Hazard identification: efficiency of short-term tests in identifying germ cell mutagens and putative nongenotoxic carcinogens.

For more than a decade, mutagenicity tests have had a clearly defined role in the identification of potential human mutagens and an ancillary role in the identification of potential human carcinogens. The efficiency of short-term tests in identifying germ cell mutagens has been examined using a combined data set derived from the U.S. Environmental Protection Agency/International Agency for Research on Cancer Genetic Activity Profile (EPA/IARC GAP) and EPA Gene-Tox databases. Our review of these data indicates adequate sensitivity of batteries of in vitro short-term mutagenicity tests in identifying germ cell mutagens. The analysis also supports the inclusion of an in vivo assay as suggested in proposed regulatory testing guidelines. In the context of carcinogenicity testing, the ability of short-term bioassays to detect genotoxic or mutagenic carcinogens is well established. Such tests are not considered to be as sensitive to nongenotoxic or nonmutagenic carcinogens. However, analyses presented in this report using the EPA/IARC GAP database demonstrate that many putative nongenotoxic carcinogens that have been adequately tested in short-term genetic bioassays induce gene or chromosomal mutation or aneuploidy. Further investigation should reveal whether the mutagenicity of these agents plays an important mechanistic role in their carcinogenicity.


Introduction
In regulatory practice, mutagenicity tests have a clearly defined, primary role in the identification of potential human mutagens and an ancillary role, when considered with other relevant toxicological information, in the identification of potential human carcinogens. Certain assumptions have been made in the deployment of these tests for the detection of germ cell mutagens. For example, all substances capable of mutating mammalian germ cells are viewed as being able to mutate somatic cells in vivo. It is also assumed by many that the great majority of (if not all) substances capable of mutating somatic or germ cells of mammals can be detected with in vitro tests. Two in vitro tests (one for bacterial mutagenicity and one for chromosome breakage in cultured mammalian cells) are assumed to be sufficient to detect a majority of mutagens. However, the alternative of an additional test (induction of gene mutations in cultured mammalian cells) is considered by some to be preferable, either as a complement to the twotest battery or in lieu of in vitro cytogenetics when carried out using a protocol that permits evaluation of clastogenicity.
Because of the purported sensitivity of in vitro systems and the desire to reduce the requirement for animal testing, tests involving mammalian germ cells are not used for screening purposes. It has generally been regarded as prudent, however, to require at least one in vivo mammalian mutagenicity test for substances with relatively high human exposure, such as food additives and drugs. The detection of chromosome breakage (aberrations or micronuclei) in rodent bone marrow has been the test of choice.
ing the ability of short-term tests to detect such agents. The results of our analysis indicate an acceptable level of sensitivity for the detection of germ cell mutagens and suggest some ways in which the tests might be deployed in the future.
How Well Do Short-Term Tests Detect Carcinogens?
As mentioned above, an ancillary role of short-term tests and one that has provoked much controversy (1) is the detection of potential human carcinogens. The utility of short-term tests in the detection of the so-called genotoxic carcinogens has not been seriously questioned. However, the earlier tauted overall correlations between mutagenicity in short-term tests and carcinogenicity in rodents have deteriorated since the 1970s. The apparent failure of short-term tests in the detection of carcinogens may be attributed in part to the identification of more and more putative nongenotoxic carcinogens. These agents often induce site-selective carcinogenicity in rodents when administered at or close to the maximum tolerated doses (2). Although short-term tests are not considered to be sensitive to such agents, the overall database for the putative nongenotoxic carcinogens has not been carefully examined.
