Short-term testing — are we looking at wrong endpoints?

Mutation Research, 205 (1988) 13-24 Elsevier

13

MTR 04401

Short-term testing - - are we l o o k i n g at w r o n g endpoints? Claes Ramel Department of Genetic and Cellular Toxicology, WallenbergLaboratory, University of Stockholm, 106 91 Stockholm (Sweden) (Received 17 March 1987) (Accepted 1 April 1987)

Keywords: Short-term testing; Animal carcinogenicity; Oncogenes; Amplification; Transposition; Hypomethylation; Polygene mutation; Recombinogenic effect.

Summary Short-term testing has been performed and interpreted on the basis of correlation between these tests and animal carcinogenicity. This empirical approach has been the only feasible one, due to a lack of knowledge of the actual genetic endpoints of relevance in carcinogenicity. However, the rapidly growing information on genetic alterations actually involved in carcinogenicity and in particular activation of oncogenes, provides facts of basic importance for the strategy of short-term testing. The presently used sets of short-term tests focus on standard genetic endpoints, mainly point mutations and chromosomal aberrations. Little attention has been paid in that connection to other endpoints, which have been shown or suspected to play an important role in carcinogenicity. These endpoints include gene amplification, transpositions, hypomethylation, polygene mutations and recombinogenic effects. Furthermore, indirect effects, for instance via radical generation and an imbalance of the nucleotide pool, may be of great significance for the carcinogenic and cocarcinogenic effects of many chemicals. Modern genetic and molecular technology has opened entirely new prospects for identifying genetic alterations in tumours and in its turn these prospects should be taken advantage of in order to build up more sophisticated batteries of assays, adapted to the genetic endpoints actually demonstrated to be involved in cancer induction. Development of new assay systems in accordance with the elucidation of genetic alterations in carcinogenicity will probably constitute one of the most important areas in genetic toxicology in the future. From a regulatory point of view the prerequisite for a development in this direction will be a flexibility of the handling of questions concerning short-term testing also at a bureaucratic level.

The increasing evidence that genetic changes constitute a critical event in tumorigenicity has enabled the use of short-term mutagenicity assays also to trace carcinogenic properties of chemicals. Correspondence: Dr. C. Ramel, Department of Genetic and Cellular Toxicology, Wallenberg Laboratory, University of Stockholm, 106 91 Stockholm (Sweden).

Considering the vast number of chemicals and various combinations of chemicals, to which humans are exposed, this approach has in fact been the only feasible one. However, the use of shortterm assays to predict human cancer entails a number of difficulties, causing both false positive and false negative results at least for certain groups of chemicals. There has been much discussion of

0165-1218/88/$03.50 © 1988 Elsevier Science Publishers B.V. (Biomedical Division)

14 how to apply and interpret short-term assays and some voices have even been raised, questioning the predictive value of short-term assays altogether. Although the discovery and analyses of oncogenes in particular have demonstrated the role of genetic alterations in carcinogenicity it must be stressed that tumour formation is a multistep process in all probability also involving epigenetic events. The extrapolation from mutagenicity data in short-term assays to carcinogenicity therefore does not only involve a correlation between these two endpoints, but it can be divided into a series of components, constituting the sequence of events leading to cancer - - usually more or less arbitrarily described under the concepts initiation, promotion and progression. Fig. 1 shows a diagram of some basic components of interest in the present context illustrating this sequence of events from mutagenicity to carcinogenicity. Moving from short-term assays (A) at the left of the diagram in Fig. 1 to carcinogenicity at the right end (E) essentially implies an increasing relevance for cancer prediction, but at the same time a decreasing resolving power and increasing cost and time expenditure. The most relevant measure of carcinogenic potency of environmental agents in man is direct epidemiological observations, but obvious obstacles such as the long latency period and difficulties in identifying sufficiently large populations with a defined and homogeneous exposure in relation to adequate controls, preclude any general use of epidemiology for the prediction of carcinogenicity of chemicals. But apart from these difficulties epidemiology essentially implies the counting of dead bodies, which evidently constitutes a limitation in cancer prevention. At the other extreme the use of short-term assays for the prediction of human carcinogenicity ( A - E ) is simple, inexpensive and sensitive, but it comprises an

A

B

C

IN SHORT-TERM ASSAYS

CHANGESIN CARCINOGENESIS

D

E

AND S E ~ R Y GENETICCHANGES IN CARCINOGENESS

I............ H.......... H.....

I INCREASING RELE~kNCE INCREASINGCOST & TIME DECREASk'~G RESOLVINGPOWER

Fig. 1. Endpoints in testing for carcinogenicity.

extrapolation across the whole series of events in carcinogenesis. It must be emphasized that extrapolation of short-term assays to carcinogenicity in practice implies the comparison of mutagenicity with animal carcinogenicity, while animal carcinogenicity is approximated also to measure human carcinogenicity. As pointed out previously (Ramel, 1986) the correlation between animal and human carcinogenicity is, however, by no means evident. Apart from species differences in susceptibility, etc., animal carcinogenicity normally measures the ability of a test agent to function as a complete carcinogen, which probably differs substantially from the exposure of humans to various complex mixtures of initiating and promoting agents. The prediction of cancer from short-term assays is largely a question of overcoming the intervening and mostly unknown sequence of items between the short-term testing and human carcinogenesis as indicated in Fig. 1. The application of short-term assays for this purpose can either be performed on the basis of empirical observations of the relation between short-term tests and carcinogenicity or else by a mechanistic approach unravelling the effect of the test agents on the various steps leading to malignancy. These two approaches are of course not mutually exclusive, but they imply different conditions, analyses and problems.

