7
Using Model Animals to Assess and Understand Developmental Toxicity

The recent advances in developmental biology described in Chapter 6 have established the central importance of a small number of highly conserved signal transduction pathways that mediate cell interactions crucial for animal physiology, reproduction, and development. It seems likely that many developmental toxicants might affect development by acting on those pathways. Application of the methods that have been so successful in elucidating them should now allow scientists to investigate that possibility and to determine the mechanisms by which developmental toxicants act. This chapter reviews the experimental approaches primarily responsible for the recent advances in knowledge about animal development and discusses how those approaches might be applied to developmental toxicology. Chapter 8 discusses how those approaches might lead to improved qualitative and quantitative risk assessment.

MODEL ORGANISMS AND THE GENETIC APPROACH

Single-Cell Organisms

Model organisms have been important throughout the study of modern biology. In the 1940s and 1950s, biochemical analysis of bacteria was important in working out the enzymatic pathways of metabolism. In the 1960s and 1970s, bacteria, especially Escherichia coli and its viruses (called phages), provided models for the new science of molecular biology and the elucidation of basic mechanisms for deoxyribonucleic acid (DNA) replication, transcription, and translation in prokaryotes. Since then, the budding yeast Saccharomyces cerevisiae and more recently the fission yeast Schizosaccharomyces pombe have



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 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment 7 Using Model Animals to Assess and Understand Developmental Toxicity The recent advances in developmental biology described in Chapter 6 have established the central importance of a small number of highly conserved signal transduction pathways that mediate cell interactions crucial for animal physiology, reproduction, and development. It seems likely that many developmental toxicants might affect development by acting on those pathways. Application of the methods that have been so successful in elucidating them should now allow scientists to investigate that possibility and to determine the mechanisms by which developmental toxicants act. This chapter reviews the experimental approaches primarily responsible for the recent advances in knowledge about animal development and discusses how those approaches might be applied to developmental toxicology. Chapter 8 discusses how those approaches might lead to improved qualitative and quantitative risk assessment. MODEL ORGANISMS AND THE GENETIC APPROACH Single-Cell Organisms Model organisms have been important throughout the study of modern biology. In the 1940s and 1950s, biochemical analysis of bacteria was important in working out the enzymatic pathways of metabolism. In the 1960s and 1970s, bacteria, especially Escherichia coli and its viruses (called phages), provided models for the new science of molecular biology and the elucidation of basic mechanisms for deoxyribonucleic acid (DNA) replication, transcription, and translation in prokaryotes. Since then, the budding yeast Saccharomyces cerevisiae and more recently the fission yeast Schizosaccharomyces pombe have

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment served as models for intensively investigating the molecular mechanisms of these and other functions unique to eukaryotic cells, such as the cdk-cyclin-based cell cycle, mitosis, meiosis, ribonucleic acid (RNA) splicing, regulation of chromatin structure, secretion, dynamics of the cytoskeleton, stress pathways, checkpoint pathways, and, to some degree, intercellular signaling and differentiation, the last two associated with yeast mating. Most of these cellular functions have been highly conserved during eukaryotic evolution, so that knowledge gained from yeast research is directly applicable to understanding human cell processes. However, understanding the interactions of cells and tissues in development and physiology of higher eukaryotes requires study of metazoans (i.e., multicellular animals). It should be appreciated, though, that as the processes are understood in metazoa, the components of each process can be introduced into yeast and the individual processes reconstituted there for further detailed study. For example, it has been found that a number of human cell-cycle proteins function well in the yeast cell cycle, when replacing the yeast cell’s components. Utility of Model Animals Much has been learned about human development and physiology through the study of model animals, a small set of diverse metazoans that have particular advantages for laboratory research. There are several reasons for their utility. Research on humans and other primates is expensive and limited by ethical considerations. The most commonly studied model animals are relatively inexpensive to maintain and are well suited for experimental manipulation. Most important, as outlined in Chapter 6, recent research has shown that there is a remarkable degree of similarity in the developmental mechanisms of all animals. Not only individual genes and proteins but also entire pathways of signaling and response and their functions in developing embryos appear highly conserved throughout evolution. This means that, although the embryology of simpler animals might appear superficially very different from that of humans, knowledge gained from those models can often be applied directly to understanding human developmental mechanisms. On the other hand, there are important developmental and physiological attributes that can be investigated only in vertebrates, such as the adaptive immune system, or in mammals, such as placentation and lactation. Therefore, it is useful to study a representative range of model animals—from invertebrates that are only distantly related to humans but have particular experimental advantages, to rodents and other mammals that are less convenient but more closely related to humans. Model Animals for Study of Development For study of development, the currently most intensively investigated model animals, in order of increasing complexity, are the free-living soil roundworm

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment (nematode) Caenorhabditis elegans, the fruit fly Drosophila melanogaster, the frog Xenopus laevis, the zebrafish Danio rerio, the chick, and the laboratory mouse. Also particularly useful for certain investigations are sea urchin, sea slug (Aplysia), puffer fish, and a few mammals, including the rat. This set of model animals is somewhat different from those most widely used in the 1950s. Why have these species been chosen for recent intensive study? For four of them, the principal answer is genetics. The genetic approach has become established in the last three decades as one of the most powerful tools for elucidating biological mechanisms. It allows researchers to compare wild type with a mutant phenotype and to identify new genes involved in controlling a biological process and to determine their functions in the organism. Genes that control important functions are identified by mutations that cause defects in those functions. These genes are then mapped, cloned, and identified at the molecular level so that the proteins they encode can be studied using methods of biochemistry and cell biology. This approach has proved to be extremely powerful, not only for basic research in model organisms but also for medical research on heritable human diseases. The approach was followed, for example, in the mapping, cloning, and subsequent study of the cystic fibrosis gene, the breast cancer susceptibility gene, and many others. The four model animals chosen primarily on the basis of their convenience for genetic analysis are C. elegans, Drosophila, zebrafish, and mice. All are relatively small, easy to maintain in large populations in the laboratory, and have short generation times, which allow for rapid analysis of breeding experiments. The remaining animals are not well suited for classical genetic analysis, primarily because of much longer generation times, but have compensating advantages of convenience and manipulability or simplicity. Sea urchins, because of their reproductive properties, have been particularly valuable in studies of fertilization and gene regulation in early embryos. Aplysia are used in nerve growth and development studies. Puffer fish are useful for genomics because of their remarkably small genome size (400 megabases (Mb)) compared with most other vertebrates (about 3,500 Mb, including humans). The frog Xenopus has eggs and embryos that can be obtained in quantity and are relatively large (about 1 millimeter (mm) in diameter). The eggs and embryos are convenient for biochemical analysis as well as microsurgery and can easily be microinjected with cloned genes, RNAs, proteins, drugs, and so forth to study the developmental effects of those molecules. The embryos have been used in toxicant tests, such as the frog embryo teratogenesis assay–Xenopus (FETAX). FETAX is currently under consideration for validation (Bantle et al. 1996; NIEHS 1998). Chick embryos, more closely related to mammalian embryos, are readily accessible for observation and microsurgery (unlike those of mice, which develop in the uterus) and are convenient for tissue transplantation experiments. Putative developmental toxicants can be added directly to the embryo, thereby bypassing the modifying effects of maternal metabolism and selective transfer by the placenta. Rat, rabbit, and

