities of cells of multicellular organisms, including all embryonic stages, are contingent on extracellular signals. These activities include secretion, motility, adhesion, proliferation, differ-
Recommendation 2. The committee recommends that it is important to study how the new information about development and developmental toxicity can address the uncertainties in quantitative and qualitative risk assessment.
Recommendation 3. To improve the interdisciplinary advances in developmental toxicology, the committee recommends that the databases of developmental toxicology, developmental biology, and genomics be better linked on the Internet, and that multidisciplinary outreach programs be established for the effective exchange of information and techniques related to the analysis of developmental defects and to the assessment of toxicity for risk assessment.
entiation, and transcription. The components of signaling pathways and genetic regulatory circuits are key components to investigate. Further arguments for emphasizing these components as potential targets are the following:
When signaling is interrupted in mutant animals, development is affected, with specific effects depending on the component that is inactivated.
A few toxicants currently are known to affect signaling in mammals, such as cyclopamine (the Hedgehog signaling pathway; Chapter 4), retinoic acid and its many derivatives (the nuclear hormone signaling pathway and transcriptional
regulator; Chapter 4), smoking and transforming growth factor (TGF) variants in humans (the receptor tyrosine kinase pathway; Chapter 5), and smoking and MSX/BMP variants in humans (the TGF pathway; Chapter 5).
The committee acknowledges that signaling and gene regulation have not been proved to be particularly vulnerable points of development. An argument can be made that, because these pathways are so essential for all aspects of development, evolutionary selection has made them particularly resistant to perturbation, including that by toxicants. However, data from experimentally constructed mutant animals contradict this argument. In invertebrates, such as Drosophila and C. elegans, null mutants are often lethal, and hypomorphic mutants (i.e., ones having partial activity) show specific developmental defects, probably reflecting the times and places where the inadequate component is required at highest activity. In vertebrates, including mammals, the picture is somewhat different. The genes encoding most components have undergone duplication and slight diversification, so there are functionally equivalent members for each kind of component. This is termed “redundancy”. The members are expressed at different times and places in development. In vertebrates, null mutants frequently show, not lethality, but specific local losses, reflecting the few times and places where a lost component’s function is not overlapped by a related component. Some of these null phenotypes in mice closely resemble human developmental defects (e.g., see Tables 6-4 and 6-5). Thus, the committee’s hypothesis, namely, that signaling pathways and the associated genetic regulatory circuits are critical molecular points of susceptibility of development to toxicants, has not been proved.
A final point should be made about the appropriate use of model organisms for studies of molecular mechanisms of toxicity. As mentioned above, although these organisms have different organs from mammals, the developmental processes involved in their organogenesis are similar to those of mammals. Organogenesis in various species represents different combinations and orderings of conserved processes, such as signaling pathways and the responses they engender (e.g., proliferation, locomotion, and secretion). Thus, for example, in the course of a study of toxicity mechanisms, Drosophila might be scored for the effects of a chemical on its wing development, but what is really being scored is the effect of the chemical on the kinds of signaling pathways and genetic regulatory circuits also used in human development of different organs. Wing development serves as a well understood set of conserved molecular components and interactions. Effects on the wing can be readily recognized, and because of the advanced understanding of its development, the targeted developmental processes can be surmised. From the fly results, predictions can be made of the effects of those chemicals on mammalian organogenesis, for organs in which these same components operate. Zebrafish, as a vertebrate test animal, are expected to share many developmental processes with mammals (e.g., more details of specific organogenesis) and can be used as an intermediate model. All these model systems share with
mammalian embryos the need to coordinate and integrate the many temporally and spatially distinct cell regulatory networks operating during development. A fully developed, functional fly wing, fish fin, or human arm are all products of comparable hierarchically organized processes.
1.1.2. Molecular-stress and checkpoint pathways as potential targets of toxicants.
There is evidence that some toxicants primarily damage basic cell functions, such as those of cellular reproduction, and that the damaged cells subsequently fail to participate in development. Such toxicants can act indirectly on development, although the failing of basic cell functions can soon enough impede development. Molecular-stress and checkpoint pathways are the cell’s defenses for counteracting such damage. At least 10 of these pathways are known. They respond to different kinds of damage and decisions to make different counteractions. Apoptosis is the ultimate pathway, by which the cell is self-destroyed. This is thought to happen when the other counteractions are not sufficient to restore the cell to a minimal state of basic function.
