212 Diana Anderson mentioned above and a relevant whole mammal test in germ cells. Other tests may be added if the chemical warrants it, such as cytogenetic analysis of human lymphocytes of exposed workers, or any others that may be considered relevant. Substances found positive in the first screening, depending on their economic or medical importance, could be subjected to more rigorous testing. According to Matter (19), however, very few compounds (0'1•o) originally produced in pharmaceutical re- search ever have a chance of being marketed. He suggests that extensive early muta- genicity testing of drugs is of little help since benefit/risk decisions cannot be made at that stage, and that such testing should be concentrated on those compounds scheduled for clinical trials or commercial introduction. However, it would seem better to avoid mutagenic products at an early stage where possible. Before a substance is tested, consideration should be given to its chemical structure, to determine if it is related to compounds which already have known mutagenic, carcino- genic, teratogenic or general toxic effects. Such a consideration might give an indication of its mutagenic potential and this may be useful in setting testing priorities. If it is suspected of having high mutagenic potency, then this factor may override the other two parameters which determine exposure. COMMENTS ON SOME OF THE METHODS A great deal has already been published about recommended protocols, e.g. (17, 60). However, some general rules apply for all testing systems. All assays should be run with concurrent negative and positive controls. Positive controls should, where possible, be structurally or mechanistically related to the compound under test. MICROBIAL METHODS Much work in different laboratories has been carried out on the bacterial Salmonella typhimurium mutagenicity system with rat liver microsomal activation, and as such it is well validated both for the detection of some mammalian and human carcinogens and bacterial mutagens. However, it is not well validated for human germ cell mutagens since none are unequivocally known. This does not mean that other microbial systems would not be equally valid for detecting mutagens/carcinogens if as much work had been undertaken with them. The Salmonella typhimurium strains of Ames (3) are able to detect base-pair substitution and frameshift gene mutations. A drawback to the system is that it is not quantatitive and the potency of mutagens cannot be ranked. However, the system is useful for a qualitative answer. The system of Salmonella typhimurium and other systems such as that of the bacterium Escherichia coli (61) and yeasts (62, 58, 59) can provide a quantitative answer if used in liquid cultures in combination with viability studies. Yeasts, however, are used to measure not only gene mutation but more commonly gene conversion and mitotic recombinational events. A very sensitive test for measuring gene mutation is claimed when E. coli is used in a fluctuation assay (63). Generally, compounds in a screening programme should be tested both with and without microsomal activation. Strains to detect both base-pair and frameshift mutations should be used, as well as at least five concentrations of the test compound over a wide dose range in order to maximise the chance of obtaining a response, with the highest dose if possible inhibiting the growth of the microbes. This ensures that the compound has entered the microbes. Replicate experiments should be undertaken to determine a

The current state of mutagenicity testing 213 reproducible response. Simpler 'spot' tests measuring the degree of differential killing in repair-deficient microbial strains by comparison with wild type strains (64, 65) provide an indirect answer to the problem of mutation testing, since they detect damage to DNA but do not measure gene mutation directly. Such tests at the present time are not con- sidered particularly sensitive, however. MAMMALIAN CELL SYSTEMS Cell 'transformation '/mutation assay In this laboratory, concurrent with the Ames' test the 'cell transformation' assay system of Styles is used (8, 66). It is based on the ability of mammalian cells transformed by a carcinogen to form colonies in soft agar and this is only one of the accepted criteria for cell transformation. A recent paper (67) suggests that the ratio between transformation and mutagenesis for ouabain resistance in normal diploid cells is about 20: 1 (after treatment with benzo(a)pyrene and its 7,8-dihydrodiol), which suggests that any one of 20 genes may be involved in transformation as opposed to 1 for mutation. This ratio is also apparently substantiated for normal hamster embryo cells treated with benzo(a)- pyrene. If transformation is, therefore, a genetically based system, it may be useful for preliminary genetic monitoring. We have found it to be equally predictive for carcino- genicity as the Ames' test, and by inference, therefore, it is equally predictive for mutas genicity. Baby hamster kidney (BHK 21/C12) and either human diploid lung fibroblast- (WI 38) or human liver cells (Chang) were treated as described (47). Five doses of com- pound were used with the S-9 mix of the Ames' test. Survival was assessed independently and a transformation frequency calculated. The test is not as rapid as the Ames' test, but in our 'blind' study for 120 compounds with these two tests in combination the tests only missed detecting as carcinogens/mutagens diethylstilboestrol and vinyl chloride. This latter compound was later detected when tested in the gaseous phase. This therefore seems a very promising system for a preliminary mutation screen. The limitations of using these two test systems for a simple screening programme have been discussed elsewhere (8). Lymphoma Cells in Culture Whilst human cells or normal diploid cells are obviously desirable for mutation assays in culture, they are more difficult to handle from a screening viewpoint than malignant cells in culture due to their low plating efficiencies and lack of perpetual proliferation. Lymphoma cells, which grow in suspension, are even easier to handle than cells which grow as a monolayer. They do not require trypsinisation, are very easily subcultured and they are not subject to metabolic co-operation which can cause a loss of mutants. Also they can be used in a host-mediated type assay (68) or in a fluctuation test (69). Lymphoma cells in culture can readily detect direct-acting mutagens and carcinogens (70-78) and may also be used in combination with S-9 mix to detect indirect acting muta- gens and carcinogens (Anderson, unpublished). The lymphoma cell mutation system is manageable in that induced mutation frequencies are readily detectable and vast numbers of plates do not have to be used to detect a spontaneous frequency (e.g. 105 cells per petri dish in soft (0.3 •o) agar and selective medium gives a background frequency of about 2-20 colonies in P388F cells with selective media containing 5-iodo-2-deoxyuridine and excess thymidine respectively). Induced frequencies can increase from 10- to 100-fold above these