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment guinea pig have long been standard systems for physiological and toxicological investigation. However, because of the power of genetic analysis, the four genetically tractable model animals (C. elegans, Drosophila, zebrafish, and mouse) have become mainstays of recent research in developmental biology and, for the same reason, are also likely to be particularly valuable in emerging approaches to developmental toxicology. These systems are described in more detail below, following a brief review of methods in genetic analysis. Rationale and Strategy of the Genetic Approach Genetic analysis has a powerful advantage in that it can “dissect” functionally and define the important components of any biological process without knowing anything about the process in advance—simply by isolating mutations that affect it, using those mutations to define the genes that control the process, and then cloning and characterizing those genes and their gene products, thereby revealing molecular mechanisms. Over the past two decades this approach has been successfully applied to many aspects of animal development, as indicated in Chapter 6. It can also be applied to elucidating the mechanisms of action of developmental toxicants. The general steps in the standard genetic approach, described below, are sometimes referred to as “forward genetics” (going from the mutant phenotype to the gene) in contrast to the more recently developed methods of “reverse genetics” (going from the gene back to a phenotype) made possible by molecular biology and genomics (see Chapter 5 for some of the genomic methods). Although the terms forward and reverse genetics are now generally accepted, it should be noted that the term “reverse genetics” has had a history of use in earlier medical genetics literature to describe the progression from mapping of a heritable disease state to cloning of the responsible gene (called “forward genetics” elsewhere). Forward Genetics The steps in this approach are as follows: Choose a defective phenotype of interest (e.g., failure to develop a particular structure or increased sensitivity to a toxicant) that is specific and selectable or easily recognizable. Using mutagenized populations, carry out a saturation screen for mutants with the defective phenotype (i.e., a screen large enough so that mutations are likely to be found in every gene required in development of the normal phenotype). Use classical genetic analysis of these mutations to define the genes they represent by genetic mapping and complementation tests and to determine their null phenotypes (i.e., the effects of complete loss of gene function). The incisive-

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment ness of studying null mutants is worth mentioning in the context of developmental toxicology. Their phenotypes match the toxicologist’s ideal of what the “perfect” toxicant would generate for observation if it completely inhibited just one target component of the organism. If possible, establish the order of function of the identified genes by constructing double mutants to determine which of two distinguishable phenotypes takes precedence (epistasis test). Identify additional modifier genes by using suppressor and enhancer screens in a sensitized genetic background for secondary mutations that make the defective phenotype of an existing mutant less or more severe. Using fine-structure genetic mapping and positional cloning, obtain genomic clones of each gene for molecular analysis and verify their identities by demonstrating that each gene in the corresponding mutant animal carries a DNA sequence alteration. From suitable complementary (c) DNA libraries of cloned cDNA copies of the animal’s messenger (m) RNA population, isolate cDNAs corresponding to each gene, sequence them to determine the predicted amino acid sequence of each encoded protein, and carry out a similarity search, comparing those sequences with the sequences available in databases, which often can be used to discern motifs and reveal the functional class to which a protein belongs. (Function was initially deduced for the class from other kinds of studies—biochemical, cellular, developmental, and physiological.) Determine when and where the mRNA and the protein encoded by each gene are found during development by using, respectively, nucleic acid probes and antibodies made to fusion proteins. A faster but sometimes less reliable alternative is to make reporter constructs, which carry the promoter region of the cloned gene fused to a gene encoding a reporter protein that can be detected by its activity (e.g., the E. coli β-galactosidase gene lacZ) or fluorescence (e.g., green fluorescent protein (GFP)). Embryos into which such a construct has been introduced (by DNA transformation) can be observed at various stages to determine when and in what cells and tissues the promoter is active. Generally (but not always), an active provider will reflect the expression pattern of the normal gene. Supplement that information with genetic mosaic analysis, by producing animals in which only certain cells or tissues are mutant, to discover where a gene must normally function and whether its functions are cell autonomous (i.e., intracellular) or cell nonautonomous (i.e., intercellular). Isolate and biochemically analyze proteins encoded by the mutationally identified genes to study further the function of the proteins. All these steps are not always carried out. The most important and difficult step, once mutants have been obtained, has been positional cloning of the gene. However, shortcuts are becoming available with the accumulation of genomic mapping and sequence information and the development of new technologies