From a mechanistic perspective and in the context of risk assessment, it is of obvious importance to be able to distinguish between "genotoxic" and "nongenotoxic" carcinogens. In recent years, much attention has focused on the detection and definition of the latter group of agents. Ashby and Tennant (3) recognized that most nongenotoxic carcinogens are not detected in the Ames test and are devoid of alerting (DNA reactive) structural features. They used the combination of positive results in the Ames test and the presence of structural alerts as an index of genotoxicity. Butterworth (4) broadened the definition of a genotoxic agent as "one for which a primary biological activity of the chemical or a metabolite is alteration of the information encoded in the DNA ... point mutations, insertions, deletions or changes in chromosome structure or number." Conversely, Butterworth (4) defined nongenotoxic chemicals as "those that lack genotoxicity as the primary biological activity. While these agents may yield genotoxic events as a secondary result of other induced toxicity, such as forced cellular growth, their primary action does not involve reactivity with the DNA." In this report we carefully examine a group of putative nongenotoxic carcinogens and classify them on the basis of their mutagenicity per se (ability to induce alterations in DNA structure or content). We have defined criteria for evidence of mutagenicity (and nonmutagenicity) and have applied these criteria to 66 agents that have been cited as nongenotoxic in the pablished literature (3)(4)(5)(6)(7). Three end points (gene mutation, chromosomal aberration, and aneuploidy) were used to evaluate the mutagenicity of these agents. The data clearly demonstrate that many of the putative nongenotoxic carcinogens that have been adequately tested in short-term bioassays induce gene or chromosomal mutations or aneuploidy.

Methods
In approaching the first analysis (detection of germ cell mutagens), we have considered those substances for which peer-reviewed evidence of germ cell mutagenicity is available. The primary data source was the U.S. Environmental Protection Agency/International Agency for Research on Cancer Genetic Activity Profile (EPA/IARC GAP) computer program (available from M.D.W.), which was supplemented with data from the EPA Gene-Tox database (now available on TOXNET). These peer-reviewed databases (8,9) were constructed using the available published literature and contain a preponderance of positive data.
For our purposes we have assumed that the mutagenicity test results were being used predictively to screen for germ cell mutagens and we therefore present an analysis of the sensitivity of these tests in detecting germ cell mutagens. We have not been able to perform an analysis on the specificity of short-term mutagenicity tests, i.e., their performance with substances that are established as nonmutagenic to mammalian somatic cells or germ cells in vivo.
The use of short-term tests in mutagenicity testing is not to define a germ cell mutagen but rather to indicate the potential to cause germ cell mutations. The agent must be shown to reach the germinal tissues to satisfy the presumption that it is a germ cell mutagen. Most published germ cell mutation data have been derived from the evaluation of substances known to be carcinogenic or mutagenic in vitro or to be structurally related to such substances.
The data used for the second analysis (detection of putative nongenotoxic carcinogens) are from the GAP database only. A list was compiled of 66 agents cited in the published literature (3-7) as putative nongenotoxic carcinogens. Fifty-three of these agents were present in the GAP database, and 39 of these were represented by five or more test results (Salmonella tester strains considered as individual tests). A table was constructed for the 39 agents emphasizing mutagenicity, i.e., gene mutation, chromosomal aberration, and aneuploidy. Each test result was verified by carefully reviewing the original publication from which data had been extracted. A summary tabular matrix was then prepared based on the strength of the positive data.
The following criteria were applied to the results in assessing the evidence of mutagenicity (or nonmutagenicity) in vivo or in vitro: a) Evidence of mutagenicity in vivo is provided when an agent exhibits activity in at least two tests for one or more of the following end points: gene mutation, chromosomal aberration, or aneuploidy and positive results are obtained in at least one mammalian study in vivo; b) Evidence of mutagenicity in vitro is provided when an agent exhibits activity in at least two tests in vitro for one or more of the above end points. c) Evidence for nonmutagenicity in vitro is provided when an agent demonstrates negative results in tests representing all three end points above and includes one mammalian test in vitro for gene mutation and one for chromosomal aberration. d) Evidence for nonmutagenicity in vivo is provided when an agent demonstrates negative results in mutagen. There are obvious limitations to this standard. tests representing all three end points above and includes In many cases only one test has been used, and confirmaa minimum of two mammalian studies in vivo. tory testing is warranted. Moreover, in some instances reviewers may have been overly generous in their accep-Results and Discussion tance of the data. All mammalian germ cell mutagens for which data from relevant in vitro tests were available are Efficiency of Short-Term Tests in shown in Tables 1 and 2. Although protocol adequacy and Identifying Germ Cell Mutagens the quality ofthe data are major considerations for regulatory purposes, this analysis deals only with the qualitative The germ cell assays that comprise the standard to test results. These data have been peer reviewed and are which other tests are compared are shown in Tables 1 and accepted for the purposes of our present analysis. In a few 2. For analytical purposes, a positive result in any of these cases, new data have been added based on the recent tests is considered to define a mammalian germ cell published literature.  (10) and nitr ogen mustard (11). "Acrylamide results are updated from the riecent literature (12)(13)(14).
