The U.S. National Toxicology Program evaluation of transgenic mice as predictive models for identifying carcinogens.

National Institute of Environmental Health Sciences researchers have invested considerable effort in exploring the utility of transgenic mice to detect carcinogens and study mechanisms of carcinogenesis. Work has assessed several mouse models genetically altered to enhance their expression of chemically induced tumors. Results with the p53def (hemizygous for the tumor-suppressor gene) and the Tg.AC (carrier of an activated H-ras oncogene) mice have been used as a basis for a proposed new strategy for identifying chemical carcinogens and assessing risk. The U.S. National Toxicology Program is conducting a series of studies with these two transgenic strains to further examine their strengths and weaknesses for identification of documented rodent and human carcinogens and to explore their ability to provide information concerning the effective dosimetry for target organ mutation.


History
Federal regulatory agencies such as the U.S. Environmental Protection Agency and the U.S. Food and Drug Administration evaluate all available toxicity data on specific chemicals or drugs to establish potential human risk through their use or exposure to them. For these agencies and other groups with human health interests, specific types of data are needed to best provide an adequate chemical toxicologic evaluation. Two sources of the more critical data are epidemiologic studies and the long-term rodent bioassays.
The most relevant data come from epidemiologic studies because they associate human health effects with environmental conditions. When a single environmental agent can be shown to be directly related to a specific health problem in an exposed population, the risk to humans for that agent is clear. The fundamental limitation of epidemiology is that it is difficult to relate a health problem to one environmental agent because most exposures are to mixtures of agents and there is an inherently low power to detect an effect. In addition these studies are retrospective evaluations of hazards (1). An additional complication to the interpretation is that any study must account for differences in lifestyles within the exposed population.
In contrast, animal studies allow single-agent evaluations using appropriate exposure routes under controlled environmental conditions and these studies may be prospective. The 2-year bioassay using both sexes of two rodent species is the current protocol most commonly used to produce chemical carcinogenicity data. This protocol has become fairly standardized (2)(3)(4)(5) and data from these studies generally are required by regulatory agencies to evaluate chemical toxicity from long-term exposures. It has been estimated that approximately 70,000 chemicals currently are in use in commerce (6), most without appropriate toxicity data. The number of bioassays completed per year is limited by the magnitude and length of time it takes to complete the studies. Thus it would take decades to provide the data needed to establish human risk on all chemicals using the bioassay alone. Yet with this number of chemicals and the widespread uses of many of them, there is potential for human exposure and it is important to provide toxicity data to protect public health.
The U.S. National Toxicology Program (U.S. NTP) has directly or indirectly provided a large component of the basic scientific data including the 2-year studies that other federal and state scientific and regulatory agencies use to protect human health. Although still considered the best test model for obtaining carcinogenicity data, the 2-year rodent bioassay is limited because it uses large numbers of animals and normally takes 5 to 8 years to complete and evaluate before results are reported, after which results must be extrapolated to humans.
One of the primary objectives of the U.S. NTP is to develop and validate improved test methods, including alternative test systems (7). In addition, the National Institutes of Health Revitalization Act of 1993 directed the National Institute of Environmental Health Sciences (NIEHS), the National Institutes of Health's component of the U.S. NTP, to establish criteria for the validation and regulatory acceptance of alternative testing methods and recommend a process through which scientifically validated alternative methods can be accepted for regulatory use (8).