Empirical approach Observations of the relation between mutagenicity and carcinogenicity for different assay systems and types of chemicals have constituted the main foundation for the extrapolation and use of short-term assays in predicting carcinogenicity. However, the high correlation between Ames' Salmonella/microsomal assay and animal cancer data reported 10 years ago has not been substantiated for all groups of chemicals. Recent cancer data by the National Toxicology Program in the United States have comprised more randomly selected chemicals than the ones used in previous comparisons with Salmonella and other tests and notable deviations from previously determined correlations between mutagenicity and carcinogenicity have been revealed for certain chemicals and groups of chemicals (Zeiger and Tennant, 1986).

15

Although the introduction of the Ames' SalmoneUa/microsomal assay no doubt has been of decisive importance for various aspects of mutagenicity testing, it is a generally recognized fact that one single assay cannot possibly constitute an adequate reflection of mutagenic alterations occurring in mammalian cells in vivo. Therefore, various batteries of test systems have been recommended. The addition of test systems, however, implies the problem of interpreting conflicting test results. This has usually been done in a subjective and often intuitive way against the background of general experience. It is, however, clear that a more systematic and objective use of the acquired knowledge from short-term testing with different assay systems and different types of chemicals should provide a better basis for a carcinogenicity risk evaluation. In recent years the systematic collection and evaluation of available mutagenicity data, particularly through the GeneTox Program in the United States, have opened up new possibilities of applying modern computer techniques to make a more efficient use of this massive background experience for cancer prediction on the basis of short-term assays. The most advanced computer system specifically designed for this purpose has been worked out by Rosenkranz and Klopman (Rosenkranz et al., 1985). It is in fact composed of two different programs, one dealing with the relation between chemical structure and biological effects and one with the empirical performance of test systems in order to optimize the selection of test batteries.

Mechanistic approach The use and interpretation of short-term assays have so far mainly been performed on the basis of an empirical approach, simply due to our lack of understanding of molecular and biological mechanisms behind tumorigenicity. This situation is, however, changing and in the future an empirical testing approach may be gradually replaced by testing and evaluation procedures based on the actual genetic alterations responsible for the development of cancer. Such an approach can be expected to lead to a more selective and sophisticated use of short-term assays in the future, but it implies an intimate cooperation with basic re-

search concerning biochemical processes in carcinogenesis. Strangely enough up to now the spectacular progress in oncogene and related research has had a remarkably minimal influence on the discussions concerning the application and interpretation of short-term testing for carcinogenicity. It seems to be high time to introduce these aspects into the discussions of short-term testing and this is the purpose of the present paper. Some of the points brought up in this paper have also been discussed at some length in a previous one (Ramel, 1986). In recent years the analyses of viral induction of cancer, of oncogene activation and of hereditary forms of malignancy have revealed a number of points of importance in the present context. It is clear first of all that alterations at the level of DNA are crucial in carcinogenesis. Secondly tumour formation usually implies genetic alterations not only at the initiation stage, but various chromosomal aberrations and other mutational events occur also at later stages. Whether these genetic alterations occur independently of each other, if they are triggered by some kind of common mechanism or if they constitute an interrelated sequence of events, is not known. Thirdly, almost all conceivable types of mutations or DNA alterations seem to be able to be involved in carcinogenesis. Fourthly, observations point to the fact that mutations associated with carcinogenesis may exhibit a frequency far above what could be expected on the basis of established mutation rates. It seems inevitable that this information should be seriously considered when establishing a strategy for short-term testing and for the development of genetic toxicology in general in the future. The rationale behind the use of short-term assays for the prediction of cancer is the correlation between mutagenicity and carcinogenicity, but what this really implies in terms of genetic mechanisms has been largely unknown due to the fact that the genetic alterations actually responsible for tumour development have been somewhat of a black box. However, the elucidation of oncogenes and their activation has, in particular, offered a peep-hole into this black box and data are being obtained, which enables a comparison between

16

genetic lesions found in mutagenesis and those involved in cancer induction. Although short-term assays embrace over 200 established test systems and a wide array of genetic endpoints (Garrett et al., 1984), it cannot be excluded that carcinogenesis comprises DNA and chromosomal alterations not or badly covered by any of these short-term assays. It should be stressed that the endpoints measured in short-term testing, be they point mutations, chromosomal aberrations or ploidy changes, may not be the ones critical in many forms of cancer. The fact that there is a qualitative correlation between mutagenicity in short-term tests and carcinogenicity may be brought back to an association between different effects on DNA rather than some kind of a causative correlation. A hint whether the mechanism of mutations in short-term assays and the genetic alterations in carcinogenicity are closely related, should be obtained by examining not only the qualitative but also the quantitative relationship between shortterm test data on mutagenicity and animal cancer bioassay data. Bartsch et al. (1983) made a quantitative comparison of 10 direct-acting alkylating agents concerning their alkylating properties, mutagenicity in Salmonella and their carcinogenic potency in animal bioassays. There was no correlation between the alkylating properties, measured as the alkylation of 4-(4'-nitrobenzyl)pyridine or the electrophilicity (the Swain-Scott constant) on the one hand and the mutagenicity on the other hand, but there was a significant correlation between these alkylating properties and the carcinogenic potency. In accordance with these findings there was no correlation between the mutagenic and carcinogenic potency either. Although this lack of association between the mutagenic and carcinogenic potency of these alkylating chemicals may have several possible explanations, an obvious one nevertheless is that the molecular mechanism of mutagenicity, as measured in Salmonella, is not identical to the one initiating cancer.