As shown in Figure 1, results from the combined GAP and Gene-Tox databases are as follows: Of 56 germ cell mutagens, 52 had been subjected to bacterial mutagenicity and/or in vitro mammalian cell chromosome breakage assays, and of these 87% gave a positive result. Evaluation of those substances for which both tests were done showed that 32 out of 34 mutagens were detected. The two substances not identified as mutagens by this battery (diazepam and norethisterone acetate) depended on a single mammalian germ cell test (dominant lethal) for their inclusion in the germ cell mutagen list. These agents are not well studied but, like ethanol (negative in a wide range of mutagenicity assays), have been shown to cause malsegregation of chromosomes (aneuploidy), which could account for the dominant lethal results.
If a test for mammalian cell mutagenesis is added to the above two tests, the following results are obtained. Of the germ cell mutagens tested in one or more of the three tests, 87% were identified as mutagens, i.e., three additional positives were identified. Analysis of substances for which all 3 tests were performed showed all 26 were identified as mutagens. Although these figures are marginally higher than those for the two-test battery, the difference reflects the more restricted number of compounds tested.
Because one of the most widely used gene mutation assays (selecting for thymidine kinase deficiency in L5178Y mouse lymphoma cells) also responds to chromosome deletions formed by chromosome breaking agents, there is in principle no reason why a mammalian gene mutation assay such as this should not be used to replace the in vitro cytogenetics and mammalian cell mutagenesis assays described above. This combination, i.e., bacterial mutation test and mammalian cell gene mutation assay (with a protocol for assessment of chromosome breakage), demonstrates the following results. As shown in Figure 1, of 52 substances tested in either one or both of tests just described, 42 were positive. Of the 34 that were tested in both tests, 30 were identified as mutagens. These figures are similar to those for the previous two test batteries, which suggests (albeit based on a relatively small data set) that the use of the cheaper mammalian cell gene mutation assay as an alternative to the assay for mammalian cell chromosome breakage deserves consideration. The four germ cell mutagens missed were DDT, sodium saccharin, ethanol, and tris(2-chloroethyl) phosphate (Fig. 1).
Given that a clear positive result in any of the tests mentioned above is normally enough to warrant a requirement for an in vivo assay, is it necessary to do both in vitro assays if the first one done is positive and, if not, which assay should be done first? As shown in Figure 1, there is no doubt that the mammalian cell chromosome breakage assay identifies more accurately the germ cell mutagens (93%) than does the bacterial mutation assay (75%). [The mammalian cell gene mutation assay is also high (86%)]. The mammalian cell chromosome breakage test, although the more sensitive, is also more expensive. A case could be made for performing the bacterial mutagenicity test first, and only performing a mammalian cell chromosome breakage assay if a negative result is obtained.
In view of the fact that there are bacteria in the gut of mammals and that the liver may be able to carry out metabolic transformations that cultured cells cannot, even with S-9 microsomal fraction, it seems prudent at the present time to subject bacterial mutagens to in vivo testing. This would normally involve looking for the induction of chromosome breakage in cells of the rodent bone marrow, either scoring micronuclei or broken chromosomes at metaphase. The rationale for this is that not all substances that are mutagenic in vitro are able to express significant mutagenic activity in the whole animal. Bone marrow assays are preferred because they are widely used and reliable and because it is reasoned that a mutagen capable of reaching the gonads ought also to reach the bone marrow. We may therefore ask whether the data show that germ cell mutagens are a subset ofbone marrow mutagens.
