Predictions for the Outcome of Rodent Carcinogenicity Bioassays : Identification of Trans-species Carcinogens and Noncarcinogens

Thirty chemicals or substances currently undergoing long-term carcinogenicity bioassays in rodents have been used in a project to further evaluate methods and information that may have the capability of predicting potential carcinogens. In our predictions the principal information used includes structural alerts and in vitro test results for Salmonella mutagenicity, relative subchronic toxicity, and the sites and types of pathology found in subchronic (90-day) studies. This group of chemicals differs significantly from those used previously to evaluate predictive methods in that 23 of 30 are defined as nonmutagenic by conventional criteria. The goal of this predictive effort is


Introduction
An effort was initiated in 1990 to utilize ongoing rodent carcinogenicity bioassays to test the capability of using both chemical and biological parameters to predict potential carcinogens (1). It was anticipated that this would be an iterative process in which insights and knowledge gained in the first phase would be utilized to improve subsequent prediction capability. In the first phase of the project 44 chemicals undergoing rodent bioassays by the U.S. National Toxicology Program (NTP) were utilized as the target and our predictions were published to encourage the application of predictive methodologies by other investigators. Seven other prediction efforts for these chemicals were also published (2). We believe that this initial effort successfully demonstrated that the parameters of chemical structure, genotoxicity, and rodent toxicity can be used to predict potential carcinogenicity (3). Further, among the various predictive methodologies reported, the highest degree of concordance was achieved for chemicals showing the clearest carcinogenic and noncarcinogenic effects. The most problematic chemicals were those showing either sex-and species-specific or weak carcinogenic effects and those producing equivocal results.
The same biological and chemical parameters cited above and used in our previous effort (1) have been used in this second predictive exercise that involves 30 chemicals currently undergoing rodent bioassay. The principal difference in our approach to predicting the activity of this second group of chemicals is that we are demarcating more precisely the carcinogenic activity that we are attempting to predict. Extensive analysis of the results of bioassays over the past few years has provided insights into the nature of the carcinogenic responses induced in F344 rats and B6C3F1 mice, the two rodent strain/species (4,5) predominantly used in the NTP 2-year bioassay. We have proposed a stratification of bioassay results that is indicative of the relative potency of carcinogens (6). The category that reflects the greatest carcinogenic potential is that of the trans-species carcinogens that induce tumors at one or more sites in both rats and mice; a particular subcategory is represented by those that induced tumors at the same sites or of the same histogenic type in both species. The least carcinogenic potential is represented by the chemicals that induce tumors at only one site in only one of the four sex/ species groups. This stratification is based on the recognition that inbred rodents possess an allelic distribution that is uncharacteristic of the type and frequency found in feral or outbred populations. That is, various alleles of polymorphic genes are represented in outbred populations with variable frequency (7). As a consequence of selected or random inbreeding, the rodent lines lose many of the polymorphic alleles and subsequently possess a more limited number of specific alleles, which are uniformly distributed in the inbred animals. Two consequences of this allelic enrichment are that the various inbred strains demonstrate specific patterns of spontaneous tumors and that they can exhibit strain-specific responses to chemicals (6)(7)(8).
Both of these consequences may be in part responsible for much of the controversy surrounding the interpretation of bioassay results and the appropriateness of their extrapolation for human risk. For example, chemicals that are identified as carcinogens based solely on increases in tumors that occur spontaneously at a high frequency (e.g., >10%), may only modulate the expression of the gene(s) that determine the disease. Since these tumors are a genetic disease of the specific rodent strain, the chemical effects may be of little consequence in other species or even other strains of the same species. The action of the trans-species carcinogens are independent of strain-specific influences and may therefore pose the greatest risk in other species, whereas the risk of the strainspecific carcinogens may affect only certain individuals in an outbred population.
The implications of the stratification of carcinogens for the prediction of carcinogenic potential is that any prediction schema must be risk-averse. That is, it must Environmental Health Perspectives * Vol 104, Supplement 5 * October 1996 .M be possible to reliably predict trans-species carcinogens. Conversely, since the genetic and biochemical basis of strain-specific effects induced by carcinogens is not generally known, it is unlikely that they can be reliably predicted and this category of carcinogens should not be a principal basis for judging how effective prediction methods may be. A corollary to this is that the methods should also be able to reliably predict which chemicals are unlikely to induce tumors in either bioassay species.

Results
As in the previous exercise (1), our predictions reflect a weight-of-evidence call based on structural alerts and Salmonella mutagenicity data, the subchronic bioassay toxicity data, and the maximum tolerated dose (MTD) levels set for the two-year bioassay. The goal of our effort to predict the carcinogens among the 30 current chemicals is to identify all of those chemicals with trans-species carcinogenic potential and those that are noncarcinogens. In our view the data available for predicting the other chemicals cannot be reliably used to infer which may induce strain-specific tumorigenic effects; therefore all of our predictions of carcinogenicity reflect a primary concern for trans-species effects. Some data may be an alert for potential strain-specific effects, such as subchronic target-organ toxicity for tissues in which spontaneous tumors are frequent; but we have not been able to identify any reliable parameters. Likewise, we have not identified parameters of toxicity for chemicals that may induce single site/single sex/species tumors that are not associated with spontaneous tumors. This uncertainty will result in such chemicals being identified as probable noncarcinogens.