The activation of the family of ras oncogenes involved in many cancer forms illustrates the extreme specificity in mutation pattern connected with tumour induction. If such specific mutations can operate only in combination with other genetic alterations, one can seriously question whether the

mutation rate required is in accordance with known mutation processes, or if alternative explanations have to be sought (see further Ramel, 1986). The remarkably high frequency of cancer in the offspring of male mice treated with X-irradiation and chemicals, as reported by Nomura (1982, 1986), likewise raises the question of the nature and frequency of mutations involved. It must also be realized that the formation of tumour cells involves a cascade of genetic alterations, pointing to a genetic instability, the nature of which awaits further analysis and explanation. One conclusion, which can be drawn from these and similar observations is the fact that carcinogenesis may not only comprise mutational processes and genetic endpoints revealed by standard short-term assays, but also less "conventional" genetic alterations, responsible for a high specificity, high genetic instability and high mutation rate. Although the actual nature of such "unconventional" mechanisms and processes of genetic alterations can only be a matter of speculation at present some conceivable possibilities in that direction will be discussed below. The established genetic alterations of proto-oncogenes leading to their activation may be used as a basis for the discussions of these matters. Activation of oneogenes

There are principally two types of oncogenes, viz. those operating at the level of the nucleus and the cytoplasm (or cell membrane) respectively (cf. Weinberg, 1985). The activation of nuclear oncogenes such as myc implies a change of gene expression, while the activation of cytoplasmic oncogenes at least in most cases is connected with a structural alteration of the protein. The mutational events behind these effects consequently differ. The activation of nuclear oncogenes usually does not affect the gene itself but implies an increased expression, either through an increased transcription of the gene or an increased number of copies of the gene. An increased transcription can be provided by a translocation of the oncogene to a constitutively operating promoter such as the promoter of immunoglobulin genes or by the insertion of a viral promoter. A possible alternative mechanism for increased transcription rests

17 on the role of methylation of cytosine residues in DNA involved in differentiation. A dedifferentiation of DNA can occur by a demethylation, resulting in an increased transcription. The activation of nuclear oncogenes through increased copies of the gene can be brought about both by amplification or by increased ploidy level. The cytoplasmatic oncogenes have different modes of overcoming the normal regulation of cellular proliferation, but generally structural changes of the gene product occur as the result of mutations in the oncogene. It should be mentioned, however, that in some tumours cytoplasmatic oncogenes also occur in highly amplified forms, evidently contributing to malignant transformation (Schwab et al., 1983; Hayashi et al., 1983). These processes of oncogene activation apparently do not always entail mutational mechanisms on which standard short-term assays are focused and I will therefore discuss some specific genetic effects, which can be visualized in this connection.

Amplification The level of expression of oncogenes has a profound effect on their carcinogenic properties. This constitutes a logical consequence of the role of oncogenes in growth and differentiation in general. The involvement of oncogenes in non-malignant cell proliferation can be illustrated by the analysis of oncogene expression during rat-liver regeneration, which exhibits sequential waves of increased expression of c-myc, c-fos and c-ras from increased transcription or extended life length and translation of mRNA (Thompson et al., 1986). Many malignancies are associated with a breakdown of the normal control and a constitutive increase of the expression of oncogenes, notably those oncogenes acting in the nucleus. A predominant mechanism for this effect is amplification of oncogenes. Recent experimental data indicate that amplification is a much more widespread phenomenon than previously thought and it can occur in virtually any locus for which there is a suitable selective agent to bring forth an amplified cellular population (for a review see Hamlin et al., 1984). The best studied manifestation of amplification

comes from cytological observations of homogenously stained regions (HSR) and double minutes (DM), which are connected with resistance to drugs. The resistance to the cytostatic drug methotrexate through an increased level of dihydrofolate reductase by means of amplification of the responsible gene, has been studied for 20 years, but a rapidly increasing number of other cases of amplification are being revealed. It should be pointed out that various neoplasms often exhibit HSRs and DMs (Cowell, 1982), but the causal connection between the neoplastic transformation and the occurrence of these chromosomal anomalies remains to established. There are essentially two major models to account for gene amplification. One implies unequal exchanges between repeated sequences of DNA. The result of such staggered recombination events is the formation of one deleted and one duplicated chromatid. Intrastrand recombination between repeated sequences gives rise to loss of genetic material through the formation of circular episomal elements. The other model of amplification implies multiple rounds of synthesis of those DNA sequences involved in amplification. This "onion skin" synthesis therefore would constitute an unscheduled DNA synthesis. Both models require some kind of recombination event. There are experimental data supporting both of these models and presumably both occur (see further Hamlin et al., 1984). It is of particular importance in the present context that agents acting on DNA synthesis and on the nucleotide pool as well as mutagens and carcinogens have been shown to enhance the frequency of amplification as indicated by the induction of methotrexate resistant cells, in some cases 1000-fold (see Hamlin et al., 1984). Furthermore, a synergistic effect of these agents and the tumour promoter TPA on amplification of the dehydrofolate reductase gene has been shown (Brown et al., 1983). It is obvious that the data on the process of gene amplification in conjunction with the well established occurrence of amplified oncogenes during carcinogenesis strongly suggest an important role of amplification in this context. But it is likely that the function of gene amplification includes several other biological phenomena. Thus