Especially with regard to environmental toxicants, these pathways might deserve special attention, because some toxicants react widely with cellular components. As mentioned in Chapter 6, many environmental toxicants set off an enlarged domain of apoptosis in the embryo. The particular spectrum of activated pathways would give an indication of the impacted cellular processes and components. Also, low- and high-dose effects could be discriminated. At low doses, cells might recover or a fraction might die and be replaced by the proliferation of others, in which case, development might continue normally. At high doses, recovery might be exceeded by impaired proliferation or apoptosis and defects occur.
The investigation of toxicant targets in these pathways can exploit C. elegans, Drosophila, zebrafish, and mouse. The molecular-stress and checkpoint pathways have been well analyzed in these organisms (in fact, discovered in them in some cases). The pathways are nearly identical in all those organisms, although mammals have a greater variety of closely related components. In general, there has been less study of these pathways in embryos than in adults, so there might be some unexpected differences.
1.1.3. Toxicokinetic components, especially drug-metabolizing enzymes.
As noted above, there are many steps in absorption, distribution, metabolism, and excretion that determine whether the conceptus will be exposed to a toxicant. The molecular components of absorption, distribution, and excretion are not well known, but those of metabolism, the DMEs, have been extensively elucidated in recent years. There might be a few hundred kinds of DMEs, with different substrate specificities and time and place of expression in the mother and conceptus. Studies are well under way to analyze the roles of DMEs to
metabolic potentiation or detoxification of chemicals, as part of the research program of ecogenetics, pharmacogenetics, and the Environmental Genome Project. Much study of the roles of DMEs is done in the mouse, for example, with null mutants of individual enzymes. The committee fully favors this direction of study and the support for it. As described later, it will also be valuable for animal assays of toxicants to know the DME similarities and differences from humans, as a part of the validation of the extrapolation of data from animals to humans. Some studies with other model organisms are probably of use. Some DMEs are widely conserved among animals (e.g., CYP1A1 in fish) and would be easier to study in other model organisms.
1.2. In pursuing mechanisms of toxicity, the committee recommends research to explore how molecular perturbations lead to dysmorphogenesis and other adverse outcomes of development in different species.
A decade ago it would have seemed impossible to analyze how a toxicant’s initial interaction with a molecule is connected with the ultimate developmental defect. The complexity of development seemed daunting, and there was little knowledge of the activities and interactions of components and their roles in developmental processes. In the past decade, however, the situation for analysis has improved greatly. The activities of numerous components are known, and there are insights into the organization and coupling of processes. The understanding of the early developmental steps of axis formation, gastrulation, and neurulation is increasing, and various examples of organogenesis are available for study. The new information on development provides the framework for analysis of toxicant effects on developmental processes and the connection of dysfunctional processes to structural and functional defects in the newborn. The analysis of development in model organisms has been crucial to the progress in mammals, specifically the mouse. Some efforts in connecting molecular effects to dysmorphologies have been successful. For example, the analysis of cyclopamine (a plant alkaloid) in causing cyclopia (a diminished head and single median eye) in cattle was possible once it was realized that mouse mutants of some components of the Hedgehog signaling pathway are also cyclopic; once the basic developmental studies showed the importance of sonic Hedgehog signal for inhibiting eye development in the ventral midline of the prospective diencephalon, leaving bilateral eye development, the mechanism of the cyclopamine-induced birth-defect became better understood.
Scrutiny of the developmental defects of mutants of developmental components in genetically favorable model organisms, and the scrutiny of toxicant-caused developmental defects, might provide informative parallels. In some mutants, a component is inactive due to a mutated gene; in others, a mutated component can be overactive or underactive, but not absent. If a toxicant interacts with one component and modifies its activity, single-gene mutant studies serve as a guide for the interpretation of how molecular perturbations result in
dysmorphogenesis and other adverse outcomes. The phenotypes for the large variety of mouse null mutants, prepared by targeted gene disruption, are a resource for such analysis. Many of these show well-defined developmental defects at birth, or prenatal death at various stages.
If a toxicant affects several developmental components, the mutant comparisons are not as good as those of one component, although some multiple mutations have been prepared. If a toxicant affects cellular activities, setting off molecular-stress and checkpoint pathways, and causes cell death, comparison mutants could be generated and analyzed to understand the consequences for development.
