- Email: [email protected]
Carcinogen risk assessment: a necessary dilemma?
Cancer Letters 117 (1997) 209–215
Carcinogen risk assessment: a necessary dilemma? David B. Clayson P.O. Box 248, 3886 Carp Road, Carp, Ontario K0A 1L0, Canada Received 1 February 1997; accepted 28 February 1997
Abstract Carcinogen risk assessment is the process by which an attempt is made to estimate human risk due to carcinogens, from the results of animal studies. It is based upon a number of prudent default assumptions, that is, assumptions that cannot be proved scientifically because either the basic concept is philosophical in nature or because the amount of scientific evidence required is too costly to obtain even on a world-wide basis. Recently, scientific effort has shown that more and more examples have been described suggesting these examples do not behave in the way indicated by the default assumptions. Since carcinogen risk assessment processes were initiated, it has been demonstrated that cancer may arise by four or more different mechanisms. It is the purpose of this paper to enquire whether consideration of these basically different mechanisms may facilitate carcinogen risk assessment. 1997 Elsevier Science Ireland Ltd. Keywords: Carcinogen risk assessment; Default assumptions; Carcinogenesis mechanisms; Electrophile generation; Oxidative DNA stress; Cellular proliferation; DNA-protein ligands and DNA expression
1. Introduction The Delaney Clause of the United States Food and Drug Act, 1958 [1] specifically forbade the addition to food of any chemical that caused cancer in man or in animals. This clause contains two highly prudent default assumptions: first, that any agent found to be a carcinogen in animals will likewise be cancer-forming in humans, and second that there is no safe level of exposure for humans to any carcinogen. It effectively removed the ability of the United States Food and Drug Administration to make reasoned judgments on the relevance of any animal carcinogen to humans. Such a draconian measure could not be applied to other areas of the environment without serious damage to the national economy. The United States
Environmental Protection Agency and the Occupational Safety and Health Administration therefore decided to establish methods to calculate the risks to humans posed by exposure to various animal carcinogens. Two further default assumptions were necessary. First, as most animal bioassays are conducted at very high doses of test agent (the maximum tolerated dose, MTD) whereas human exposures are generally much lower, it was assumed that cancer formation would be linearly related to the level of carcinogen exposure. Second, because cost severely restricted the number of animals employed in any one bioassay, it was decided to use the upper confidence limit of the bioassay results rather than the actual experimental data. Barnard [2] has estimated that, in total, these default assumptions may introduce an
0304-3835/97/$17.00 1997 Elsevier Science Ireland Ltd. All rights reserved PII S0304- 3835 (97 )0 0232- 2
210
D.B. Clayson / Cancer Letters 117 (1997) 209–215
exaggeration of the effect of exposure of humans to an animal carcinogen of 16- to 10 800-fold. This is not very satisfactory. 1.1. Exceptions to the default assumptions Clayson and Iverson [3] discussed the exceptions to the current default assumptions in some detail. Only a few examples will be mentioned here. Not all animal carcinogens are effective in humans. This is well illustrated by the non-genotoxic male rat-specific kidney carcinogens [4–6]. The male rat, but not the female, nor humans nor mice, elaborates a specific low molecular weight alpha2u-globulin that forms ligands with a variety of substances, such as the flavoring agent dlimonene. These ligands are absorbed from the blood stream at a specific part of the renal tubules, where hyaline droplets and eventually tumors are formed. The lack of the specific globulin in humans makes it highly unlikely that kidney tumors will be formed. Again, rats and mice that develop bladder stone later develop bladder cancer. The human response to bladder stone is certainly much less marked than in rodents and may be due to stones forming as a result of bladder tumors rather than the tumors resulting from the presence of stone [7]. That is, non-genotoxic rodent bladder stone-formers are either non-carcinogenic in humans or, at the most, exceptionally weak tumor inducers. It is not true that no carcinogen exhibits a threshold or exposure level below which it is safe. This may be illustrated by the food-use phenolic antioxidant, butylated hydroxyanisole (BHA). BHA when administered at a level of 2% in the diet leads to rat forestomach carcinomas [8]. While BHA is antioxidant at lower exposure levels, it is prooxidant at higher levels. At the point at which this property changes (ca. 0.1%) there is a threshold involving not only tumorigenesis but also the induction of pathological lesions such as inflammation, irritation and cellular proliferation [9]. Dose-tumor response is not always linear. This is clearly illustrated by the costly ED01 study [10] in which a variety of levels of 2-acetylaminofluorene (AAF) were fed to a total of over 23 000 female BALB/C mice to determine the shape of the tumordose response curve to a tumor-response level of 1%. While the curve for liver tumors approximated to
linearity, that for urinary bladder tumors was more akin to an ice-hockey stick with little tumorigenicity at lower exposures to AAF but a steep rise thereafter. The concept of using upper confidence levels of induced tumors presents few problems if most of the animals develop tumors in response to the agent. It represents an attempt to compensate for the relatively few animals used in the bioassay. The result is catastrophic if there is no increase or only a minor increase in tumors. Let us suppose that a bioassay of 5% water in the diet had been performed and there was no increase in the number of tumors. Use of the upper confidence limit to determine a ‘safe’ level of water would lead to a recommended human consumption of water of only a few milliliters/day. As Barnard [2] points out, these default assumptions leave much to be desired. An attempt needs to be made to determine whether mechanistic considerations can help.