18 the transposition and insertion of unstable genetic elements probably involves a similar sequence of events with intrastrand pairing and recombination. This is also supported by experimental data, for instance the work by Wahl et al. (1984) on the resistance to the drug PALA through amplification of the gene CAD, which is involved in the biosynthesis of uridine. They studied cells transfected with CAD and the selection of amplification of this gene gave different results dependent of the site of integration. Amplification in situ, release of amplified DNA and transposition of such released sequences to new sites were recorded. It is likely that the occurrence of repeated sequences, if not a definite prerequisite, in any case greatly facilitates further amplification and a general genetic instability. Flavell has stated that "The combination of repeated sequences and all of the enzymes capable of responding to them are biological dynamite" (Marx, 1984). It is noteworthy in this connection that the process of tumour formation, as mentioned above, implies a cascade of genetic and chromosomal alterations suggesting a high degree of genetic instability, which at least to some extent may rest on the occurrence of repeated DNA sequences, particularly involving oncogenes. The biological significance of amplification as a response to the exposure to noxious agents as mentioned above is obvious, but one can speculate whether such reactions may not be a more general adaptive response to environmental stress situations. The mechanism of gene amplification implies a far more efficient, flexible and rapid mechanism than ordinary mutations to meet the requirement of adaptation to environmental changes. The extraordinary mutation rate and peculiar behaviour often encountered with polygenic inheritance (see Ramel, 1983, for a review) may be explained by the fact that the variation in polygenes does not depend on mutations but rather on amplification of relevant gene sequences. An interesting set of data pointing to a possible connection between polygenes and cancer induction has been provided by Nomura (1982, 1986). He performed large scale experiments, in which he treated male mice with X-rays or urethane. He obtained an increased cancer incidence in the F2 and F3 offspring of such a high frequency that it can only

be explained by a very large number of genetic sites being involved, a remarkably high mutation frequency or a mechanism other than ordinary mutations. It is tempting to draw parallels with available data on the behaviour and mutation frequency of polygenes, mostly obtained on Drosophila. A conceivable explanation of these data again may be gene amplification.

Hypomethylation of DNA Recent data indicate that amplification of genes may be initiated by an overexpression of the genes in question (Kolata, 1986). This observation constitutes a link with another phenomenon, which has been implicated in both cellular differentiation and cancer that is methylation of cytosine residues in DNA. Methylation of cytosine is closely associated with the expression of genes in mammals. Thus dividing cells have a low level of such methylation and conversely differentiated and nondividing cells are highly methylated. On the basis of this relationship Holliday has presented a model of differentiation implying that the number of cell divisions is regulated by successive methylation of cytosine in repeated DNA sequences (Holliday and Pugh, 1975). In an extension of this hypothesis (Holliday, 1979) he suggests that cancer is initiated by a hypomethylation of genes involved in carcinogenesis. Recently, Holliday has also included ageing in the same model of methylation pattern (Holliday, 1985) and he proposes the name "epimutations" for this kind of DNA alterations. The hypothesis that hypomethylation of DNA is an important mechanism behind changes of gene expression leading to malignancy has gained experimental support. In various human tumours an overall decrease of DNA methylation (GamaSosa et al., 1983) as well as hypomethylation of individual genes (Feinberg and Vogelstein, 1983; Goelz et al., 1985) have been recorded. In rat hepatomas induced by diethylnitrosamine (DEN) and aflatoxin B1 Ha- and Ki-ras oncogenes were found to be undermethylated in cytosine as compared to the normal liver (Strom et al., 1986). C-myc was also hypomethylated in tumours induced by DEN but in a tumour induced by aflatoxin B1 a 12-fold amplification of c-myc was observed, but methylation did not show any devia-

19 tion from the normal liver. However, the relationship between cytosine methylation, expression of genes and tumorigenicity is apparently composite and local characteristics of methylation may be important. Baylin et al. (1986) have shown that the methylation pattern of the 5' region of the gene for calcitonin (CT) exhibits specific patterns of methylation of cytosine dependent on the type of tumour. While the CT gene was hypomethylated in medullary thyroid carcinoma, in which the gene is expressed at high levels, the same gene shows an increased methylation in lymphomas and different types of lung cancer as compared to normal tissues. The increased methylation in these cases occurred at sites different from the one exhibiting hypomethylation in thyroid tumours. Although the functional background of this hypermethylation of the CT gene is not clear the authors suggest that it may be used as a molecular marker for lung cancers and lymphomas.

Recombinogenic effects Genetic recombination constitutes an integral part of several fundamental biological processes, such as the segregation of linked genes through meiotic crossing-over in germ cells, DNA repair and the formation of antibodies. Recombination events include several mechanisms, which may be rather disparate at the molecular level, such as meiotic and mitotic crossing-over, gene conversion and sister-chromatid exchanges. There are several reasons to suspect that recombinogenic events are fundamental in carcinogenicity. It has already been emphasized above that amplification, transposition and insertion of DNA require some form of genetic recombination although the actual mechanisms at the molecular level are not known. The recognition of particular genes in oncogenesis, oncogenes as well as tumour suppression genes, raises the question of the mode of action of these genes - - whether the activated forms need to be homozygous to carry out their carcinogenic effect. Some genes responsible for hereditary cancer forms, notably the genes for retinoblastoma and Wilm's tumour, function as recessive genes and the normal dominant allele suppresses tumour formation. In heterozygous individuals the tumorigenicity is expressed through genetic alteration rendering the recessive gene homozygous or