Some of the aspects of subchronic toxicity that have been identified in Table 1 have been associated with the observed strainor species-specific responses seen in bioassays of other chemicals. They have been highlighted in the table as a possible alert to such specific carcinogenic effects even though they have no reliable predictive value. For example, hyalinelike droplets are often deposited in rat kidney tubular epithelium as a consequence of chronic toxicity and also have been associated with "2p-globulin related kidney tumors in male rats (9)(10)(11).
In this exercise we have identified 3 of the 30 chemicals as probable (trans-species) carcinogens, of which gallium arsenide is a nonmutagen. Ten of the 30 are predicted to be noncarcinogens. We have abstained from predicting for 2 chemicals because the outcome of the bioassay is known. This group of 30 chemicals poses a particular challenge, compared to the first exercise (1), since 23 of them are not mutagenic in the Salmonella assay. Though a few parameters related to the carcinogenicity of nonmutagens are known (e.g., peroxisome proliferation (11)(12)(13)(14)(15) and a2p-globulin hyaline droplet induced nephropathy, etc. (9-11) none are clearly represented among this group. This is particularly important since many of the chemicals that induce strain-specific carcinogenic effects are not mutagens. Consequently, 15 of the chemicals have been classified as being of "uncertain predictability" to indicate that it is possible that strain-or species-specific neoplasia could be a consequence of chronic exposure. The above caveats may seem like a rationale for anticipated missed predictions. That is not the case because we are categorical in our efforts to predict transspecies carcinogens and noncarcinogens and the success or failure of our predictions should be judged critically on that basis.

Discussion
In evaluating the potential carcinogenicity of these chemicals we have again placed primary emphasis on structural alerts, mutagenicity for Salmonella, relative toxicity (reflected in the maximum dose tested) and the sites and type of subchronic pathology. These properties are described in Table 1, together with brief comments or explanations of the basis for our prediction. The large number of chemicals deduced to be of uncertain predictability reflects a different distribution of chemicals than those for which predictions were offered in 1990 (1). It is not anticipated that any of this group of 15 chemicals will induce trans-species carcinogenesis, but some will be associated with equivocal effects or the induction of site or species-specific carcinogenesis. A majority are anticipated to be noncarcinogens but some aspect of the subchronic toxicologic evaluation of the chemicals raises the possibility that they could be associated with siteor speciesspecific neoplastic effects.
It is unlikely that speciesand strainspecific carcinogenicity will be predictable until the genetic basis of these occurrences is better understood. The rapid evolution in mouse genetics, in the utilization of transgenic and knockout technology to study chemical-gene interactions, and in sequencing the mouse genome to improve the identification and cloning of genes, holds the promise that specific genes and mechanisms underlying carcinogen susceptibility can be understood. We believe that at least two discrete mechanisms are involved: interaction of some chemicals with genes or gene products that modify the spontaneous tumor incidence of inbred rodents, and interactions with genes that influence tissue-specific susceptibility not associated with spontaneous tumor sites. These are potentially important issues in utilizing rodent bioassay for protecting public health and the environment. The speciesor strain-specific carcinogens are unlikely to pose a public health problem that is as important as that ascribed to trans-species carcinogens. However, the genes that influence or govern species/ strain specificity may be useful in identifying susceptible individual humans. To know if this is the case, it is necessary to clone such rodent susceptibility genes to determine if humans possess structural analogs. However, since cancers are the consequence of multiple genetic effects, the predominant influence of the susceptibility genes could be principally the consequence of extensive inbreeding.
We have proposed recently that specific transgenic mouse lines can be utilized in short-term carcinogenesis bioassays (8). The results available thus far suggest that the transgenic bioassays offer advantages not only in time and cost savings and a reduction in the number of animals needed for bioassays, but that they also minimize species-/strain-specific influences.
Environmental Health Perspectives * Vol 104, Supplement 5 * October 1996 Carcinogen; mutagen: chronic inflammation in lungs, bronchiolar epithelium regeneration; a NOEL in the lungs of rats or mice was not found. Carcinogen; mutagen: hyalinelike droplets in MR kidney; absolute liver wt increase at every dose; periportal degeneration at two highest subchronic doses; these same doses selected for chronic study. Carcinogen; nonmutagen: chronic inflammation in lungs; inorganic arsenic is a known carcinogen in humans and an enhancer of induced tumors in rats and mice; a NOEL was not reached in the subchronic rat study.