The committee recommends basic research on toxicant-affected developmental processes in well-understood model organisms. Drosophila mutants can be prepared with a chosen sensitized signaling pathway in a chosen kind of organogenesis, such as the wing or compound eye. In such animals, the effects of toxicants can be most favorably associated with specific processes of organogenesis.
1.3. The committee recommends research to define the genetic and epigenetic basis of variability in human response to developmental toxicants.
Variability is a large problem, covering a variety of issues. It is apparent even in the heterogeneous developmental response of individuals in a group of inbred test animals (rodents) exposed to a toxicant under controlled conditions. Two approaches to address this large problem might be useful: (1) a human epidemiological approach making use of genome information, and (2) a model animal research approach making use of new molecular biological techniques and insights to learn about the sources of variability. The following aspects of individual variability in response to developmental toxicants deserve study:
1.3.1. Individual toxicokinetic differences, especially in the metabolism and transport of chemicals.
Human individuals are known to differ substantially in their levels of DMEs of both the P450 oxidative group and the group of conjugating enzymes. For example, differences are found in ethnic groups from different parts of the world and in individuals with varied lifestyles (smoking and alcohol intake) and nutrition. These differences are being explored rapidly, and data indicate that, in cases of genetic variability, the genes encoding these enzymes might be unusually polymorphic compared with other kinds of genes. In current research, which this committee favors, the differences are being defined in terms of base sequence, protein function (e.g., loss of function, reduction of function, increase of function, and change of function), and the basis for the change (e.g., altered time and place of expression and altered catalytic activity). Differences in metabolism of chemicals by individuals with different gene combinations are also being analyzed, because some allele combinations are strongly synergistic. Current genetic, epidemiological, and genome data can greatly assist in identifying human
polymorphisms, and the Environmental Genome Project has such identification as a goal, as do the National Cancer Institute and various pharmaceutical companies. Genotype-environment interactions are suspected to underlie a variety of developmental defects, and identification of human polymorphism will shed light on those interactions.
As these polymorphic genes are identified, experiments are being done to eliminate them by targeted inactivation in the mouse and to assess the animal’s sensitivity or resistance to various chemicals. Studies are also being done to observe the protective or sensitizing effects of overexpression and ectopic expression of enzymes encoded by these genes. The committee supports this work enthusiastically.
The information about individual differences in the metabolism of exogenous chemicals is being extended to research on model organisms (the mouse, Drosophila, and C. elegans), because the validity of animal test results for predicting human toxicity depends not only upon an understanding of the toxicodynamics of the animals’ developmental process but also on an understanding of the drug-metabolizing capacity of the animal and how it is similar or dissimilar to that of humans.
1.3.2. Individual toxicodynamic differences in developmental components.
Another large domain of possible human variability concerns genetic differences in components of developmental processes. Toxicants might interact with these components. Although signaling pathways and genetic regulatory circuits are particularly attractive for study, there is little information on their genetic differences at this time compared with that on DMEs. However, many mouse mutants in genes encoding such components show developmental defects similar to those in humans, and a few human polymorphisms of components are known to correlate with disease (e.g., Patched mutations and a predisposition to basal-cell carcinoma) and predisposition to developmental defects from certain toxicants (e.g., TGF and tobacco smoke). Some signaling component polymorphisms behave as complex traits with sensitivity to genetic background (e.g., APC mutants in mice represent a Wnt pathway intermediate). Gene locations for many signaling components will soon be known through genetic studies of mouse and zebrafish development, cancer databases, and genome studies. This information will be deposited in widely accessible databases on the Internet.
The current experimental situation is favorable for determining the relationship between mutations in these components and susceptibility to toxicants. Mice offer an opportunity for a survey of the importance of genetic background for toxicant sensitivity and for toxicant specificity of effects. Genetic variants, usually produced by targeted gene knockout (but also by gene replacement to make hypomorphs and gain-of-function types and ectopic expression types), can be made with relative ease in mice, which can be used in developmental toxicity studies. When a null allele of any of a variety of kinds of signaling components is
present in the heterozygous condition, the animal is usually viable and apparently normal, although sometimes with minor developmental defects. These animals could be tested for toxicant sensitivity, to see if the genetic background biases certain outcomes. In general, the connection made between mutational alterations in developmental components and chemical induced developmental defects should be strong.