hemizygous (Knudsen, 1971; Murphree and Benedict, 1984). One obvious mechanism in this connection is through mitotic recombination. However, the property of oncogenes with respect to their dominant or recessive nature is not clear. Zarbl et al. (1985) found in MNU-induced tumours in rats the restriction pattern of both the normal and the activated Ha-ras-1 indicating that the gene may act in heterozygous form. However, as emphasized by Sager (1986) the argument for dominance of oncogenes is based on transfection experiments and cell transformation, in which the introduced gene may integrate in multiple copies and the question of dominance and recessivity becomes difficult to unravel. There are also other arguments against dominance of oncogenes. Santos et al. (1984) and Capon et al. (1983) thus reported homozygosity for an activated Ki-ras in different carcinomas. Furthermore cell-fusion experiments speak in favour of a recessive nature of oncogenes (Sager, 1986). If homozygosity is required for activated oncogenes to cause neoplastic transformation, it is easy to visualize the importance of genetic recombination also in this context. Yuasa et al. (1986) have recently presented an interesting observation, which may be of relevance in this connection. They found that a colon carcinoma had an activated ras oncogene, but beside that a homozygous mutation, presumably a terminal deletion, in another oncogene, myb. The normal tissue was heterozygous for this myb mutation and genetic recombination seems to be a likely mechanism, which made the mutation homozygous in the tumour tissue. It should be mentioned that Kinsella and Radman (1978) reported the induction of sister-chromatid exchanges by the tumour promoter TPA and they suggested that genetic recombination is a mechanism of relevance in promotion by rendering genes involved in carcinogenesis homozygous. This suggestion has been supported by recombinogenic effects of different promoting agents by Fahrig (1984). Gene conversion probably is the most important mechanism causing homozygosity of recessive genes in heterozygous tissues. The significance of this mechanism for rectifying repeated genes has been stressed by Baltimore (1981) and

20 observations by Liskay and Stachelek (1983) on cultured mouse cells showed that intrachromosome gene conversion between repeated genes was several orders of magnitude more frequent than point mutations. In yeast the frequency of gene conversion between duplicated genetic elements was far higher than reciprocal recombination events (Jackson and Fink, 1981). Genetic recombination may be of interest also as a possible mechanism to help explain the puzzling specificity and high mutation rate of oncogenes. As pointed out previously (Ramel, 1986), the specificity and high rate of oncogene activation recorded for instance in experiments by Guerrero et al. (1984) is difficult to reconcile with ordinary mutation frequencies. This fact makes it necessary to search for alternative explanations and to consider some kind of targeted mutagenesis. Genetic recombination may come into the picture here. It has been shown that the insertion of homologous DNA sequences into nuclei of cultured mammalian cells give rise to mutations in cognate chromosomal genes through mismatch, heteroduplex formation and incorrect repair and the genetic alterations by this mechanism is orders of magnitude higher than reverse mutation frequencies (Thomas and Capecchi, 1986). Perhaps the occurrence of several versions of oncogenes, for instance of the ras oncogenes with slight variations in the nucleotide composition, constitutes a possible source of mismatch pairing through which the above sequence of events leads to a targeted mutagenesis. It is of interest in this context that recombinogenic events, both gene conversions and reciprocal recombination are influenced by the structure of the chromatin. An association has been shown between hotspots for genetic recombination and palindromic structures (Krawinkel et al., 1986) as well as tandem repeats in histocompatibility complex (Kobori et al., 1986). This may open the possibility of localized effects also concerning activation of oncogenes. These observations are also of interest in relation to transposing elements and viral insertion sequences, which exhibit reverse and direct repeats, which may facilitate the recombination events leading to transpositions. Insertion of viral sequences is an established mechanism of oncogene activation and it is con-

ceivable that transpositions of intrinsic gene sequences may have similar consequences. It should finally be emphasized that many chemicals have been shown to induce genetic recombination and evidently they may act as carcinogens by means of this property. Wang et al. (1986) studied the intrachromosomal recombination between duplicated genes in mouse cells. They found an increased recombination with chemical carcinogens and UV, but not with 7-rays. At least 90% of the recombination events appeared to represent gene conversion. No evidence of recombination was found in cells containing only a single copy of the gene. It has recently been indicated by work on Ustilago that Z-DNA configuration initiates chromosome pairing (Kmiec and Holloman, 1985). No effect of Z-DNA occurred on recombination at a later stage after the binding of rec-1 to the recombination sites. Also illegitimate, non-homologous, recombination is localized and can be linked to sites for topoisomerase I, most sites of which contain CTT or GTT immediately to the 5' side of the cleavage sites (Bullock et al., 1985). These localized recombination mechanisms may again be of relevance also for localization of certain mutagenic events in carcinogenicity, depending on chromatin configurations.

lndirect mutagenesis Chemicals, which are genotoxic by more or less indirect pathways or act only in combination with other agents constitute problems in testing for carcinogenicity. Obvious examples of synergistic effects not revealed by the testing of single agents emanate from the fact that turnout formation is a multistep process, which needs the combination of initiation and promotion. Not all initiating agents are capable of performing the promotion step and vice versa. But there are also other examples of synergistic actions not readily detected in routine testing, neither in short-term assays nor in bioassays for carcinogenicity. This applies for instance to agents interfering with the protection against oxygen radicals, generated spontaneously or by chemicals. Rannug and Rannug (1984) have given an illustrative case in point. Tetramethylthiuram disulphide (TMTD) is a potent mutagen in Salmonella, but it does not show any correspond-