As shown in Figure 1, of36 germ cell mutagens tested in a bone marrow assay (either aberrations or micronuclei), IN VITRO 33 were identified as mutagens. The 3 not identified were ethanol, sodium saccharin, and triflupromazine. Adler and Ashby (15) found the evidence for germ cell mutagenicity of sodium saccharin and triflupromazine questionable. Sodium saccharin is usually included in lists of reference nongenotoxic carcinogens. Ethanol (also reviewed by these authors) may act by inducing aneuploidy, which the bone marrow assays were not designed to detect. Obe and Anderson (16) state that ethanol shows strain dependent differences in the bone marrow assays. Overall, the data are consistent with the view that germ cell mutagens are probably not unique and that some of the bone marrow assays in the database were not performed using current protocols and therefore were not of adequate power (17). It should be noted that the bone marrow assay detects not only substances shown to cause chromosomal damage in rodent germ cells, but also those known to induce gene mutations in the specific locus coat color test. This reinforces the generalization that specific gene (as distinct from chromosomal) mutagens are rare. Thus, there is some security in using an in vivo chromosome breakage assay for presumptive germ cell mutagens. Adler and Ashby (15) concluded that "the general observation that rodent germ cell mutagens are also genotoxic in somatic cells in vivo remains valid." We would only add the comment that the problem of strain variability among rodents contributes to the discrepancies with triflupromazine and ethanol and is an underlying problem in all of toxicology.

Evidence of Mutagenicity of Putative Nongenotoxic Carcinogens
Using the criteria for evidence of mutagenicity presented in Methods, we were not successful in segregating the putative nongenotoxic carcinogens into only four categories. Thus, the word "limited" was added to identify those agents that met most of the stated criteria. Figures  2 and 3 present a summary of the short-term test results for the 39 agents that could be grouped according to the resulting six categories of evidence. The qualitative data for these agents are presented according to the three genetic end points; gene mutation, chromosomal aberration, and aneuploidy. A miscellaneous category is included for data on cell transformation and sister-chromatid exchange. Test systems are identified by three-letter code words (see appendix for definitions) and in vitro tests are separated from in vivo tests for each end point. In general, the agents investigated fall into two groups: 23 agents showing evidence of mutagenicity (9 of  aniline, chlorodibromomethane, and carbon tetrachloride (3,18).
Evidence of Mutagenicity. The positive data for the seven agents that display evidence of mutagenicity in vivo (Fig. 2) represent primarily chromosomal effects (chromosomal aberrations and/or micronuclei in mouse bone marrow) or, in the case of diethylstilbesterol (DES) and NTA, induction of aneuploidy. Although induction of aneuploidy does not necessarily indicate that a chemical has interacted directly with DNA, clearly, a structural genetic alteration (alteration in DNA content) has resulted from chemical treatment. Results of cell transformation and sister chromatid exchange assays generally strengthen the case for classifying these compounds as a group.
As with the previous category, the seven agents displaying evidence of mutagenicity in vitro exhibit a large body of positive data for chromosomal mutation; all except asbestos appear to be eukaryotic gene mutagens as well.  for two or more end points. Agents in this category should not be considered to pose a serious mutagenic hazard.
Evidence of Nonmutagenicity Only 2 agents, polychlorinated biphenyls (PCB) and (4-chloro-2-methylphenoxy)acetic acid (MCPA) (shown at the bottom of Fig.  3), clearly conform to the criteria for evidence of nonmutagenicity in vivo. Negative test results representing all three end points are reported for these 2 compounds including two or more in vivo mammalian studies for chromosomal aberrations. The remaining compounds are listed under the categories of limited evidence for nonmutagenicity in vitro (6 agents) and limited evidence of nonmutagenicity in vivo (8 agents). The limitations reflect inadequate testing for the three genetic end points. The agents showing evidence of nonmutagenicity are not as thoroughly tested as are those agents with evidence of mutagenicity. Nonetheless, these 16 compounds may be considered as nonmutagenic carcinogens based on the data available in the EPA/JARC GAP database.