2. Carcinogenic mechanisms The seminal observations of Berenblum and Shubik [11–13] provided the first firm experimental evidence that the induction of cancer consisted of two or more stages. They showed that many mainly benign tumors in mouse’s skin could be induced by a single limited administration of a potently active polycyclic aromatic hydrocarbon such as 7,12-dimethylbenz(a)anthracene, followed by repeated painting with croton oil. Neither the hydrocarbon nor the croton oil alone led to the induction of tumors on an appreciable scale. The use of two quite different agents made it clear that two different targets were involved in skin tumor formation; these were named ‘initiation’ and ‘promotion’. As is usual with important observations, Berenblum and Shubik’s work engendered much further study aimed at increasing the number of agents that were initiating or promoting in mouse’s skin, were able to induce similar effects in other tissues, and importantly to demonstrate firmly that initiated cells could lie dormant in their tissues of origin for extended periods. This effort is still continuing 50 years after the initial reports appeared. Growing appreciation of the carcinogenic process has led to the belief that initiation is due to mutations
D.B. Clayson / Cancer Letters 117 (1997) 209–215
in oncogenes and tumor suppressor genes. Cohen and Ellwein [14], amongst others, have suggested that cancer formation is, at its simplest, a cascade of specific mutations interspersed with cellular proliferation to increase the size of the clones of mutated cells. This does not exclude other possibly important factors such as modification of the immunological status of the host, but nevertheless provides a convenient and largely valid model on which to base a newer approach to risk assessment. 2.1. Initiation involving electrophile generation The initial molecularly-based mechanism of carcinogenesis was due to Drs. James and Elizabeth Miller and their colleagues, who investigated the metabolic activation of 2-acetylamino-fluorene (AAF). They first identified a new metabolite, the N-hydroxy derivative, which they demonstrated was a somewhat more potent carcinogen than the parent acetamide in several rodent species. However, the N-hydroxy derivative lacked the ability to interact with DNA in vitro, a property of AAF in the living animal. It was shown that in rat liver, a sulfotransferase enzyme was able to create the sulfate ester of N-hydroxy-2-AAF. This, instead of being more easily excreted than the Nhydroxy derivative, was highly unstable. The sulfate ion acted as a leaving group, liberating the highly reactive 2-acetylimino-fluorene moiety that interacted with a wide variety of cell constituents including DNA [15–18]. This demonstration engendered a great deal of further work that showed N-hydroxylation followed by conjugation by one of a variety of groups was a major route in aromatic amine carcinogenesis. It was also found that other carcinogens such as the polycyclic aromatic hydrocarbons, the nitrosamines, and agents such as the aflatoxins all induced cancer through their conversion to a reactive entity, an electrophile. Certain other carcinogens such as the nitrosamides, the sulfur and nitrogen mustards broke down spontaneously into electrophiles in biological media [19]. 2.2. Initiation involving oxygen (and nitrogen) radicals Radiation-induced cancer has been recognized to
211
be associated with radical formation for many years. Only recently has the possible importance of radical formation in chemically-induced cancer begun to be appreciated [20–22]. Fat metabolism provides a useful example of what is happening. During fat metabolism, a hydroxy radical is produced that is sufficiently stable to diffuse from its site of origin but is sufficiently reactive to interact with DNA. In addition, although hydrogen peroxide does not interact directly with DNA, if it is not degraded by catalase, it may react with transition metal ions including those derived from iron and copper to produce a highly reactive oxygen radical. This radical, if produced in close proximity to the DNA, like the less active radical mentioned previously, may produce one of about 20 different lesions on the DNA. These include single and double strand breaks, mono- and dihyroxylated bases, and so on [23]. Oxygen radical formation differs from electrophile generation in so far as it is a major naturally-occurring and ongoing process. Studies in acellular systems have shown that the most adequately studied oxygen radical-induced lesion, 8-hydroxy-2-deoxyguanosine, leads to mutation on DNA replication [24,25]. Ames and Gold [21] estimated that the DNA from an untreated adult rat liver contained an average of about 106 oxygen radical-induced lesions in each cell genome. Others have reported similarly high levels in other human and animal cells. After 66% partial hepatectomy in rats, the liver returns to near its normal size and functionality, indicating that these cells must be able to protect themselves efficiently from mutational events caused by these lesions (for references see [22]). This implies that the risks associated with these oxygen radical-induced lesions may differ at least quantitatively from those due to electrophile generators and that this factor needs attention in quantitative risk assessment. 