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment (see following sections). For example, if a mutation defining a gene of interest has been mapped to a region of the genome for which the entire DNA sequence is known, the “candidate gene” approach can be used to identify it. Computer analysis of the genomic sequence can predict which sequences in the region represent coding sequences and open-reading frames (ORFs) of genes and what proteins these DNA sequences encode. It is then often possible to guess one or a few most likely candidate genes and confirm that one of these is correct by sequencing one (or preferably more) mutant allele and finding the responsible sequence alteration(s) or by expressing the candidate gene to see if its encoded product reverts the mutant phenotype back to wild type. A new method called genomic mismatch scanning (GMS), using DNA microchip technology, will allow more rapid identification of the candidate gene and the mutational lesion in one step. Oligonucleotides representing the entire sequences of all candidate genes in the region to be tested, as well as all possible single base-change mutational variants of each sequence, are synthesized and fixed in an indexed array on a microchip (see description of the method in Chapter 5). The chip is then annealed to differently labeled probes from nonmutant and mutant forms of the cloned gene. By comparing these patterns, both the correct candidate gene and the nature of the mutational lesion can be determined. Reverse Genetics With the increasing availability of genomic sequence information, the following somewhat different approach is becoming more useful for studying biological processes, especially in organisms such as mammals, for which the forward genetic approach is difficult. It is called reverse genetics, because it starts with a cloned gene of potential interest. The cloned gene is then used to obtain animals with defects in the gene or its expression for functional analysis. The steps in this approach are as follows: Identify a gene of interest from its sequence (e.g., the mouse homolog of a developmentally important gene in Drosophila) and obtain a clone of the gene by standard methods based on sequence similarity (such as screening a mouse library (collection) of genomic DNA clones with the cloned Drosophila gene). Determine its expression pattern (as described above) for clues to its function. Inactivate the gene (often referred to as “knocking out,” “targeted inactivation,” or “homologous recombination” of the gene) and observe the phenotypic consequences for more definitive information on function. This can be done either transiently, by injection of an antisense or double-stranded mRNA that specifically prevents gene expression, or permanently (preferable, but requiring considerably more effort), by generating animals that carry a null mutation in the gene. In the nematode C. elegans, the double-stranded mRNA method works

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment particularly well (described in more detail below). In the mouse, mutations can be obtained efficiently by targeted recombination of mutant DNA constructs introduced into the germ line (described in more detail below). In flies (Drosophila) and nematodes (C. elegans), the desired mutant individual can be screened from a large population after random transposon insertion or chemical mutagenesis. Once mutations are obtained, they can be subjected to any of the genetic analyses described above. Again, emerging technologies, such as microchips carrying ordered arrays of cDNAs to allow rapid analysis of how a mutation affects mRNA populations, will accelerate and enhance the above approaches. Extrapolation to Humans For many genes identified by forward or reverse genetics in model animals such as the mouse, and particularly for genes relevant to human disease states, the next step is to isolate and characterize the corresponding (orthologous) gene in humans. Several recent developments have simplified the task of cloning human homologs for molecular analysis. Extensive and detailed maps of molecular markers are now available for many areas of the human genome, and rapid progress is being made on the remainder in connection with the Human Genome Project. Comparison of mouse and human maps demonstrate extensive linkage conservation (synteny) between the two genomes (i.e., the arrangement of orthologous genes has been conserved over large regions from the last common ancestor). Considerable linkage conservation is found even between fish and mammals. As ancestral species diverged hundreds of millions of years ago and evolved into present-day species, local gene order in most instances has been maintained while large blocks of contiguous genes have been rearranged. For example, genes A-B-C-D-E found on mouse chromosome 12 might be found as A-B-C-D-E or, in reverse order, as E-D-C-B-A on human chromosome 7. As a result, if the chromosomal location of a gene responsible for a trait in the mouse is known, it is now possible to predict quite accurately the chromosomal location of its ortholog in humans (see Web site at http://www.informatics.jax.org, under mammalian homology and comparative maps). This approach will also be useful in defining human genes that affect responses to developmental toxicants (e.g., the genes for various enzymes that metabolize exogenous chemicals). There also are large libraries (expressed sequence tag (EST) libraries) of sequences representing pieces of mRNAs transcribed from genes at various times and tissues in the human (see description of EST methods in Chapter 5). Transcripts from almost 90% of all human genes are estimated to be sequenced and present in these libraries. The transcripts are of great value for isolating the human homologs of genes and gene products that have been well characterized in other organisms.

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment THE MAJOR MODEL ANIMALS FOR GENETIC ANALYSIS The four genetically tractable model animals, C. elegans, Drosophila, zebrafish, and mouse, are useful for somewhat different reasons. Relevant characteristics of each are described briefly below, along with some of their experimental advantages and disadvantages (see also Tables 7-1 and 7-3). The potential utility of each animal for identifying and investigating mechanisms of developmental toxicants is discussed later in this chapter. The Nematode Caenorhabditis elegans History, Biology, and Genetics Caenorhabditis elegans is a roundworm found commonly in soils all over the world. It has become widely exploited as a model animal largely because of the early efforts of Brenner (1974), who recognized its experimental advantages and pioneered its genetic analysis. The adult is about 1-mm long, just visible to the naked eye. It feeds on bacteria, such as the common bacterium E. coli, and is easy to grow and breed on agar plates in the laboratory. C. elegans is one of the simplest animals known, with a small fixed number of somatic cells: 959 in the adult hermaphrodite and 1,031 in the adult male. It is transparent throughout the life cycle, so that its entire development can be analyzed in living animals with the light microscope. Its generation time is only 3 days, and development is rapid (Figure 7-1). Embryogenesis is complete by 14 hours after fertilization. The first-stage (L1) larva hatches from the egg, growing and molting through three larval stages (L2, L3, and L4) as its reproductive system develops before the final molt to adulthood. Adult males make sperm and can mate with hermaphrodites, making genetic crosses possible. The hermaphrodites are essentially females but produce some sperm during late larval development and can self-fertilize, which simplifies genetic analysis. C. elegans has a genome size of about 100 megabases (Mb) packaged into six small chromosomes, including five autosomes and a sex (X) chromosome (hermaphrodites have two and males one). Extensive genetic and physical maps have been constructed, and its genome has recently become the first in a metazoan to be completely sequenced under the auspices of the Human Genome Model Organisms Project (C. elegans Sequencing Consortium 1998). The genome includes about 19,000 genes. Because of its transparency and the invariance of cell-division patterns throughout C. elegans development, it has been possible to describe embryonic and larval development completely at the cellular level. By observation of developing animals using Nomarski microscopy, Sulston and coworkers were able to define all the larval cell lineages (Sulston and Horvitz 1977) and later the entire embryonic cell lineage (Sulston et al. 1983), so that the ancestry of every cell in the adult organism is now known. Perturbation of normal development by laser