A summary of NTP and IARC evaluations of animal and human carcinogenicity classifications for the 39 putative nongenotoxic carcinogens presented in this paper is given in Figure 4. The NTP evaluations are presented by species and sex together with the number of the NTP Technical Report from which the information was obtained. The IARC evaluation represents animal and human studies as well as an overall evaluation of the carcinogenic potential for humans. Possible mechanisms of carcinogenesis are also indicated (refer to Conclusions).
Based on IARC's evaluation of 22 of the remaining 27 agents for which there was insufficient mutagenicity data for inclusion in this analysis, 17 are "not classifiable as to carcinogenicity in humans" and the remainder are considered as "possibly carcinogenic to humans." Limited evidence of mutagenicity and carcinogenicity for these agents suggests that it may be inappropriate to classify them as nongenotoxic carcinogens at present.
Overall, when comparing agents with evidence of mutagenicity to those showing evidence of nonmutagenicity (Fig. 4), it is difficult to find any remarkable distinctions between the two groups based on the weight-of-evidence for carcinogenic:ty. Closer examination of the NTP carcinogenicity data yields similar results. Figure 5 shows a graphic representation of the data organized by tumor site, species, and sex. The first column indicates evidence of mutagenicity or nonmutagenicity. Again, no clear relationship was apparent between the carcinogenicity of these agents in any site and their mutagenicity in shortterm tests. The data do show, however, that 7 of 11 mouse hepatocarcinogens tested are nonmutagenic, and this observation warrants further investigation.

Conclusions
Dearfield et al. (19) have discussed recently the testing approaches used by the EPA Office of Pesticide Programs (OPP) and Office of Pollution Prevention and Toxics   (OPPT). The scheme outlined in Figure 6 represents the proposed OPP mutagenicity test guideline. The first two components in the guideline essentially represent the three-test battery previously discussed (Fig. 1). The in vivo tests in the OPP screening battery include the same bone marrow assays (aberrations or micronuclei) mentioned above. Under the OPPT testing scheme shown in Figure 7, the initial battery is essentially the same as that of the OPP as discussed above. Most chemicals would require no further testing if found negative in all three of the tests in the first tier. However, subsequent germ cell tests or cancer bioassays may be warranted based on other information such as human exposure, compelling chemical structureactivity relationships, or other factors of concern. Positive test results in the initial tier would trigger further testing in the second tier; additional positive results in the second tier tests would trigger the appropriate germ cell tests. The results of the present analysis lend support to the regulatory strategy for mutagenicity testing that has been outlined by the EPA. There are additional implications in terms of carcinogenicity testing as discussed below.
In any sizeable series of chemicals subjected to mutagenicity testing, two groups of agents will be found that are mutagenic in vitro but that are not mutagenic to mammalian germ cells. The first group are those that produce chromosomal damage in the rodent bone marrow. These (or their active metabolites) may not penetrate to the gonads but must be presumed to have the potential to produce chromosomal damage in somatic tissues in vivo. The regulatory position is that, in the absence of evidence to the contrary, it is prudent to assume that any chemical capable of causing mutations in vivo in mammals is a potential carcinogen. In the light of this view, it was in the past rare to proceed to a mammalian germ cell assay because the available mutagenicity information was generally considered sufficient for regulation based on carcinogenic potential. The new regulations mentioned above will require further eValuation of mutagenic potential per se.
The second group of compounds is more difficult to deal with. They are those that are positive in vitro but that do not produce detectable chromosome damage in the bone marrow. This group will include some that are not absorbed, are not activated in vivo, or are immediately inactivated or detoxified. These may be presumed not to present a mutagenic or genotoxic hazard in animals or in man. Others, however, will be active, but their active species do not reach the bone marrow. The gut and the liver are clearly potential targets for such agents. Most of the evidence for such "local" activity pertains to directacting alkylating agents. Ashby (20) has pointed out how unwise it is to neglect these agents because their genotoxicity is not systemic. The typical regulatory position is that further work is needed to provide reassurance that there is no mutagenic or genotoxic action in the gut or liver. The nature of such work is determined on a case-by-case basis. The techniques currently available often leave much to be desired in terms of reproducibility, and many have not been adequately validated. Some of the most difficult problems in genetic toxicology occur in this area, and it is one in which high hopes are held for the future application of recently developed assays with transgenic mice.