A further determination to be made is how much of the apparent specificity of the outcome of a toxicant-induced developmental defect is due to the specificity of the toxicant and how much is due to the particular genetic background of the exposed animal. This determination should be applied especially to broad-acting toxicants (e.g., ones causing widespread cell death). In model animals, such as Drosophila, C. elegans, or the mouse, various genetic constructs that are sensitized in various ways (e.g., a slightly reduced Hedgehog pathway or a slightly reduced TGF pathway) could be exposed to the same toxicants to see how the outcomes differ. Already there are human toxicant-induced differences in profiles of birth outcomes in Tp53 (-/-) knockout mice that vary with genetic background. The full implication of these observations has not yet been exploited for assessing developmental toxicants.
Ultimately, it is not known how similar a group of animals can ever be in individual responses to toxicants. In the standard rodent tests for toxicants the response of a group of animals exposed to a single intermediate dose is heterogeneous. The developmental outcome is affected in some animals and not affected at all in others. How much of the heterogeneity is genetic and how much is “epigenetic” (i.e., associated with variable histories of nutrition, disease, stress, or other chemical exposures) is an important issue. Conventional inbred strains, which are more than 98% genetically identical after 20 generations of inbreeding, are a valuable and easily available resource for studying epigenetic contributions.
1.3.3. Individual differences in molecular-stress and checkpoint pathways, which normally operate to counteract failure of cell function.
These components are part of the organism’s line of defense and might be activated by broad-specificity toxicants. The extent to which individual humans differ in their molecular-stress and checkpoint pathway components and, hence, in their responses to environmental chemicals, is unknown. These pathways are known to have important roles in other organisms and to be conserved across phyla. Individual differences in these pathways among humans should be explored. Model animals should be prepared with mutated components of these pathways in order to determine whether their sensitivity to toxicants is increased or decreased. Mouse and Drosophila mutants are already available.
1.4. In seeking to understand molecular mechanisms of toxicity, it is important to clarify how the approaches and information can be applied to a comprehensive assessment of human developmental risk.
The committee has stated that the developments in developmental biology and genomics present an unprecedented opportunity to understand the mechanisms of action of toxicants at a molecular level and that, at some time in the future, perhaps a decade, risk assessors will have primarily “mechanism-based data” from test animals to use in arriving at human toxicity estimates. However, standard toxicant bioassays on mammals do not yet yield comprehensive data on mechanisms and consequences valid for extrapolation to humans.
What can be done in the interim to build on recent advances? In Chapter 8, the committee outlined a multilevel approach to risk assessment that incorporated various assays intended to provide information ranging from molecular interactions to developmental consequences. Advantages and disadvantages of the assays of each level were outlined as well. This approach is briefly described here. There are two domains of information. One contains results from model systems and model-animal tests of toxicant effects on development and of genetic alterations affecting toxicant susceptibility. The results need to be extrapolated to humans. The second domain contains results from human studies of toxicant exposure, toxicant susceptibility (including polymorphisms), and toxicant effect.
There are four levels of model systems for providing information for assessing the effects of toxicants (or the absence of effects): (level 1) in vitro tests and cell tests; (level 2) nonmammalian animal tests of development and the role of genotype; (level 3) mammalian tests of development and the role of genotype; and (level 4) mammalian tests of mechanism and susceptibility. In general, expense increases with each level, and the number of chemicals that can be tested decreases. The questions of extrapolation to humans are greatest at the low levels. As described in Chapter 8, the committee did not develop a tiered approach, but rather showed how information from each level could be used to improve developmental toxicity risk assessments.
Various examples of this information follow.
1.4.1. The metabolism of developmental toxicants.
Knowledge of DMEs is sufficient to devise level 1 tests of the conversions of a large variety of chemicals by a large variety of enzymes. For example, human metabolism genes have been introduced into test cells, such as yeast or human lymphoblast lines, to generate assay systems. Once the specificity range of human enzymes is well characterized, it should be possible to make predictions about the capacity of a battery of enzymes to modify yet-untested chemicals. It is also clear that various animals (Drosophila, C. elegans, and mice) can be constructed with deficiencies or excesses of various metabolizing enzymes to determine whether the developmental toxicity of a chemical is increased or decreased. Much of this work is under way. There is substantial research inquiry about DMEs, their roles, and their synergisms, especially among oxidases and conjugating enzymes. Although the study of the DMEs is relatively advanced, re-
search is still needed on other aspects of absorption, distribution (including the multidrug transporter proteins), and excretion.