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment TABLE 7-1 Comparison of Four Model Animals for Genetic Analysis and Humans As a Reference Animal Adult Size (cm) Genome Size (Mb) Period of Organogenesis (d) Generation Time (wk) Experimental Advantages Nematode (Caenorhabdits elegans) 0.1 97 0.2-0.4 0.4 Convenient forward and reverse genetics, complete genome sequence known, complete description of development available, simplicity, transparency Fruit fly (Drosophila melanogaster) 0.4 180 0.5-1 2 Most convenient forward genetics, many genetically defined signaling pathways known, extensive knowledge of development Zebrafish (Danio rerio) 3 1,700 1-4 12 Vertebrate, good forward genetics, transparency, external, well-studied development, accessible to test chemicals in water Mouse (Mus musculus) 6 3,000 6-15 10 Placental mammal, closest model to humans, good forward and reverse genetics, well-studied development For comparison:   Human 170 3,500 14-60 27 yr (1,400 wk)   Abbreviations: cm, centimeter; d, day; Mb, megabase; wk, week; yr, year. ablation of specific cells has provided information on inductive cell interactions during embryogenesis and larval growth. This knowledge has been extremely useful in analyzing the genetic control of cell-fate determination and the roles of cell signaling pathways by using genetic approaches, as described further below. For more comprehensive reviews on current knowledge of C. elegans, see Wood et al. (1988) and Riddle et al. (1997). Transgenic Technologies DNA Transformation. Cloned genes can be reintroduced into the C. elegans genome by injection of DNA into the syncytial region of the hermaphrodite gonad (Mello et al. 1991). The injected DNA recombines to form large replicating extrachromosomal arrays, which become incorporated into developing oocytes

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment FIGURE 7-1 Life cycle of Caenorhabditis elegans. The numbers 10 through 40 indicate hours after fertilization of the egg. L1 through L4 indicate larval stages, each ending in a molt, a shedding of the tough cuticle. The dauer larva is a diapause stage entered when food (usually bacteria) is in short supply. Source: Wood (1999). Reprinted with permission from Encyclopedia of Molecular Biology; copyright 1999, John Wiley & Sons. and then embryos. These arrays can be transmitted to most cells of the resulting animals and through the germ line to their progeny. Although the genes on such arrays are present in high and somewhat variable copy number, they are efficiently expressed and can be useful for many types of investigations, such as transformation rescue in positional cloning and analysis of a cloned gene’s expression patterns using lacZ or GFP reporter-gene constructs. From transmitting lines, more stable integrated lines can be obtained in which the array has inserted randomly into a chromosomal locus, allowing various gene-trapping technologies for identifying loci with tissue-specific expression patterns. Targeted insertion of transgenes by homologous recombination has not yet been achieved. Reverse genetics using targeted gene disruption is therefore difficult but can be accomplished by random transposon insertion (Plasterk 1995) or deletion mutagenesis followed by appropriate screens, or it can be accomplished by RNA-mediated gene interference (RNAi), as discussed next.

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment RNAi. A powerful tool for reverse genetic analysis has been provided by the discovery that introduction of double-stranded mRNA for a particular gene into C. elegans will specifically inactivate that gene, resulting in loss-of-function phenotypes that generally mimic the gene’s null phenotype for at least a generation or two (Fire et al. 1998). Although the mechanism of this inactivation, referred to as RNAi, is not yet understood and gene expression in some tissues is more susceptible to inactivation than expression in other tissues (Montgomery et al. 1999), it is clear that RNAi will be extremely useful for rapid functional tests of genes identified by genome sequencing as potentially important, for example, in development or in responses to environmental toxicants. Moreover, recent results indicate that the technique is applicable to Drosophila (Kennerdell and Carthew 1998) and perhaps to other organisms as well. Signaling Pathways in Development Most of the progress in understanding C. elegans development has come from application of forward genetics as described above, combined with laser ablation experiments to identify required cell interactions. A variety of inductive events, which in C. elegans can be analyzed at the single-cell level, are mediated by signaling pathways that are still under investigation. However, it is already clear that nematode development uses most of the pathways described in Chapter 6, often in developmental contexts similar to those found in more complex metazoans. Two exceptions are the Hedgehog and cytokine signaling pathways, which C. elegans appears to lack (Ruvkun and Hobert 1998). The Fruit Fly Drosophila History, Biology, and Genetics Drosophila melanogaster is the common fruit fly found worldwide in orchards, where adult flies lay eggs on rotting fruit. Since the beginning of this century, fruit flies have been cultured in the laboratory in half-pint milk bottles and more recently in shell vials and plastic tubes by using a solid food, typically composed of agar, cornmeal, dried yeast, and molasses. At 25°C the life cycle takes approximately 2 weeks. Embryogenesis and the first two larval stages require 1 day each; the third larval stage, 2 days; and the pupal stage, 4-5 days (Figure 7-2). Two-day-old adults begin to lay eggs. Because of the short life cycle, ease of rearing in large populations, and the many diverse phenotypes readily visible under a simple dissecting microscope, many mutations have been accumulated in the organism since its initial use by T. H. Morgan and his associates at Columbia University in the 1920s. The study of fly genetics has been instrumental in many classic discoveries in eukaryotic genetics, such as linkage, gene mapping, recombination frequency, and chromosomal aberrations. Discov-