2.3. DNA lesions and cellular mutation It should not be imagined that every induced DNA lesion will lead to a cancer mutation. Some lesions will so poison the cell as to lead to cytotoxicity. Others will be faithfully repaired by the appropriate DNA repair enzymes. Mutations may occur as the result of base mispairing if the cell and its DNA replicates before the DNA lesion is repaired.
212
D.B. Clayson / Cancer Letters 117 (1997) 209–215
2.4. Cellular proliferation The importance of cellular proliferation in carcinogenesis has not received the attention it merits because of the amount of labor involved in counting individual cells in tissue slides. Hopefully, this demand on labor will be overcome in the near future by the use of computer-based image analysis. Electrophile-generating and possibly oxygen radical-forming carcinogens induce proliferative cell regeneration in response to cytotoxicity and possibly through their ability to interfere with apoptosis, which is the mechanism of programmed cell death that keeps tissue mass in proportion to the needs of the total organism. In some cases, induction of cellular proliferation appears to be the major effect of a particular carcinogen. This is certainly the case with non-genotoxic agents leading to stone formation in rodent bladders and thence to bladder cancer [7,26]. A naturally occurring or surgically implanted stone may increase the rate of cell turnover many times. Increased cell proliferation has also been reported with a number of rodent liver carcinogens, such as chloroform [27,28]. There is a need to determine the nature of the initiating stimulus before carcinogenesis risk assessment is attempted. Initiation in such cases could occur naturally in a tissue (genetic inheritance), could result from naturally occurring carcinogens or accidental exposure to another carcinogen, or to non-faithful DNA replication. In my view, the former is marginally more likely especially in animals that demonstrate a significant background tumor yield. A further complication in the study of cellular proliferation is that initiated cells at different stages of malignancy may differ in their response to a particular agent. This is well illustrated by the non-genotoxic liver carcinogen, butylated hydroxytoluene (BHT) [29–31]. When administered acutely to rats, BHT induced a relatively minor, time-limited increase in cell turnover. This relapsed to below the levels observed in the untreated liver tissue after 4 days. However, BHT had quite a major effect when administered to rats with enzyme-altered foci in their livers [32,33]. Attempts to derive dose-response relationships for cellular proliferation to the relatively low exposure levels required for carcinogen risk assessment for
humans is seldom attempted because of the labor involved. The derivation of computer-based image analysis is essential for this purpose. 2.5. Ligand formation and DNA interaction In some cases, application of a carcinogen leads to physical interaction with a specific protein and the ligand combining with a specific site on the DNA to modify its expression. This was first demonstrated by Jensen [34] with estrogen. The ligand DNA interaction led to enhanced cellular proliferation in estrogenresponsive tissues such as the mammary gland. Other examples include 2,3,7,8-tetrachloro-para-dibenzodioxin [35] and the peroxisome proliferators [36,37]. Where possible, it is probably better to consider such agents in terms of the more general tissular changes that they engender (e.g. cell proliferation) than to try to assess the risk they pose in terms of either ligand formation or DNA interaction. It has been suggested that agents that act in this way will de facto exhibit a threshold or no effect level because of the need to attain a certain critical concentration of ligand before DNA expression is altered. This, like the additional assumption that all non-genotoxic carcinogens may be thresholded, is not tenable. Estrogen, for example, is continually controlling the rate of cellular proliferation in the mammary gland and in other responsive tissues. Addition of any level of excess estrogen should have the potential to modify tissue behavior. Thresholds require real evidence, not assumptions, before they can be accepted.