In general, this research is an exemplary area for the exchange of information across levels with the ever closer approximation of the fast, inexpensive level 1 tests to human metabolizing conditions.
1.4.2. Toxicodynamics: toxicant effects on developmental components—information about mechanism and susceptibility.
Other assays of toxicants should be used to focus on their effects on developmental processes, particularly on the intercellular signaling pathways and genetic regulatory circuits, which operate repeatedly and pervasively in the development of animals of all phyla. Some assays can be done at level 1, since many signaling components have been identified and their genes isolated. Relevant proteins can be produced for in vitro tests or single-cell tests. For example, to test for agents interfering with Hedgehog signaling, the signal transduction intermediates could be introduced into cultured cells. To some extent, cell functions—such as secretion, entry into mitosis, motility, or specific gene expression—that depend on signaling could be scored for interruption by chemicals. The availability of components for nonanimal assay systems is already considerable. Many signaling components have been mapped on the mouse genome, and their corresponding location in the human genome is predictable and will soon be sequenced.
The committee recognized at the outset that the information about chemical impacts on cell signaling components is of little use for risk assessment if there is no organ and mammalian relevancy. Therefore, a comprehensive approach was envisioned to allow this initial information to be placed directly into an overall assessment framework. For testing the effects of toxicants on the activity of these components in development, level 2 assays on Drosophila, C. elegans, and zebrafish development are incisive. Multiple pathways are used in the development of complex organs. One pathway at a time can be sensitized in a specific aspect of development by genetic means. An animal, when exposed to a toxicant affecting that pathway, will have altered development of that organ, whereas the rest of the animal will probably be unaffected or less affected. Altered development is, then, the scored end point of the toxicant’s effect on the specific pathway. Although the model organisms have different organs from those of humans, the signaling pathways and genetic regulatory circuits that operate in the development of that organ also operate in the development of mammalian organs of other kinds. Thus, the effects of chemicals on fundamental processes, such as signaling and transcription, can be detected. The signaling pathways operating in the various kinds of organogenesis in mammals are known; therefore, a prediction can be made and tested in mammals, using level 3 testing approaches. Because level 2 assays are inexpensive and fast, many compounds can be tested, and patterns of toxicity effects can be recognized in advance of the rodent tests. In the multilevel approach, level 3 tests are ones with specially modified mice (containing sensi-
tized pathways and reporter-gene constructs), and more information is obtained with them than with standard animals. In this way, the levels are connected to each other and kept relevant to risk assessment.
1.4.3. Molecular-stress and checkpoint pathways.
Other assays should be directed toward the detection and characterization of cellular responses to chemicals by way of these defense pathways. Approximately 10 of these conserved pathways are now known. Their activation might be relevant to detecting the effects of broadly acting toxicants on maternal and embryonic cells—that is, toxicants such as antimitotic agents or inhibitors of replication, transcription, or translation—that interact with many targets in many cells.
Research remains to be done to connect the damaging effects of toxicants on cells to the disruption of particular steps of development. The connection could be established with level 2 organisms genetically sensitized in one or more of their molecular-stress and checkpoint pathways in a particular organ. The established connection might provide leads for level 3 tests with mice.
Recommendation 2. The committee recommends investigating how the new information about development and developmental toxicity can address the uncertainties in quantitative and qualitative risk assessment.
The committee believes that the new information and approaches of developmental biology and genomics will be useful in improving the quantitative as well as qualitative components of risk assessment. As they are currently designed, the rodent tests for developmental toxicity are limited in their capacity to provide mechanistic information. They are costly in time and resources, and, therefore, only a small percentage of the more than 80,000 chemicals in commercial use (or the even larger number—about 6 million—of natural products) can be tested. Rodent-test end points are frequently limited to effects on growth, organogenesis, and viability of the conceptus and do not include functional, molecular, or cellular effects, nor do they include early developmental losses. The relevancy of animal toxicity outcomes for humans is often questioned, as is the significance of high dose animal exposure conditions for human exposures. Hence, risk assessors must often resort to large default corrections when extrapolating animal results to define safe exposure concentrations for humans. The validity of the extrapolation of particular test results from animals to humans is itself usually not assessed.