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment Fibroblast-Growth-Factor (FGF) Pathways. The FGF pathway, which is a subset of the RTK pathways, is not used in either the eye or the wing in Drosophila, but it is required for proper tracheal branching, a process that is readily visible in living larvae by a variety of techniques. Consequently, specific interference with this pathway might also be detectable (Klambt et al. 1992). The tracheae are branched airways of the respiratory system of Drosophila. Their development has interesting similarities to angiogenesis in vertebrates, which involves FGF signaling. Stress Pathways. Much of the original characterization of these pathways (e.g., the heat-shock response) was done in Drosophila. For convenient scoring of chemical effects, a variety of transgeneic strains have been constructed with reporter genes that are activated under conditions of stress. Cell-Cycle Control. Drosophila might also be useful for the analysis of compounds that interfere with control of the cell cycle or that activate various checkpoint control pathways. Compounds stored in the egg during oogenesis support nearly all of the critical cell divisions required to achieve the free-living larval stage. During the larval stages, growth of the larva occurs by polyploidization and cell enlargement rather than by cell division. The only dividing cells in the larva are the precursors of the adult structures, and they increase in number prior to metamorphosis in the pupal stage. Most loss-of-function mutations in genes that are required for the cell cycle result in larval lethality, often at the larval and pupal boundary (metamorphosis), when the absence of adult precursor cells becomes critical. Consequently, it should be possible to test various chemicals for their ability to thwart the growth of adult precursor cells. It could be argued that toxicant effects of the cell cycle are better studied in defined cultured conditions with, for example, culture-adapted differentiated cells. Indeed, cell lines are available, or could be prepared, with one or the other of the 17 intercellular signaling pathways functioning in culture, and these would be valuable for assessments of toxicant effects (e.g., effects on activin’s action as an erythroid differentiation factor in human erythroleukemic cell lines). The advantage of a proliferating developing system, such as the Drosophila imaginal discs described above, is that several signaling pathways simultaneously influence proliferation, in their full complexity of signal release via the Golgi and signal modification in the extracellular space, in addition to signal reception and transduction. Thus, the initial net cast to find toxicant effects might be wider than that in a cell culture system. Still, the latter could be very useful in focused identification and analysis of toxicant effects. Possible Assays of Behavioral Development in Drosophila Because of complex courtship behavior, feeding habits, and biological

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment rhythms, flies have become an excellent organism for identifying genes involved directly in behavior. Some of the major discoveries of genes controlling specific behavior traits have been accomplished in Drosophila. For example, the recent identification of genes controlling circadean rhythms in mammals (Takahashi 1995) owes its origin to the genetic screens that revealed the per locus in Drosophila (Konopka and Benzer 1971). Subsequent Drosophila screens have disclosed other components of the molecular clock—such as Timeless, Cycle, and Clock—which are also conserved in other organisms for photo-period regulation. Similarly, genetic screens in Drosophila have identified the mutants, such as Dunce, Turnip, Cabbage, Amnesiac, and Rutabaga that affect associative learning (Dubnau and Tully 1998). These genes have been shown to be part of the adenyl cyclase pathway. Possibly the most elaborate behavioral patterns of Drosophila involve courtship behavior, which includes a series of species-specific activities essential for successful mating (Hall 1998). Starting from these successful pioneering studies, many of which were initiated in the laboratory of S. Benzer, Drosophila has become a useful organism for exploring the genetic basis for alcoholism (Bellen 1998), susceptibility to drugs (McClung and Hirsh 1999), aging (Lin et al. 1998), and other neurobehavioral traits. Because of the simplicity of its life cycle and rearing needs, the fly will continue to be an important first step in the identification of genes controlling a wide variety of behaviors. It should also be extremely useful for detecting deleterious effect of toxicants on specific behaviors, and as described above, the tests could be done on sensitized strains. Zebrafish The zebrafish is an appropriate model organism in which to test potentially toxic chemicals for several reasons. First, the zebrafish is a vertebrate and therefore of more direct relevance to the role of pathways in development of vertebrate-specific tissues and organs, such as the neural crest (Kelsh et al.1996; Schilling et al. 1996) and parts of the heart (Stainier et al. 1996), which are affected in many congenital anomalies. Second, chemicals can be added directly to the water (Stainier and Fishman 1992). Third, the zebrafish embryo is transparent, so tissue and organ development can readily be assayed. Viable, preferably dominant, mutations would be needed as the sensitized target to make such screens achievable on a large scale. Some viable pigmentation mutations have different heterozygous and homozygous phenotypes, indicating a dose responsiveness to the defect (Haffter et al.1996), and therefore are candidates, but these mutations have not been cloned, so the affected pathways remain to be determined. The potential is great for devising assays for chemical effects on development. The zebrafish embryo has been used for toxicological assays (Ensenbach and Nagel 1995, 1997; Henry et al. 1997; Mizell and Romig 1997); however, it has not been developed for sensitized assays. Such assays are now feasible and would

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment be advantageous. Also, there are now a number of mutants for components of signaling pathways (e.g., BMP2, Nodal, and Cripto). One of the original hopes for the study of zebrafish was that it would permit the genetic dissection of behaviors, including learning. In principle, the transparency of the embryo offers the opportunity for simultaneous assay of neuronal activity. Several behaviors are established early during development. For example, the embryo becomes motile and will respond to touch within the first day of postfertilization (Saint-Amant and Drapeau 1998), and eyes follow a striped drum (termed the optokinetic response) by 3 days (Brockerhoff et al. 1995). Screens have already identified genes that modify these activities (Brockerhoff et al. 1995; Granato et al. 1996). Mutations that perturb locomotory behavior have been shown to affect a variety of sites, including receptors, the CNS, or muscle (Granato et al. 1996). Rhythmic activity, such as circadian rhythms (Cahill et al. 1998) and cardiac pacemaking (Baker et al. 1997), are embryonic in their time of onset and can be dissected by genetics (Baker et al. 1997). Whether these behaviors may be modified is not known, but modification would provide a means to garner genes for learning, addiction, and memory. The effects of chemicals on the development of these behaviors remain to be examined, but the availability of specific behavioral assays is at hand. A radiation hybrid map of the zebrafish genome has been completed, the map coverage being 81.9% of the genome. The map is based on a panel of 94 radiation hybrids (Geisler et al. 1999). A large-scale insertional mutagenesis screen in the zebrafish, with the goal of isolating about 1,000 embryonic mutations, is under way (Amsterdam et al. 1999). This approach is similar to the Nobel Prize-winning approach of isolating a large number of Drosophilia mutants—described in an earlier chapter. Mouse Transgenic Animals and Developmental Toxicology Mice and rats have long been the mammals of choice for toxicological tests. Their advantage over the previously discussed model organisms is their similarity, as mammals, to humans. An advantage of mice over rats is the advanced state of the procedures for genetic manipulation and the large number of mutant and inbred strains already available. The disadvantage of mice and rats, compared with the other model organisms presented above, is their expense. To use them to test tens of thousands of compounds under a range of exposure conditions or in combinations is not feasible. Thus far, little use has been made of sensitized strains and reporter strains to improve toxicant detection and to learn more about mechanisms of toxicity. Furthermore, little analysis has been done on the differences between mice and humans that lower the validity of cross-species extrapolations. There are