Unlikely carcinogen; nonmutagen: if high dose in the chronic study exceeds 2.5% in the diet, then tumors at a sex/species specific target organ site could occur. Unlikely carcinogen; nonmutagen: hyperplasia and hyperkeratosis of the forestomach epithelium in rats and mice, probably compound related; bone marrow atrophy and hemorrhage in rats; atrophy of ovarian corpus luteum in FM.Toxic lesions occurred at highest subchronic dose, which is 4x highest chronic study dose. (low, low risk; chronic irritation in the forestomach by a nonmutagen does not easily result in tumors). Unlikely carcinogen; nonmutagen: the absence of any histopathologic lesions observed in the 13-week subchronic study suggests that codeine will prove to be a noncarcinogen in the 2-year chronic study; rats exhibited poor weight gain in the subchronic study because of lowered food intake. Unlikely carcinogen; nonmutagen: lung was the only target organ in exposed rats, with alveolar focal accumulations of inflammatory cells; these lesions were accompanied by lymphoid hyperplasia of the bronchial and mediastinal lymph nodes; the rat lung is a possible target for neoplasia.Lack of any subchronic lesions in mice suggests an absence of any neoplastic effects in the chronic study. Unlikely carcinogen; nonmutagen: chronic epithelial hyperplasia of the forestomach in rats and mice induced by a nonmutagen could result in a low incidence of forestomach tumors; a NOEL was NOT reached for FM; a NOEL for hematologic effects in FR was NOT reached; most of the target organ toxicity in both sex/species occurred at highest subchronic (500 pm) dose; a sex/species specific target organ tumor response cannot be ruled out.
Unlikely carcinogen; nonmutagen: minimal hypertrophy of goblet cells lining the nasopharyngeal duct in most caudal nasal sections occurred at all doses; no NOEL observed in rats; 1,3-butadiene and chloroprene, analogs of isoprene, are carcinogens in rats and mice (NTP); high dose of isobutene in the chronic study is the same as the high subchronic dose; possible nasal passage carcinogen in rats. (continued) Possible carcinogen; mutagen: liver hypertrophy and inflammation at all doses in rats and highest doses in mice; kidney nephropathy in MR; hyalinelike droplets in kidneys, MR and FR (thought to be parent compound); urinary bladder effects at all doses in mice. An analog, 2-aminoanthraquinone, is a liver carcinogen in MR, MM, and FM;low to moderate risk. Unlikely carcinogen; nonmutagen: chronic hyperplasia of nasal respiratory epithelium in rats could lead to neoplasia; epidermal hyperplasia of the skin and epithelial hyperplasia of the forestomach in mice are unlikely to lead to neoplasia, though a spontaneous mutation in the cellular Ha-ras gene could lead to skin tumors; doses for chronic study were set at the NOEL or at a dose at which toxicy was minimal. Possible carcinogen; mutagen: doses for chronic study were set at or near NOEL observed in the subchronic study in both rats and mice; low risk. Unlikely carcinogen; nonmutagen: it is unlikely that chronic skin irritation caused by a nonmutagen will result in skin neoplasia; chronic systemic effects in the rat kidney or mouse liver could lead to a specific sex or species neoplastic lesion, but the low doses selected for the chronic study may minimize that possiblity; low risk. Possible carcinogen; nonmutagen: chronic skin inflammation in both rats and mice unlikely to result in tumors; possible, but likely low Unlikely carcinogen; nonmutagen: liver cytomegaly was the major lesion in all mice exposed to the highest subchronic dose (5000 ppm) which was accompanied by 'narcosis"; the next lower dose, 1800 ppm, was selected as the chronic study high dose; if chronic liver cytomegaly and/or necrosis is minimal in mice, liver neoplasia is unlikely. Possible carcinogen; nonmutagen: chronic hyperplasia and inflammation of the bronchiolar and alveolar eptithelium accompanied by fibrosis; low to moderate risk related to spontaneous incidence of lung tumors in rats and mice.
(continued) -/rats; bone marrow and spleen, histopathologic lesions observed in exposed rats; 77-09-8 MM, FM elevated incidence of micronucleated erythrocytes in mice suggests a genotoxic metabolite of the parent compound; cell depletion and necrosis occurred in mouse bone marrow; phenophthalein is unlikely to be a carcinogen unless chronic high dose in mice causes neoplasia indirectly via physiolological imbalance.
"All of the information used in these predictions, e.g., structures and subchronic toxicity, can be obtained through the NIEHS/NTP home pages: NIEHS:http://www.niehs.
nih.gov/exchange/ and NTP: http://ntp-server.niehs.nih.gov/. bStructural alert, SA (5): NA, criteria for determining structural alerts do not apply to inorganic compounds Sal. Salmonella mutagenicity assay performed with and without S9-induced liver microsome preparations. cMaximum dose selected for chronic 2-year bioassay expressed as: %, ppm, mg/kg or ppm/meter(m)3. dMajor target organs identified via histopatholog in the 13-week subchronic study: male rat (MR); female rat (FR); male mouse (MM); female mouse (FM). NOEL = a "no effect level" was not reached in the subchronic study. "Nine chemicals that have undergone peer review prior to submission of this manuscript.