21 ing binding to DNA. It turned out that the mutagenicity of this compound was dependent on the presence of oxygen. Now T M T D and other dithiocarbamic acid derivatives cause inhibition particularly of superoxide dismutase, and the mutagenicity of T M T D presumably is not caused by a direct effect, but by a decreased protection against oxygen radicals. This interpretation was supported by the observation that T M T D exerted a strong synergistic effect with a radical-generating agent, menadione. The experimental conditions necessary to reveal such as indirect effect by radical generation are seldom at hand in standard shortterm assays, but the corresponding conditions may nevertheless occur in real life. Another group of chemicals of interest from a similar point of view are those causing proliferation of peroxisomes, predominantly in the liver. They have been shown to induce liver cancer in rodents, often at a high frequency. Peroxisomes are cellular organdies involved in lipid metabolism and chemicals inducing proliferation of peroxisomes are used as hypolipidemic drugs, for instance clofibrate. This proliferation of peroxisomes implies lipid peroxidation and the generation of radicals, which may interact with D N A and cause mutations. This highly specific sequence of events requires specific experimental conditions to reveal any effect at the D N A level and it is not astonishing that those agents are notoriously negative in standard short-term testing. How important radical induced mutations are for the carcinogenicity of these compounds is obscure, however. Analyses at our laboratory of unscheduled DNA synthesis in vivo in rat hepatocytes revealed a consistent but only small increase after proliferation of peroxisomes by clofibrate (Beije, unpublished). A possibly important indirect mutagenic effect of chemicals may occur through an induction of an imbalance of the nucleotide pool providing the precursors for D N A synthesis. As emphasized by Haynes (1985) imbalance of the nucleotide pool can have profound consequences on a wide array of genetic alterations including different mutagenic events and transformation. A change of the nucleotide balance can exert a strong synergistic effect with mutagens. Thus, hydroxyurea, which causes an imbalance of the nucleotide pool by an

inhibition of ribonucleotide reductase caused a profound synergistic effect with chemical mutagens and cotreatment with thymidine, also affecting the nucleotide pool, likewise increased the response to chemical mutagens (Jenssen, 1986).

Concluding remarks and perspectives When applying short-term assays for the prediction of carcinogenicity one has to be attentive to the fact that cancer induction obviously involves a wide spectrum of genetic alterations, and it is possible that some of the most important ones are difficult or impossible to detect by means of usual mutagenicity assays. To what extent genetic alterations, decisive for the induction of tumours, remain undetected or hidden behind other genetic effects in short-term assays can only be a matter of speculation at present. However, with reference to what has been said above there are good reasons to believe that the standard mutagenicity assays are not designed to measure several genetic endpoints of known or suspected importance in carcinogenicity. It seems, therefore, that this area deserves special attention in the future both from a fundamental and a practical point of view. Although it is reasonable to assume that the intense research on oncogenes and other genes involved in carcinogenicity will gradually provide more insight into these problems, certain research approaches specifically directed to these aspects can be suggested. Remarkably little is thus known today of chemical induction of gene amplification, polygene mutations, transpositions and changes in the intrinsic methylation pattern. Experimental techniques to study these phenomena are available or can certainly be worked out, but at present data are greatly lacking. Likewise, there is a fundamental lack of data on chemical induction of recombination in connection with carcinogenesis with the exception of sister-chromatid exchange, which has been used as one of the standard indications of chemical mutagenesis in general. In order better to understand the genetic alterations leading to cancer more specific test systems for the "unconventional" genetic effects outlined above have to be developed and applied. An important approach in this context is to establish the

22

spectrum of different genetic effects, including these "unconventional" endpoints, for various groups of chemicals. Those carcinogens, which do not act through any obvious standard mutagenic mechanism and therefore are classified as "nongenotoxic" carcinogens may be of particular interest in this connection. By virtue of their lack of measurable mutagenic and initiating property such chemicals tend to be lumped together and classified as "promoters", but it seems to be doubtful if this is the whole story in most cases. For example, T C D D essentially lacks mutagenic effects but is still the most potent carcinogen in male rats (Gold et al., 1984). Although this compound no doubt exhibits characteristics in accordance with a promotion, one can question if this is sufficient to explain this strong carcinogenic effect or if there are additional properties leading for instance to activation of oncogenes at an initiating stage of tumour development - - effects possibly not recorded as ordinary mutations. Likewise the pronounced carcinogenic effect of peroxisome proliferators as mentioned above has been assigned to a promoting effect, but a wider approach to elucidate what really happens at an early stage in hepatocytes exposed to these chemicals may give valuable information. In recent years the correlation between mutagenicity in short-term assays and carcinogenicity in animal bioassays has been studied with more randomly selected chemicals than used previously. The average correlation has been considerably lower than experienced 10 years ago for other sets of chemicals and as mentioned before these findings have raised the question of the reliability of short-term tests for the prediction of carcinogenicity in general. It seems to me important in this connection to state first of all that changes at the D N A level no doubt are critical in carcinogenicity, and this fact provides the basic and logical foundation for the use of short-term assays. It is reasonable to assume that the increasing knowledge of the molecular mechanisms involved in cancer induction will gradually enable the design of more appropriate test batteries. It is essential in this context that the performance of short-term testing should not be frozen in a bureaucratically fixed form, but that close contact is kept with the front of research in this area and that a high