One approach the U.S. NTP is taking to meet these objectives is evaluation of alternative models to identify toxic agents. The overall aim of these evaluations is to develop reliable methods to test agents for carcinogenicity in a shorter period of time using fewer animals. The impetus for the transgenic mouse evaluation came in part from discussions at the U.S. NTP workshop "Mechanism-Based Toxicology in Cancer Risk Assessment: Implications for Research, Regulation, and Legislation," which was held [11][12][13] January 1995 in Chapel Hill, NC (9). At this workshop, Dr. R. Tennant presented preliminary data from studies under way in his and other NIEHS laboratories that suggested Tg.AC and heterozygous p53def transgenic mice might be good models to rapidly identify agents with carcinogenic potential. These studies have now been published (10,11). Conclusions from the workshop were that the preliminary results from transgenic mice were promising and that further studies including validation of the models should be considered.
Environmental Health Perspectives * Vol 106, Supplement * February 1998 Models The Tg.AC transgenic line was produced in FVB/N mice by pronuclear injection of a v-Ha-ras transgene linked to a fetal 4-globin promoter and an SV40 polyadenylation/splice sequence (12). Tg.AC mice behave like genetically initiated mice, rapidly developing epidermal papillomas in response to topical tumor promoter or carcinogen treatment. In some experiments the time to first tumor appearance has been as short as 3 weeks, but all chemicals tested that produced papillomas did so in less than 20 weeks. A dose-response relationship has been observed with promoters and carcinogens studied to date and in some studies mice have developed papillomas too numerous to count. Using the in vitro Salmonella assay as a measure of mutagenicity, Tg.AC mice appear to respond to genotoxic as well as nongenotoxic carcinogens. Although the response in treated mice is dramatic, untreated singly housed Tg.AC mice do not usually develop any spontaneous tumors and the histology of the skin is normal. With respect to mechanisms the v-Ha-ras transgene is not significantly expressed in nontumor-bearing Tg.AC tissues but is overexpressed in the proliferating component of benign and malignant tumors. Thus, it appears that expression of the transgene drives proliferation and subsequent tumor development in carcinogen-treated Tg.AC skin.
The heterozygous p53def mouse line has one functional wild-type p53 allele and one inactivated nonfunctional allele. The p53 gene is critical to cell cycle control and DNA repair and one often mutated or lost in human and rodent tumors (13). Mice with a single copy of the wild-type p53 allele (p53+1heterozygous) offer a single target for mutagens, a condition analogous to that in humans with some heritable forms of cancer. The hemizygous state should increase the probability for either loss of p53 tumor-suppressor function or gain of transforming activity by requiring only a single mutation.
Heterozygous p53def mice are viable and show a low background tumor incidence up to almost 12 months of age. This transgenic model is particularly attractive because in limited studies the same chemical-specific target organs appear to respond as in the 2-year bioassay; mice respond within 6 months and small numbers of mice can be used to identify the most serious potential human health risks, i.e., chemicals that are mutagenic transspecies carcinogens (10). These features together with the observation that the p53 gene has been shown to be altered in approximately 50% of human cancers, suggests that this model could be considered as an adjunct with the short-term assays for mutagenic chemicals (Salmonella and in vivo rodent micronucleus assays) as a way of detecting potential human carcinogens. Studies completed to date suggest that both the Tg.AC and the p53def transgenic models are stable within the period of time that they were studied (10,11). The preceding discussion about interpretation of preliminary NIEHS transgenic mouse studies is illustrated in Table 1.
Although U.S. NTP researchers are fully aware of the simplistic notion and the limitations of the distinction typically drawn between mutagenic and nonmutagenic carcinogens, the operational use of these categories was thought to provide a convenient framework for testing these models. The Salmonella parameter was used as a general index to discriminate among mutagenic and nonmutagenic chemicals because extensive studies have demonstrated that of all the available genotoxicity assays the Salmonella assay segregates most unambiguously with mutagenic potential. In addition the Salmonella assay has the highest predictivity for rodent carcinogenicity of four in vitro systems commonly used to identify chemical mutagenic potential (14).