The committee envisions that research directions included in the informational framework established in Chapter 8 will address various existing limits on the data available to risk assessors and have the potential to provide the scientific basis to reduce the magnitude of or replace defaults using mechanism-based extrapolation approaches.
2.1. Qualitative risk assessment: testing a larger variety of chemicals and chemical mixtures.
Expanding the number of tested chemicals is an enterprise in qualitative risk assessment. The new rapid and inexpensive model assay systems have an important potential use in such an enterprise. At level 1, which involves in vitro and single-cell assays, tens of thousands of assays could be run per year to test chemicals as substrates for DMEs, as agonists and antagonists of signaling components and genetic regulators of the kind used pervasively in development, and as triggers of molecular-stress and checkpoint pathways. At level 2, which involves tests of chemical effects on the development of nonmammalian animals, thousands of assays could be run per year. Genetically sensitized model organisms (Drosophila, C. elegans, and zebrafish) equipped with various reporter genes would facilitate analysis. Mammalian relevancy and human applications would be further defined in level 3 tests involving mammals—the mouse being the most favorable because of its ease of genetic modification, its vast libraries of mutants, and the advanced knowledge (among mammals) of its development.
2.2. Qualitative risk assessment: assessing toxicant effects across all stages of development.
As noted in Chapter 2, early fetal loss in human development is frequent (20-30% of initial pregnancies). Although many of these losses may be due to chromosomal aberrations for which there are good chemical assay methods, other mechanisms of early loss are less well understood. Recent observations have demonstrated that, contrary to what was previously believed, toxicant exposures during early times in development can not only result in fetal loss but also specific birth defects and adverse functional impacts. In addition, functional impacts occurring as a result of post-organogenesis toxicant exposure have also not been clearly delineated. Further tests of toxicant impacts during these early and late developmental time points are needed and can build on the rapidly expanding body of knowledge about early events such as axis formation, primitive streak formation, and node regression as well as an expanded understanding of functional deficits.
2.3. Quantitative risk assessment: the toxicokinetic differences of test animals and humans should be characterized to improve extrapolations.
The committee recommends that test animals be better characterized with regard to their differences from humans in DMEs and other toxicokinetic variables. With better characterizations, it can be known whether the test conceptus and the hypothetical human conceptus are indeed exposed to the same chemicals at corresponding concentrations and intervals of development. Many DMEs have been identified, and others are known to exist. The profile of activity of mice (level 3 assays) and humans should be determined so that their similarities and differences are known. Some of this effort is already under way, and the com-
parison will be accelerated by the availability of mouse and human genome data. Proteins involved in chemical uptake across the gut, distribution in the fluid space, multidrug transport of chemicals in and out of cells, and excretion from the body are less well known and should also be characterized and compared.
Moreover, once the differences are recognized, test animals such as mice can perhaps be modified genetically to reduce their differences from humans. Thus, cross-species extrapolations could be improved in this respect. It may not be feasible to eliminate all toxicokinetic differences between the mouse and human, however, the differences will be better known with such approaches when extrapolations are invoked.
In level 2 assays involving nonmammalian organisms, such as Drosophila, C. elegans, and zebrafish, it is also important to know the differences between humans and rodents in terms of drug metabolism to increase the accuracy of extrapolations to mammals. Transgenesis and mutagenesis can be done at high frequency in these animals to make them less different from mammals in their drug metabolism.
2.4. Quantitative risk assessment: toxicodynamic differences of test animals and humans should be characterized to improve extrapolations.
The differences in development of various organisms mostly reflect differences in the time, place, order, and combinations of use of conserved developmental components, such as those of the signaling pathways and genetic regulatory circuits. The committee’s recommendation to make more use of nonmammalian model animals in developmental toxicology is based on the recognition of the conservation, although with the caveat that the scoring of toxicant effects is done in these animals at the molecular level of conserved components, and not at the diversified tissue and organ levels, which are obviously not conserved across phyla.