degree of flexibility is left for appropriate adaptation to new findings. Although it is no doubt highly desirable that one examines whether present short-term assays really cover the most relevant genetic endpoints for the production of carcinogenicity and that one alters the testing procedure accordingly, we are still constrained to a great extent to resort to an empirical approach, based on known but incompletely understood correlations between mutagenicity and carcinogenicity. As emphasized above the power of this empirical approach relies on an efficient use of available data and a computerized system for this purpose seems inevitable. The computer systems worked out by Klopman and Rosenkranz (Rosenkranz et al., 1984) seem to be extremely useful in this Connection. An example of the practical use of this approach has recently been given by Ennever and Rosenkranz (1986) in a re-evaluation of the data on 70 chemicals negative in rodent cancer bioassays, but positive in selected short-term tests as presented by Shelby and Stasiewietz (1984). By using data from an alternate battery of tests, selected on the basis of known computerized performance of the test systems the specificity increased from 0.50 to 0.80 and "false positives" decreased accordingly. Our evaluation of the performance of short-term tests to predict carcinogenicity is almost entirely based on animal cancer data, but we are evidently not primarily interested in animal but rather human carcinogenicity. This implies an additional correlation between animal and human carcinogenicity and this correlation is unknown with the exception of the few established human carcinogens. A hint that one has to exert some caution on this point is provided by the fact that the correlation for carcinogenicity between mice and rats is not better than around 75%. As discussed previously (Ramel, 1986) it is not certain that animal cancer tests provide a safer foundation for the prediction of human carcinogenicity than a wellbalanced battery of short-term tests. References Baltimore, D. (1981) Gene conversion: Some implications for immunoglobulin genes, Cell, 24, 592-594. Bartsch, H., R. Terraeini, C. Malaveille, L. Tomatis, J. Warendorf, G. Brunn and B. Dodet (1983) Quantitative

23 comparison of carcinogenicity, mutagenicity and electrophilicity of 10 direct-acting alkylating agents and the initial O6:7-alkylguanine ratio in DNA with carcinogenic potency in rodents, Mutation Res., 110, 181-219. Baylin, S.B., J.W.M. H~Sppener, A. de Bustros, P.M. Steenbergh, C.J.M. Lips and B.D. Nelkin (1986) DNA methylation patterns of calcitonin gene in human lung cancer and lymphomas, Cancer Res., 46, 2917-2922. Brown, P.C., T.D. Tisty and R.T. Sc imke (1983) Enhancement of methotrexate resistance and dihydrofolate reductase gene amplification by treatment of mouse 3T6 cells with hydroxyurea, Mol. Cell. Biol., 3, 1097-1107. Bullock, P., J.J. Champoux and M. Botchan (1985) Association of cross-over points with topoisomerase I cleavage sites: A model for nonhomologous recombination, Science, 230, 954-958. Capon, D.J., P.H. Seeburg, J.P. McGrath, J.S. Hayflick, U. Edman, A.D. Levinson and D.V. Goeddel (1983) Activation of Ki-ras 2 gene in human colon and lung carcinomas by two different point mutations, Nature (London), 304, 507-513. Cowell, J.K. (1982) Double minutes and homogenously stained regions: gene amplification in mammalian cells, Annu. Rev. Genet., 16, 21-59. Fahrig, R. (1984) Genetic mode of action of cocarcinogens and tumor promoters in yeast and mice, Mol. Gen. Genet., 194, 7-14. Feinberg, A.P., and B. Vogelstein (1983) Hypomethylation distinguishes genes of some human cancers from their normal counterparts, Nature (London), 301, 89-92. Gama-Sosa, M.A., V.A. Slagel, R.W. Trewyn, R. Oxenhandler, K.C. Kuo, C.W. Gehrke and M. Ehrlich (1983) The 5-methylcytosine content of DNA from human tumors, Nucleic Acids Res., 11, 6883-6894. Garrett, N.E., H.F. Stack, M.R. Cross and M.D. Waters (1984) An analysis of the spectra of genetic activity produced by known or suspected carcinogens, Mutation Res., 134, 89-111. Goelz, S.E., B. Vogelstein, S.R. Hamilton and A.P. Feinberg (1985) Hypomethylation of DNA from benign and malignant human colon neoplasms, Science, 228, 187-190. Gold, L.S., C.B. Sawyer, R. Magaw, G.M. Backman, M. de Veciana, R. Levinson, N.K. Hooper, W.R. Havender, L. Bernstein, R. Peto, M.C. Pike and B.N. Ames (1984) A carcinogenic potency database of the standardized results of animal bioassays. Environ. Health Perspect., 58, 9-319. Guerrero, I.A., A. Villasante, V. Corces and A. Pellicer (1984) Activation of a c-K-ras oncogene by somatic mutation in mouse lymphomas induced by gamma radiation, Science, 225, 1159-1162. Hamlin, J.L., J.D. Milbrandt, N.H. Heintz and J.C. Azizkhan (1984) DNA sequence amplification in mammalian cells, Int. Rev. Cytol., 90, 31-82. Haynes, R.H. (1985) Molecular mechanisms in genetic stability and change: The role of deoxyribonucleotide pool balance, in: F.J. de Serres (Ed.), Genetic Consequences of Nucleotide Pool Imbalance, Plenum, New York, pp. 1-23. Holliday, R. (1979) A new theory of carcinogenesis, Br. J. Cancer, 40, 513-522.