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment likely to be two main kinds of differences: (1) differences in the steps by which the active or activated toxicant is introduced to the embryo or fetus, and (2) differences in the components of developmental processes, making the toxicant have more effect or less effect on developmental function and resulting in a more severe or less severe developmental defect. Enough is now known about absorption, distribution, metabolism, and excretion, especially regarding drug-metabolizing enzymes (DMEs), to make systematic comparisons between humans and rodents possible. Many components of signaling pathways and genetic regulation, which are central to development, are also now known in test animals and humans, and those components could be compared systematically as well. As summarized by Malakoff (2000), the use of mice in biomedical research and drug testing is expected to increase dramatically in the next few years, especially as the sequencing of the mouse genome approaches completion, and the increased use of mice will create opportunities in developmental toxicology. Gene-Deletion Transgenics for the Study of DMEs Only within the last 5 years have transgenic methodologies been brought to bear on the reciprocal relationship between the animal’s genetic constitution and its susceptibility to developmental toxicants. A particularly important and approachable set of genes are those encoding enzymes that metabolize exogenous chemicals. As discussed in Chapter 5, there are two major categories, the oxidizing enzymes (mostly P450 proteins) and the conjugation enzymes. There might be tens of these enzymes that metabolize most chemicals and hundreds more that metabolize a few chemicals each. In addition, several known transcription factors activate the expression of genes encoding these enzymes, and some of these factors themselves bind exogenous chemicals. Much remains to be learned about the role of the various enzymes in generating and removing active toxicants, and the gene-deletion approach in the mouse has great advantages. At least in some cases the knockout mice are viable and fertile. One pioneering example in which gene-deletion transgenics was used concerns the dioxin-inducible mouse gene battery (Nebert and Duffy 1997), a group of genes believed to play an important role in developmental toxicity. Much more of this analysis can and should be done, both to understand the role of these enzymes in potentiation and detoxification and to define human and mouse differences for better-informed extrapolations of animal-test data. Overexpression Transgenics for the Study of DMEs Another approach to the study of developmental toxicity using transgenic animals is to overexpress, or ectopically express, a gene of interest in the embryo and fetus. For example, one could ask whether overexpression or ectopic expression of a gene encoding a particular oxidizing or conjugating enzyme either sensitizes or protects the embryo and fetus from the adverse effects of developmental

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment toxicants. Expression of transgenes can be driven by ubiquitous, constitutive promoters, such as β-actin, to induce expression in most, if not all, cells of the embryo. Alternatively, transgene expression can be limited to specific tissues by using tissue-specific promoters. Finally, transgene expression can be limited to specific stages of development and specific tissues by using one of several inducible expression systems such as the Cre-recombinase described above. Transgenes as Biomarkers for the Activation of Stress, Checkpoint, and Apoptosis Pathways Transgenic mice might also play a role in testing drugs and chemicals for potential developmental toxicity (see further discussion in Chapters 8 and 9). As scientists learn more about normal development and about the mechanisms of developmental toxicity, in part from studies using gene deletion or overexpression approaches, sufficient information should become available to construct transgenic mouse lines designed to contain biomarkers of the animal’s toxic response to a chemical. For example, a transgenic mouse could be constructed that contained a transgene consisting of the heat-shock promoter linked to an appropriate reporter gene, as has been done in Drosophila. This reporter, in turn, could be used to determine whether specific drugs and chemicals induce a stress response, which is often associated with developmental toxicity. Such a biomarker transgene could be used for in vivo developmental toxicity studies, or cells from appropriate tissues could be used for in vitro testing. Other stress and checkpoint pathways could be connected to reporter genes as well, as could components of apoptosis. Signaling Pathways as Developmental Targets of Toxicants As noted above, the mouse is at present the animal of choice for the selective knockout or replacement of genes. A large number of genes encoding components of many signaling pathways have now been eliminated, one at a time, as summarized in Chapter 6 (Tables 6-4 and 6-5), and the phenotypes of the single-gene homozygous null mutants have been examined. Surprisingly, many of these mutants achieve advanced development, living past birth, and some reach fertile adulthood. Many have discrete developmental defects. The current explanation of the viability of these mutants is that vertebrates have large, partially diversified families of genes for most signaling components, and the genes, which are expressed at different times and places in the embryo and fetus, encode proteins of partially redundant function. Defects tend to occur at times and places where there is no overlap of expression. A number of double and triple mutants have been made as well, and they have more severe effects. Sensitized test mice can certainly be prepared for the testing of toxicant effects on development (i.e., mice with a particular pathway operating at or near the limiting rate in a designated developing tissue). Animals can also be prepared with lacZ or GFP reporter genes to give enhanced readout of effects on pathways.