The extent to which developmental components of different animals differ in their interactions with toxicants is not known. Some components can be exchanged between flies and mice without loss of function, but most have not been tested for interchangeability. Vertebrates also have large genomes containing two or more duplicated and slightly diversified genes for many components for which nonvertebrates have a single gene. These diversified components might differ in their toxicant interactions. The recognition of differences will be aided by the genome databases and by further genetic substitutions in test animals. Once toxicokinetic differences are minimized between mice and humans, modified mice should be tested with a battery of toxicants known to affect humans in order to make sure that equivalent developmental outcomes are obtained. If equivalent outcomes are not obtained, the difference is grounds for further analysis of toxicodynamic comparisons.
Current research in developmental biology, which includes mouse development as the exemplar of mammalian development, will increasingly use compari-
sons between mice and primates. The tests of the toxicant susceptibility of developmental components of mouse mutants will greatly guide research in humans.
2.5. Quantitative risk assessment: low-dose effects of toxicants and chemical mixtures should be better detected and characterized.
Risk assessors need data on toxicant effects covering a wide range of doses. A chemical might affect a variety of development processes at high doses but only one critically sensitive pathway at low doses, making that single pathway the most relevant for overall risk assessment. Studies of model systems—such as Drosophila, C. elegans, zebrafish, and the mouse—could provide quantitative information to improve understanding of such dose distinctions and their basis.
In risk assessment, animal test results obtained at high doses of a toxicant and with a small population of animals are frequently used to estimate the consequences of low doses in large populations. Furthermore, when a group of animals is exposed to the lowest dose of toxicant for which there is an effect, the individuals of the group usually respond in a heterogeneous way. Thus, there are many uncertainties about extrapolation to low doses. Basic questions to be explored in this area include the following: (1) What is the shape of the dose-response curve for developmental toxicants at low, environmentally relevant doses? (2) Can the increased attention on key cell-signaling pathways and genetic-regulatory circuits identify biomarkers useful for defining low-dose responses caused by developmental toxicants? (3) Do the low-dose responders represent variants with genetic susceptibility? and (4) Is there an inescapable nongenetic variability to development?
2.5.1. Low-dose cellular responses revealed through the molecular-stress and checkpoint pathways.
Some developmental toxicants might act primarily by interfering with basic cellular reproduction (e.g., DNA synthesis or mitosis). The conceptus might be more sensitive than the adult because it has a higher frequency of cell division (and fewer cells are in a nondividing differentiated state). Dose effects might be nonlinear. High doses of a toxicant might cause so much cell death that local development is impaired, but low doses might cause so little cell death that cell proliferation by the unaffected cells can restore the population and development is not detectably abnormal. At even lower doses, the various molecular-stress and checkpoint pathways might protect individual cells so that none dies, and development is completely normal. Nonetheless, the activation of the recovery pathways might be detected as an indicator of effect in this low-dose range. The committee recommends that these toxicants be explored over a range of doses, to detemine whether responses can be found in doses too low to cause developmental defects and to reveal the capacity of the conceptus for recovery.
2.5.2. Genetically sensitized animals should be tested for low-dose toxicant effects.
Does the heterogeneous response of a population of test animals to toxicants give a detectable effect that reflects genetic heterogeneity? To determine the role of genetic differences, various mouse strains can be produced with limiting levels of activity of particular developmental components, and these strains can be tested for their sensitivity to a standard set of toxicants representing a variety of suspected mechanisms of action. Such tests would be a measure of whether animals that are genetically close to abnormal development due to their genetic constitution are more sensitive than normal animals to toxicants, and if so, whether a general sensitivity or a specific one is related to the particular limiting component.
Because genetically sensitized models can approximate more closely potentially susceptible members of the human population, risk assessors can use the toxicity data more comfortably. Developmental toxicologists might want to explore the relationship between genetic variation of toxicodynamic components and toxicant sensitivity in model animals before human variants of developmental components are identified in the future. (The association of cigarette smoking, TGF variants, and cleft palate is already an example.)
Improved information on human exposures gained from other areas of study (e.g., improved biomarkers) should provide additional information to predict more accurately the human exposure range for any given chemical or mixture. Such information could then be used to set exposure concentrations for model-animal assessments in levels 2 and 3.
