Current status of the Gene-Tox Program.

The U.S. Environmental Protection Agency's Gene-Tox Program is a multiphased effort to review and evaluate the existing literature in assay systems available in the field of genetic toxicology. The first phase of the Gene-Tox Program selected assay systems for evaluation, generated expert panel reviews of the data from the scientific literature, and recommended testing protocols for the systems. Phase II established and evaluated the database of chemical genetic toxicity data for its relevance to identifying human health hazards. The ongoing phase III continues reviewing and updating chemical data in selected assay systems. Currently, data exist on over 4000 chemicals in 27 assay systems; two additional assay systems will be included in phase III. The review data are published in the scientific literature and are also publicly available through the National Library of Medicine TOXNET system. The review and analysis components of Gene-Tox comprise 45 published papers, and several others are in preparation. Differences that have been observed between Gene-Tox and National Toxicology Program databases relative to the sensitivity, specificity, accuracy, and predictivity of genetic toxicity data compared to carcinogenesis data are ascribable to differences between the two databases in chemical selection criteria, testing protocols, and chemical class distributions.


Introduction
The U.S. Environmental Protection Agency's Gene-Tox Program is a multiphased effort to review and evaluate the existing literature in genetic toxicology. Phase I of the program was devoted to the selection ofassays to be evaluated and the evaluation of literature by work groups of experts in each area. Phase II was devoted to establishing a database ofchemicals evaluated by each work group and analyzing that database. Phase m (ongoing efforts) is devoted to the continued review of selected assays and updating ofthe database, now publicly available irough the National Library ofMedicine (NLM) TOXNET system. Reports ofall three phases are published in Mutation Research Reviews in Genetic Toxicology (1)(2)(3)(4).
Phase I During phase I of the program, work groups of experts reviewed and evaluated the published literature for 23 selected assays (Table 1) to determine a) the validity of a particular system, b) the chemicals for which it was best suited, c) the proper test protocol, and c) the appropriate techniques of data analysis, interpretation, and presentation.
In addition, each work group was asked to a) evaluate the assay's ability to discriminate between mutagens and nonmutagens and/or carcinogens and noncarcinogens, b) evaluate the system's performance with chemicals of various classes and identify chemicals whose effects were not adequately detected, c) formulate generalized protocols and criteria for data evaluation and validation, d) identify areas requiring additional research or further development and validation, and e) publish an evaluation of the assay in the open literature. Literature for evaluation was provided to the work groups by the Environmental Mutagen Information Center (EMIC), Oak Ridge National Laboratory, Oak Ridge, Tennessee. EMIC selected only that portion of the available literature which met the following criteria: the article was a primary paper published in a peer-reviewed journal, it dealt with chemical mutagenesis, the agent studied was a pure chemical, the article contained quantitative data, and the article was published in English or a language for which EMIC had easy access to a translation.
Articles that met these criteria were given to the work groups, which then evaluated each article for the following elements: proper use of experimental design; use ofpositive and negative controls; proper selection of solvents and vehicles; acceptable spontaneous background mutation frequency or rate; use ofmetabolic activation systems, if necessary; use ofappropriate criteria; for positive, negative, or inconclusive results; and provision of dose-response information. This latter criterion was not considered critical if all others were met. In addition, each work group was free to apply other criteria that might be specific to its particular assay. Agents evaluated were designated as positive (+), positive with dose-response data provided (+D), positive with activation only (+*), negative (-), or evaluated but no definitive call possible (T).
At the end ofphase I, the work groups had published 37 review articles, 36 concerned with assays in genetic toxicology and 1 describing the establishment of the Gene-Tox carcinogen database (2).
been tested in only one system. Those chemicals that had been tested in more than one system were, for the most part, either direct-acting mutagens or those that are known to metabolize to reactive intermediates by liver enzyme systems. This may have made sensitivity of the various systems appear unnaturally high.
The database is skewed to the positive. With the exception of the Salmonella assay there is a paucity of negative data in the database in general and in the carcinogen database in particular, where only 61 of 506 chemicals evaluated had negative results.
Chemicals tested are unevenly distributed across the 30 classes used in the Gene-Tox classification scheme ( Table 3). The most heavily tested classes are class 25, benzene rings; class 30, heterocyclic rings not otherwise classified, unclassified compounds; class 29, alcohols and phenols; class 8, aromatic amines, aliphatic amines, amides, and sulfonamides; and class 2, acyl and aryl halides, halogenated ethers and halohydrins, and saturated and unsaturated alkyl halides. Such a distribution makes an analysis of chemical class specificity of the various assays difficult for all except the Salmonella assay, where a sufficient number of chemicals have been tested in the various classes to permit a determination of system performance according to chemical class.
The phase II analysis resulted in three publications; one dealing with the establishment ofthe database (1), one dealing with the evaluation of mutagenicity assays for the purpose of genetic risk assessment (3), and a third dealing with the developmental status of various assays for genetic toxicology testing (4).