Holliday, R. (1985) The significance of DNA methylation, in: A.D. Woodhead, A.D. Blackett and A. Hollaender (Eds.), Molecular Biology of Ageing, Plenum, New York, pp. 269-283. Holliday, R., and J.E. Pugh (1975) DNA modification mechanisms and gene activity during the development, Science, 187, 226-232. Jackson, J.A., and G.R. Fink (1981) Gene conversion between duplicated genetic elements in yeast, Nature (London) 292, 306-311. Jenssen, D. (1986) Enhanced mutagenicity of low doses of alkylating agents and UV-light by inhibition of ribonucleotide reductase, in: C. Ramel, B. Lambert and J. Magnusson (Eds.), Genetic Toxicology of Environmental Chemicals, Part A, Basic Principles and Mechanism of Action, pp. 541-549. Kinsella, A.R., and M. Radman (1978) Tumor promoter induces sister chromatid exchanges: Relevance to mechanism of carcinogenesis, Prec. Natl. Acad. Sci. (U.S.A.), 75, 6149-6153. Kmiec, E.B., and W.K. Holloman (1985) Homologous pairing of DNA molecules by Ustilago rec 1 protein is promoted by sequences of Z-DNA, Cell, 44, 545-554. Knudsen, A.G. (1971) Mutation and cancer: Statistical study of retinoblastoma, Prec. Natl. Acad. Sci. (U.S.A.), 68, 820. Kobori, J.A., E. Strauss, K. Minard and L. Hood (1986) Molecular analysis of the hotspot of recombination in the murine major histocompatibility complex, Science, 234, 173-179. Kolata, G. (1986) Why do cancer cells resist drugs? Science, 231, 220-221. Krawinkel, U., G. Zoebelein and A.L.M. Bothwell (1986) Palindromic sequences are associated with sites of DNA breakage during gene conversion, Nucleic Acids Res., 14, 3871-3881. Liskay, R.M., and J.L. Stachelek (1983) Evidence for intrachromosomal gene conversion in cultured mouse cells, Cell, 35, 157-165. Marx, J.L. (1984) Instability in plants and the ghost of Lamarck, Science, 224, 1415-1416. Mayashi, K., T. Kakizoe and T. Sugimura (1983) In vivo amplification and rearrangement of the c-Ha-ras-1 sequence in a human bladder carcinoma, Gann, 74, 798-801. Murphree, A.L., and W.F. Benedict (1984) Retinoblastoma: Clues to human oncogenesis, Science, 223, 1028-1033. Nomura, T. (1982) Parental exposure to X-rays and chemicals induces heritable tumours and anomalies in mice, Nature (London), 296, 575-577. Nomura, T. (1986) Further studies on X-ray and chemically induced germ-line alterations causing tumors and malformations in mice, in: C. Ramel, B. Lambert and J. Magnusson (Eds.), Genetic Toxicology of Environmental Chemicals, Part B, Genetic Effects and Applied Mutagenesis, Liss, New York, pp. 13-20. Ramel, C. (1983) Polygenic effects and genetic changes affecting quantitative traits, Mutation Res., 114, 107-116. Ramel, C. (1986) Deployment of short-term assays for the detection of carcinogens: genetic and molecular considerations, Mutation Res., 168, 327-342.

24 Rannug, A., and U. Rannug (1984) Enzyme inhibition as a possible mechanism of the mutagenicity of dithiocarbamic acid derivatives in Salmonella typhimurium, Chem.-Biol. Interact., 49, 329-340. Rosenkranz, H.S., C.S. Mitchell and G. Klopman (1985) Artificial intelligence and Bayesian decision theory in the prediction of chemical carcinogens, Mutation Res., 150, 1-11. Sager, R. (1986) Genetic suppression of tumor formation: A new frontier in cancer research, Cancer Res., 46, 1573-1580. Santos, E., D. Martin-Zanca, E.P. Reddy, M.A. Pierotti, G. Della Porta and M. Barbacid (1984) Malignant activation of a K-ras-oncogene in lung carcinoma but in normal tissue of the same patient, Science, 223, 661-664. Schwab, M., K. Alitalo, H.E. Varmus, J.M. Bishop and D. George (1983) A cellular oncogene (c-Ki-ras) is amplified, unexpressed and located within karyotypic anomality in mouse adrenocortical tumour cells, Nature (London), 303, 497-501. Shelby, M.D., and S. Stasiewicz (1984) Chemicals showing no evidence of carcinogenicity in long-term, two-species rodent studies: The need for short-term data, Environ. Mutagen., 6, 871-878. Strom, S., J. Giles and A. Mitchell (1986) Amplification and methylation of cellular oncogenes in chemically induced rat hepatomas, Proc. AACR, 27, 96. Thomas, K.R., and M.R. Capecchi (1986) Induction of homologous DNA sequences into mammalian cells induces mutations in cognate gene, Nature (London), 324, 34-38.

Thompson, N.L., J.E. Mead, L. Braun, M. Goyette, P.R. Shrank and N. Fausto (1986) Sequential protooncogene expression during rat liver regeneration, Cancer Res., 46, 3111-3117. Wahl, G.M., B.R. de Saint Vincent and M.L. de Rose (1984) Effect of chromosomal position on amplification of transfected genes in animal cells, Nature (London), 307, 516-520. Wang, Y., V.M. Maher, J.J. McCormick, and R.M. Liskay (1986) Carcinogen-induced homologous recombination between duplicated chromosomal sequences in mammalian cells, Environ. Mutagen., 8, Suppl. 6, 89. Weinberg, A. (1985) The action of oncogenes in the cytoplasm and nucleus, Science, 230, 770-776. Yuasa, Y., E. Premkumar Reddy, J.-S. Rhim, S.R. Tronick and S.A. Aaronson (1986) Activated N-ras in a human rectal carcinoma cell line associated with clonal homozygosity in myb locus-restriction fragment polymorphism, Gann, 77, 639-647. Zarbl, H., S. Sukumar, A.V. Arther, D. Martin-Zanca and M. Barbacid (1985) Direct mutagenesis of Ha-ras-1 oncogenes by N-nitroso-N-methylurea during initiation of mammary carcinogenesis in rats, Nature (London), 315, 382-385. Zeiger, E., and R.W. Tennant (1986) Mutagenesis, clastogenesis, carcinogenesis: Expectations, correlations and relations, in: C. Ramel, B. Lambert and J. Magnusson (Eds.), Genetic Toxicology of Environmental Chemicals, Part B, Genetic Effects and Applied Mutagenesis, Liss, New York, pp. 75-84.