2.6. Quantitative risk assessment: modeling extrapolation from test animals to humans.
The informational framework in Chapter 8 should provide a guide for obtaining the kinds of test-animal data that are needed for a comprehensive cross-species toxicokinetic and toxicodynamic model of exposure and development of test animals, such as mice, to humans. As outlined in Table 8-1B, information from improved human biomarkers of exposure, susceptibility (both genetic and nongenetic), and effect would be used in the model as well. Such models, difficult as they are to devise and fill with satisfactory data, are needed if a chemical’s potential for developmental effects are to be extrapolated to humans in a meaningful way. For example, complex computational models and abundant data are needed to estimate in utero and postnatal exposures in mammals and to link this information with toxicological impacts.
Interest in such models is demonstrated by the efforts of the National Institute of Environmental Health Sciences to link exposure information with mechanistic toxicity data. However, only a few such specialized models exist, and none adequately combines toxicokinetic and toxicodynamic information, especially the information on molecular and cellular impacts. The lack of an adequate frame-
work into which mechanistic data can be incorporated has limited the usefulness of the recent advances in developmental biology for quantitative risk assessment. Instead, default corrections continue to be used, despite recommendations such as those published in Science and Judgment (NRC 1994) calling for the incorporation of new scientific information into the risk assessment process. The committee believes that the framework laid out in this report has the potential to bring the information gained from recent advances in developmental biology into developmental toxicity risk assessment.
Recommendation 3. To improve the interdisciplinary advances in developmental toxicology, the committee recommends that the databases of developmental toxicology, developmental biology, and genomics be better linked on the Internet and that multidisciplinary outreach programs be established for the effective exchange of information and techniques related to the analysis of developmental defects and to the assessment of toxicity for risk assessment.
The committee concludes that increased multidisciplinary efforts and exchanges of information in chemistry and biology are essential to improve risk assessment for developmental toxicity. As mentioned at the beginning of this chapter, developmental toxicology is a broad and complex field.
In recognition of the interdisciplinary nature of future work in developmental toxicology, this interdisciplinary NRC committee was formed. It has been a struggle for the scientists from different relevant disciplines to communicate the research and public-health challenges in developmental toxicology within the committee. Members of this committed group were unfamiliar with each other’s discipline and with the differing connotations of such terms as “mechanism.” The committee soon realized the need for future activities fostering communication and joint research efforts within the scientific community.
3.1. Development of cross-disciplinary, linked databases of relevance for developmental toxicity.
To support the growth of knowledge in developmental toxicology and organize information in a way useful for risk assessment, this committee proposes that cross-disciplinary, linked databases of relevance for developmental toxicology be established with entries from industry, academia, and government. To capture and collate information about chemical toxicants that are important as developmental toxicants, internal organization of the data should reflect knowledge about chemical structure and should include known molecular targets, organotypic effects, and defined associations with developmental anomalies primarily from animal tests but also, when available, from humans. The database should link to genomics databases, for example, with epidemiological information on human variation, such as the large number of human DME polymorphisms. Another important link would be the historical control database for developmental and reproductive toxicity.
Ideally, a separate but linked relational database would be established, grouped by signaling pathways and genetic regulatory circuits, and referred to when chemicals are identified as interacting with an element of the pathway. The database could be helpful in identifying potential biological interactions of a chemical with other chemicals that affect components of the same pathway. A signal-transduction database was recently activated at www.stke.org. This or a similar database should keep track of the involvement of signaling pathways and genetic regulatory circuits in all aspects of development for a wide range of organisms. This information should also be connected to the large and growing database of phenotypes of mouse mutants, many of which are being generated by targeted gene disruption and transgenesis of signaling components or combinations of components. The mouse mutant collection represents the most systematic library of mammalian birth defects associated with known genetic defects. In addition to homozygous null mutants, the library should include phenotypes of hypomorphs, heterozygous null mutants, and suppressor loci, as they become available.
3.2. Enhancement of multidisciplinary research interactions.
The challenges that investigators face when trying to work across fields, such as developmental biology, developmental toxicology, and risk assessment, are a key issue that the committee identified early in its deliberations. This issue previously impeded the successful application of the new scientific information to improve developmental toxicity risk assessment. For the successful application of this report’s findings, the committee believes that multidisciplinary educational and research programs must be conducted. Programs, such as workshops and professional meetings, should be organized so that researchers of developmental toxicology, developmental biology, genomics, medical genetics, epidemiology, and biostatistics can come together to exchange new insights, approaches, and techniques related to the analysis of developmental defects and to risk assessment. By accelerating the necessary research, cooperative research projects would move forward the recommendations of this report.