Phase 11
In addition to the published reports, a database of more than 2000 chemicals had been established at EMIC (3). At the outset of the Gene-Tox Program, it was anticipated that this database would be amenable to the type of analysis that would answer a series of fundamental questions about genetic toxicology (Table  2). However, certain characteristics of the database have made such an analysis difficult, if not impossible, to perform.
Chemicals are unevenly distributed across assay systems. For example, of the more than 1000 chemicals in the phase I Salmonella database, approximately 200 had been tested in a cancer bioassay. In comparison, 59 of the approximately 200 chemicals in the mouse lymphoma L5178Y phase I database had been tested in a bioassay.
There is little basis for studies of comparative mutagenesis. In the phase I database, 1559 chemicals, or 59% of the total, had Table 2. Goals of phase II of the Gene-Tox Program. Identify genetic and related end points that are of concern to human health Distinguish those systems that are most ready for extensive use in testing from those that should be regarded as developmental Determine the sensitivity ofeach assay to respond to specific classes of chemicals and identify major strengths and weaknesses Examine the qualitative correlation between mutagenesis and carcinogenesis end points Devise specialized batteries of bioassays that detect with high probability the various types of genetic and related damage induced by various classes of chemicals Consider the potential utility of in Witro mutagenesis and carcinogenesis bioassays for potency estimation Identify information gaps and future research needs and establish a mechanism for evaluating the status of test systems on a continuing basis As part of the ongoing Gene-Tox effort, certain assays from phase I have been selected for update (Table 1). In addition, two assays not evaluated in phase I, the Chinese hamster ovary (CHO) AS52 assay and the mouse biochemical specific locus assay, will be included in the updated database.
Although the update process has been simplified over that used in phase I, the overall objectives of the program and the basic work group structure remain in place. More than 1500 chemicals have been added to the database since the completion ofphase I, bringing to over 4000 the total number of chemicals evaluated. The basic features ofthe phase HI database are the same as those noted above lbr phase I. There is still a paucity ofnegative data; the majority ofthe chemicals evaluated have been tested in only one system and chemical class distribution is essentially unaffected.
The database for the Salmonella assay now totals 2469 chemicals. Ofthese, 1100 are positive, 880 are negative and 489 are classified as T. Of the 1100 chemicals that are positive, 666 are positive without activation, 416 are positive with activation, and 18 are positive without activation and negative with activation (Table 4).
The Gene-Tox and NTP databases are different in several important aspects. Most notably, chemicals in the NTP were selected according to defined criteria and tested according to standard protocols, whereas chemical selection in Gene-Tox is random, and protocols are varied. In the case of the Salmonella assay, however, the most likely reason for the reported differences in sensitivity, specificity, predictive ability, and concordance ofthe assay is probably related to chemical class distribution of the agents tested.
The Gene-Tox chemical classification scheme is based on selected organic functional groups, ring systems, biological origins, and/or organic composition. Carcinogens that have been tested in the Salmonella assay are more apt to be classified as halides, epoxides, sulfur compounds, mustards, xanthenes, nitro and nitroso compounds, nitrosamines, metals, polycyclic aromatic hydrocarbons (PAH), steroids, and benzene rings. False positive results are distributed across the data base in the following pattern: alkyl halides, 12; vinyl/allyl compounds 6; halogenated benzenes, benzeneamines and steroids, 5 each; metals and aromatic azides, 4 each; benzene/phenols and ureas/carbamates, 3 each; amides and hydrazines, 2 each; and miscellaneous, 1. Distribution ofthe first set of73 NTP chemicals (6) across the Gene-Tox chemical classification scheme shows a relatively high number of agents classified as alkyl halides, allyl and vinyl alkenes, benzeneamines, and aromatic azo compounds; the same classes in which Gene-Tox has found a high proportion of false negative responses. If this distribution holds true for the combined set of 114 chemicals, it could account for the lower sensitivity observed by the NTP and accordingly the differences noted in the other parameters.
Noncarcinogens in the Gene-Tox Salmonella database are found primarily in classes 2 (halides), 8 (aromatic amines), 11  (carbamates and ureas), 18 (nitro compounds), 22 (PAH), and 25 (benzene rings). It appears from this analysis that the Salmonella assay can serve as a useful tool for identifying in vivo carcinogens, providing attention is paid to the importance of chemical class when interpreting results.
The Gene-Tox database is now publicly available through the National Library of Medicine's TOXNET system. The 'TOXNET unit record for the Gene-Tox database is shown in Figure 1; a partial record for formaldehyde is shown in Figure 2. Update ofthe database now that it is public will be primarily the responsibility ofthe EPA with the assistance ofEMIC. The update will continue to make use of the peer-review system although in a slightly modified form. Chemicals to be added to the database will be published in a series of short papers in Mutation Research Reviews in Genetic Toxicology; simultaneously with submission ofthe manuscript, the chemicals evaluated for each assay will be added to the publicly available database. At present, the TOX-NET database contains all ofthose chemicals evaluated in phase I and results of the update for the CHO/HGPRT assay and the micronucleus assay.
The EPA wishes to take this opportunity to express its gratitude to those members ofthe genetic toxicology community who have given so generously of their time and talent to contribute to the success of the program. We also thank the staffof EMIC for their unfailing support without which this program would not be possible.