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment Assays of Altered Behavioral Development in the Mouse As noted in Chapter 2, developmental defects include functional as well as structural defects, but functional defects are less well diagnosed and less well understood. Behavioral defects comprise a large category of functional defects. The testing of environmental agents in animals for behavioral defects as an outcome of prenatal exposure is sometimes done, but not routinely. If an agent is suspected to have neurotoxic effects, a developmental neurotoxicity battery is required, and the suspected agents are mostly pesticides, fungicides, and rodenticides (see Makris et al. 1998; EPA 1998f). In the developmental neurotoxicity tests, pregnant rodents are treated with the agent from day 6 after conception until 10 days after birth to include transfer of the chemical in the milk, and then the pups are tested in various ways from day 4 to day 60. Tests include a functional observational battery (FOB) and specialized behavioral tests. The FOB includes cage-side observations of arousal; autonomic signs; convulsion; tremors; gait; mobility; posture; rearing; stereotypy; responses to touch, approach, tail pinch, and clicks; foot splaying; grip strength; righting reflex; body temperature; and body weight. The specialized tests include those of motor function, sensory function, and cognitive function. Motor-function tests involve tests of grip strength, swimming endurance, use of the suspension rod and rotorod, discriminative motor function, gait, righting reflex, and various ratings of spasms or tremors. Sensory-function evaluations include discrimination conditioning and reflex modification related to auditory, visual, somatosensory, and olfactory inputs and pain sensitivity. Cognitive-function tests include habituation of the startle reflex and classical conditioning related to the nictitating membrane, flavor aversion, passive avoidance, and olfactory conditioning. Twelve chemicals, mostly pesticides, were tested by the above protocol, reviewed retrospectively as a group, and various developmental neurotoxic effects were distinguishable (Makris et al. 1998). Concordance was found between structural alterations of the nervous system and behavioral defects. The difficulties of interpreting the developmental causes of behavioral defects were discussed. In the effort to assess functional defects, the committee favors greater inclusion of behavioral tests with other tests of toxicants. It would be useful to calibrate these tests with mice of known genetic defects affecting aspects of neurodevelopment. Null knockout mutants in homozygous and heterozygous condition could be used. The importance of modifier genes in the genetic background of each transgenic mouse line cannot be underestimated. It would also be useful to calibrate the tests with animals treated with doses of toxicants just below the level giving detectable structural defects in CNS development.

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment TABLE 7-3 Advantages and Disadvantages of Four Model Animals for Toxicant Testing Organism Advantages Disadvantages C. elegans Relatively inexpensive, short life cycle, small size, large populations, extensive information on genetics, genomics known Some signaling pathways are apparently not represented (e.g., Hedgehog); short life cycle, so test compounds might persist long enough to complicate the study of sensitive developmental periods and of primary versus secondary effects Drosophila Relatively inexpensive, short life cycle, small size, large populations, extensive information on genetics, genomics known During pupal development organism is inaccessible to toxicants; short life cycle, so test compounds might persist long enough to complicate the study of sensitive developmental periods and of primary versus secondary effects Zebrafish Simple vertebrate, transparent, external embryos, best organogenesis model Relatively expensive, long life cycle, genetics and genomics research in progress Mouse Mammalian model, most similar to humans, targeted gene replacement possible Relatively expensive, long life cycle, embryonic development in utero and not as easily accessible for manipulation as other model animals Comparative Utility of Model Organisms In summary, each of the four genetically tractable model organisms—C. elegans, Drosophila, zebrafish, and mouse—offers promise for development of tests to identify potential developmental toxicants, and each has advantages and disadvantages, as summarized in Table 7-3. A useful approach, which the committee will present more fully in Chapter 8, might be to develop the more-rapid and inexpensive systems, namely, C. elegans and Drosophila, for preliminary large-scale assays, which can be followed up by more definitive tests using the vertebrate models, zebrafish and mouse, which, although slower and more costly, are more likely to give results relevant to humans. GENE EXPRESSION AS DETERMINED BY IN VITRO AND ENGINEERED CELL TECHNOLOGIES Mammalian cells in culture, as an inexpensive and fast complement to whole-animal (e.g., mouse and rat) studies, can be used for a wide variety of specific

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment purposes related to mechanisms of developmental toxicity and polymorphic variation in susceptibility of toxicants. The incisive use of cultured cells has been enhanced greatly by molecular methods in which genes of choice (from any organism including humans) can be introduced into such cells, for example, to express variant forms of DMEs and to investigate the relation of function to allelic variation. Human allelic variants code for gene products different from those encoded by the wild-type allele, and it is important to characterize and, if possible, quantify any functional alterations. The fundamental techniques available for studying gene expression include (1) expression in vitro where DNA from the allelic variant or wild-type allele is transcribed to the mRNA and then translated into a functional, active protein in a cell-free extract; (2) high-level expression in cells from which proteins can be purified with relative ease; and (3) expression in eukaryotic cell culture in which the cell or the cell’s DNA has been modified (i.e., genetically engineered). These techniques rely on the gene (or at least the cDNA-encoded “coding region” responsible for generating the amino acid sequence of the protein). The cDNA encoding the allele being studied must first be cloned into an appropriate plasmid vector that includes regulatory sequences that drive and terminate transcription and a selectable marker gene. As outlined below, each of the aforementioned expression strategies is used by the investigator to meet specific needs. In general, these in vitro and cell-culture systems are very useful, because they are relatively quick, efficient, inexpensive, and use simple technologies when compared with the development of mouse lines or other animal-model systems. There is a major shortcoming, however, in the in vitro and cell-culture expression systems. Single-nucleotide polymorphisms (SNPs) outside the amino-acid-determining region of the gene (e.g., splice junctions; promoter sequence; 5′ and 3′ untranslated regions; and enhancers upstream, downstream, and inside the gene) can have striking effects on expression of the gene under study, and the effects would generally not be realized, characterized, or quantitated by the in vitro and cell-culture expression systems (Nebert 1999). Expression In Vitro There are both reticulocyte lysate and wheat-germ lysate combined transcription and translation kits now commercially available to assess the interaction of the protein under study with another purified protein. To use these systems, a plasmid containing the cDNA of interest and a radiolabeled amino acid are added to the lysate. Although the protein of interest is not expressed at high levels relative to the total amount of proteins in the lysate, it is the only radiolabeled protein in the reaction mixture.