Another problem area in genetic toxicology is that of interpreting short-term tests results for putative nongenotoxic carcinogens with regard to hazard identification and risk assessment. It is difficult to ignore the positive mutagenicity test results reported for many of the agents discussed in this report, even in light of evidence that the carcinogenicity of some of them may be mediated primarily through nongenotoxic mechanisms (4). Unless an agent has been tested for each of the three broad categories of DNA alterations, i.e., gene mutation, chromosomal aberration and aneuploidy, it is inappropriate to conclude that the compound is nonmutagenic. Similarly, the strength of the evidence for carcinogenicity must also be carefully evaluated. For example, malonaldehyde is obviously mutagenic in vitro, but the evidence of carcino-genicity is weak. Thus, IARC lists malonaldehyde as "not classifiable as to its carcinogenicity in humans," and the NTP data show a low incidence of tumor formation in the rat thymus gland (Fig. 4). Ashby and Morrod (2) note that many nongenotoxic agents induce selective carcinogenicity in rodents after treatments at or close to the maximum-tolerated doses and that the activity of these compounds in humans cannot always be assumed. Sodium saccharin induces bladder tumors in rats administered 5% saccharin in the diet (21), a dose much higher than the levels present in normal human diets. The evaluation of human risks associated with exposures to such compounds must be performed on a chemical-by-chemical basis considering not only the strengths and weaknesses of the mutagenicity data but of the carcinogenicity data as well.
More research is needed on the mechanisms of action of the putative nongenotoxic carcinogens to ensure proper evaluation of these agents in the context of quantitative risk assessment. Many biological activities may or may not be relevant to the ultimate mechanisms of carcinogenicity. For example, a number of putative nongenotoxic rodent hepatocarcinogens have been shown to cause peroxisome proliferation (7). Although peroxisome proliferation has been demonstrated in most rodent species tested, nonhuman primates are only weakly responsive, suggesting that humans are at limited risk of carcinogenic effects due to this phenomenon (22)(23)(24)(25). Other compounds, such as carbon tetrachloride, chloroform, tetrachloroethylene, pentachloroethane, and sodium saccharin, appear to produce tissue damage by cytotoxicity resulting in regenerative hyperplasia (26). The specific mechanisms involved or the conditions under which these agents produce their effects, as mentioned earlier in the case of sodium saccharin, may have little if any relationship to human cancer risk.
In summary, the efficiency of current procedures for screening mutagens has been examined using a combined data set derived from the EPA/IARC GAP and Gene-Tox databases. Our review of these data indicates that the sensitivity of the batteries of in vitro short-term mutagenicity tests currently used to identify potential germ cell mutagens is approximately 90%. Some possible improvements in the deployment of such tests, such as the inclusion of a mammalian assay for induction of aneuploidy, are worthy of further exploration. However, the data affirm the utility of short-term tests in hazard identification and support current regulatory testing strategies.
In the context of carcinogenicity testing, the ability of short-term tests to detect genotoxic or mutagenic carcinogens is well established. Such tests are not viewed as sensitive to nongenotoxic or nonmutagenic carcinogens.
Analyses presented in this report (Fig. 2) using the EPA/ IARC GAP database, however, demonstrate that many putative nongenotoxic carcinogens that have been adequately tested in short-term genetic bioassays induce gene or chromosomal mutation or aneuploidy. The ability of compounds to induce such alterations in DNA structure or content cannot be ignored as work continues toward understanding the mechanisms of carcinogenesis and potential cancer risks for human populations. Further investigation of the agents listed in Figure 2 should reveal whether their mutagenicity plays an important mechanistic role in their carcinogenicity.