Hypotheses
These preliminary transgenic mouse study results combined with those from the U.S. NTP 2-year study data lead to the following hypotheses. * Chemicals that are both positive in the Salmonella in vitro assay and show cross-species carcinogenic potential in F344 rats and B6C3F1 mice in 2-year studies would produce tumors in heterozygous p53def and Tg.AC mice. * Chemicals that were negative in the Salmonella in vitro assay but showed cross-species carcinogenic potential in F344 rats and B6C3F1 mice in 2year studies would be negative in the heterozygous p53def but positive in Tg.AC mice. Chemicals that were either positive or negative in the Salmonella assay and showed no carcinogenic potential in F344 rats and B6C3F1 mice in 2-year studies would be negative in heterozygous p53def and Tg.AC mice.
If the pattern of tumor response can be validated with additional studies, these two transgenic mice strains may possibly be used as a screen to identify chemicals with and without carcinogenic potential. In addition these studies have the potential to provide preliminary information about chemical-specific mechanisms involved in tumorigenesis.

Study Objectives
Based on the transgenic mice study data (10,11) presented and recommendations of participants at the "Mechanism-Based Toxicology in Cancer Risk Assessment: Implications for Research, Regulation, and Legislation" workshop (9) and following the mandate (8) to develop alternative models, the U.S. NTP decided to test the hypothesis above by evaluating the heterozygous p53def and Tg.AC transgenic mouse models further to determine their potential for use in the program.
The purpose of the U.S. NTP studies is not to simply validate the models versus the rodent 2-year assay but rather to further examine their strengths and weaknesses for identification of documented rodent and human carcinogens and to explore their ability to provide information concerning the effective dosimetry.
To more fully evaluate the apparent potential of transgenic mouse models, studies must be conducted to develop a database of responses of the heterozygous p53def and Tg.AC strains to the same chemicals. Only chemicals showing crossspecies carcinogenic potential in 2-year studies have been tested in the transgenic mouse lines retrospectively. Chemicals producing different responses, e.g., those that cause increased tumor incidences in a single species or sex or in only a specific

Chemical Selection
Chemicals for the initial U.S. NTP studies were selected to determine if the models detect human carcinogens, discriminate between carcinogens with different mechanisms of action, or detect carcinogens administered by more than one route of exposure.
When U.S. NTP data were not available for a particular chemical, the literature and medical reference books were used to establish dose levels and target organs, e.g., for the human carcinogens. However, where possible, evaluations of the transgenic mouse model use the U.S. NTP's historical 2-year rodent carcinogenicity and Salmonella assay databases as resources to select chemicals. Candidate chemicals were placed into one of four groups based on results of the 2-year and Salmonella studies. Chemicals for consideration were restricted primarily to the more recent 2-year studies, i.e., those conducted under the current standardized protocols and reported after the National Cancer Institute's Carcinogenesis Bioassay Program was transferred to the NIEHS. For the initial studies only those chemicals with positive findings in both species and usually both sexes were categorized as positive carcinogenicity studies and only those chemicals with negative results in both sexes of both species were considered negative carcinogenicity studies. Chemicals were considered mutagenic if there was one or more positive Salmonella assay response and nonmutagenic if there were only negative responses. Scientists from both government and the private sector were solicited for comments on the U.S. NTP plan to evaluate the potential of the transgenic mouse models and the set of chemicals selected for these initial studies.
Specific chemicals chosen for the initial U.S. NTP transgenic mouse studies and the rationale for their selection are listed in Table 2 together with the results from other U.S. NTP studies. 2-year + In addition, N-methylolacrylamide is well absorbed topically and administration by this method and an oral route will provide comparative information on systemic effects. (Target organs: Harderian gland, liver, lung, and ovary; nerve damage in short-term studies.) Isomers 2,4-Diaminotoluene SA + To test the ability of the transgenic mouse models to discern differences between closely related chemicals with apparent 2-year + differences in carcinogenic potential in the rodent bioassay. Also 2,4-diaminotoluene would provide an additional test on genotoxic chemicals that have caused a less marked tumor 2,6-Diaminotoluene SA + response than the more potent carcinogens in the 2-year bioassay. (Target organs: 2,4-diaminotoluene, liver and mammary 2-yeargland; 2,6-diaminotoluene, pancreas and liver.) Model comparison p-Anisidine SA + To obtain complementary information on a uniform set of chemicals in the transgenic models. This agent has been studied 2-year + in p53def but not the Tg.AC. It will also give information on comparative target organ specificity. (Target organ: preputial gland.) Noncarcinogens To examine hypothesis that rodent noncarcinogens will also be negative in the transgenic species regardless of mutagenic The U.S. NTP studies described are currently under way. Preliminary data from these and additional studies completed by NIEHS scientists (10,11) are encouraging and have stimulated the U.S. NTP to plan future transgenic model studies.

Future Studies
Because these studies require fewer animals and shorter exposure times, the use of transgenic models to identify carcinogens has generated interest among government and industry scientists. The U.S. Food and Drug Administration has decided to accept data from transgenic models in lieu of traditional mouse bioassay data for new drug registrations. Their decision is in keeping with the mission objectives of the International Congress on Harmonization to examine whether the need for long-term studies in two species could be reduced (15). The U.S. NTP is participating in efforts by the pharmaceutical industry and other federal agencies to examine the response of four transgenic in vivo models systems including the Tg.AC and the heterozygous p53def mice.