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment High-Level Expression in Cells Proteins or enzymes of interest can be expressed at very high levels, which can be advantageous for studying certain functional alterations. The most commonly used systems include plasmid-transformed bacteria (Oudenampsen et al. 1990), baculovirus-infected insect cells (Kost and Condreay 1999), and vaccinia-virus-infected mammalian cells (Chakrabarti et al. 1985; Eckert and Merchlinsky 1999). All three systems allow high-level expression of cloned genes. In general, bacterial expression is the easiest to use. However, both the baculovirus-infected and the vaccinia-virus-infected expression systems are eukaryotic, and it might be particularly important to use them if post-translational processing or intracellular accessory factors are needed in the production of a functional gene product. Moreover, proteins expressed in any of the above systems can be given “tag sequences” to allow for rapid, specific purification. Transient and Stable Transfections of Mammalian Cells in Culture Yeast (Oeda et al. 1985), African green monkey kidney fibroblast COS cells (Zuber et al.1986), and vaccinia virus (Battula et al. 1987) were among the earliest expression systems in vivo. They have been successfully used to study DMEs. Allelic DNA can be transfected (passed into the cell) by microinjection or by chemical-DNA aggregation methods including calcium phosphate precipitation and liposome-mediated transfection. By using such methods, followed by antibiotic treatment to isolate cells that house the plasmid containing the selectable marker gene, cells can be transfected either transiently or stably. In transient transfections, expression of the gene is generated from extrachromosomal copies of the transfected plasmid and persists until the expression plasmid is degraded or diluted by cell passage. In general, 5% to 50% of all cells in culture contain the incoming gene, the DNA is not stably “integrated” in the cell’s genome, and the transfected cells contain many copies of the new genetic material. In contrast, for stable transfections, the incoming DNA is integrated (albeit randomly) into the cell’s genome. Cells expressing the gene under study are initially selected on the basis of co-expression of a gene that provides antibiotic resistance. After antibiotic selection, continued cellular propagation in the presence of the antibiotic will ensure that the gene of interest is expressed in a more or less permanent (i.e., stable) fashion—remaining after the cells are passaged on through many additional generations. The copy number of integrated genes is highly variable and can range from one to several dozen, and the genes are generally arranged in tandem (head to tail) at an “integration site” that normally cannot be directed, or controlled for, on an experimental basis. For better gene expression in culture, one or more introns (even when they are artificially inserted) have been found to enhance gene activity in cultured cells (Palmiter et al. 1991); alternatively, small genes with fewer than 10 and

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment usually fewer than 5 exons and introns can often be transfected transiently or stably into cells in culture as the complete gene. Dozens of promoters have been studied over the past 15 years. They include the Drosophila heat-shock promoter (Hsp), the mouse mammary tumor virus long-terminal repeat (MMTV LTR) having a glucocorticoid response element (GRE), enhancer sequences from simian virus 40 (SV40) and herpes simplex virus type 1 (HSV-1), human cytomegalovirus (hCMV), thymidine kinase (Tk), and locus control regions of the metallothionein gene (LTR-MT) displaying variable potencies in driving transcription (Blackwood and Kadonaga 1998; Makrides 1999). Both transient and stable expression of genes in mammalian cells have many advantages. First, genes are expressed in a native environment so post-translational modifications and subcellular targeting are authentic. Second, many transformed mammalian cell types are available for transfection. This allows for the selection of cell types—with the proper intracellular accessory factors or proteins—for enzyme activities that most closely resemble their in vivo counterparts. Finally, expression of gene products can be controlled by an increasing number of eukaryotic promoters. Thus, the levels of expression, including inducible expression, can be controlled. Generations of cell lines that stably express DME genes represent the new generation of pharmacological and toxicological test systems (Langenbach et al. 1992). Expression of stably transfected DME genes in Epstein-Barr virus (EBV)-transformed human B-lymphoblastoid cell lines has provided cell lines that scientists can use to study and categorize numerous foreign compounds. Those drugs or chemicals can be classified according to the intermediates formed by way of the different metabolic pathways. More than a dozen EBV-transformed human B-lymphoblastoid cell lines are already commercially available (Crespi and Miller 1999). They contain anywhere from 1 to 12 human P450 and other DME genes (i.e., cDNAs), which retain their substrate specificity with regard to the metabolism of particular drugs or classes of pharmaceutical agents. Some of these cell lines are already being used by pharmaceutical companies to determine whether a newly developed drug is metabolized by a particular DME. If, for example, the new drug is shown to be a CYP2D6 substrate, it is already known that humans differ by 10-fold to more than 30-fold in the activity of that enzyme (see Chapter 5). Poor metabolizers (the PM phenotype, which comprises 6% to 10% of Caucasian populations) would thus be expected to metabolize the new drug more slowly than extensive metabolizers (EM phenotype) and much more slowly than ultra-metabolizers (UM phenotype). If the parent drug is toxic, the PM individual would be at increased risk; if the CYP2D6-mediated metabolite is toxic, then the EM and especially the UM individual would be at increased risk. Dosage of the drug can therefore be adjusted to the patient’s genotype before the physician prescribes the new drug. For drugs already on the market, molecular epidemiologists can search for possible associations between reported differences in birth defects or other developmental problems and genotype of the

OCR for page 151
Scientific Frontiers in Developmental Toxicology and Risk Assessment mother (or child, fetus, or embryo)—keeping in mind the range of drug doses that might have been given. During the next decade, dozens more of the DME genes (alone and in combination) are likely to be expressed in the human B-lymphoblastoid cell lines or in other similar stably transfected cell backgrounds to determine which enzymes are responsible for either detoxification or metabolic potentiation of a particular drug under study. This information will be useful in the future of developmental pharmacology and toxicology, as well as molecular epidemiology. SUMMARY The committee has evaluated the state of the science for elucidating mechanisms of developmental toxicity and concludes that such elucidation, although not yet realized, can be achieved in the next decade for the following simple reasons: Developmental biology has reached the molecular level of mechanistic explanations. The accumulation of new and relevant information about vertebrate development is rapid (assisted greatly by research on model organisms, such as Drosophila and C. elegans). The accumulation of genome sequence data for humans, mouse, rat, and Drosophila is rapid, adding to that already available for C. elegans, yeast, and many prokaryotes. Information on human polymorphisms and rare variants and their disease relatedness is increasing rapidly, as are data on the ever-increasing library of mouse mutants. The methods are powerful and widely applicable, and the species barriers to comparative study have been greatly reduced in the past few years. The committee begins in this chapter, and continues in the next, the third charge to evaluate how this information can be used to improve qualitative and quantitative risk assessment. In this chapter, the committee summarizes some of the techniques for modifying model organisms, including the mouse, for effective use in assays evaluating agents for potential developmental toxicity and for elucidating mechanisms of toxicity. The committee concludes that the methods and background knowledge are at hand to make incisive comparisons of humans and model animals so that the extrapolation of results from model animals to humans can be more accurate and useful for risk assessment.