3. Carcinogenesis mechanisms and risk assessment The preceding brief account of carcinogenesis mechanisms, which has relied heavily on an initiation/promotion or mutation/cellular proliferation model, makes it abundantly clear that a single format for the assessment of human risk from animal carcinogens will seldom be satisfactory. Reference to Williams [38] shows that, possibly, an even more complex approach will eventually be required than that suggested here. Because of the need to use the maximum tolerated dose (MTD) in rodent bioassays, the carcinogenicity data by itself is not adequate for risk assessment. Because of the relatively imprecise
D.B. Clayson / Cancer Letters 117 (1997) 209–215
methods by which the MTD is estimated, such a dose may be quite toxic to the host tissues leading to the suppression of tumor formation or, alternatively, may exhibit an enhancing or promotional effect. That is, at such high levels, the upper reaches of the dose-carcinogen response curve may be severely distorted making it virtually useless for extrapolation to the estimated levels of human exposure. The first need in assessing the risk associated with a novel carcinogen is to determine whether or not it is genotoxic. At this time positive genotoxicity data is a highly efficient way in which to identify electrophile generators. The same does not appear to be as true for agents leading to oxidative stress. Attempts are currently in progress to devise methodology to recognize agents leading to the formation of oxygen stress [39]. Such methods will require extensive validation before they can be used with confidence. Felton and Turteltaub [40] described an accelerator mass spectrometric method to measure the amount of DNA adducts formed after the administration of low levels of 2-acetylaminofluorene to rats. If this approach can be transferred from the research to the routine laboratory, it promises a far better indication of the probable shape of the dose cancer-response curve for electrophile generators than the presently used prudent but unscientific default assumptions. Such an approach makes the apparently reasonable assumption that DNA adduct formation for electrophile generators is proportional to tumor formation. Such an approach is not yet possible for agents leading to cancer through the induction of oxygen radicals. Methods need to be elaborated and validated. It should not be assumed, in the interim, that these agents will mimic electrophile generators. As already noted oxygen-induced DNA damage is a high level ongoing process in mammals and this may seriously affect the shape of the carcinogen dose response curve. It may be speculated, at this time, that substantial increases in oxygen radical formation will be necessary to overcome the cell’s apparent ability to defend itself against the natural occurrence of these lesions. However, the fact that radiation-induced cancer involves radical formation and no threshold has been suggested for radiation-induced tumors means that such a speculation must be regarded with great caution. The non-genotoxic carcinogens present even
213
greater difficulties because of the labor involved in quite primitive assessments of the dose response curve for induced cellular proliferation. It is anticipated that novel techniques, possibly computerbased image analysis, will be required to obtain meaningful exposure response curves and, where they exist, threshold values for most individual non-genotoxic carcinogens. A further problem arises that while, with the necessary effort, the approaches suggested here will yield useful data in animal studies, similar procedures in humans are unethical. To complete an effective approach to carcinogenesis risk assessment, it will also be necessary to devise methodology, largely in vitro, to determine similar parameters in humans. Without this not inconsiderable effort, carcinogenesis risk assessment will continue to be grossly inadequate and in my opinion a waste of valuable time and effort.
Acknowledgements I wish to thank Dr. Philippe Shubik most warmly for many personal kindnesses during my career including (1) inviting me to join his group in Omaha, Nebraska to continue working in chemical carcinogenesis, (2) explaining to me the nuances of international panel meetings so that I might participate more effectively, and (3) above all, for his cooperation and encouragement in founding and managing Cancer Letters for over 20 years. I must also thank Drs. W.G. Flamm and F. Iverson for reading this manuscript and making helpful comments.
References [1] United States Food and Drug Act, Public Law 85–929, 85th Congress H.R. 13254, September 5th, 1958. [2] R.C. Barnard, Risk assessment. The default assumption controversy, Regul. Toxicol. Pharmacol. 21 (1995) 431–438. [3] D.B. Clayson, F. Iverson, Cancer risk assessment at the crossroads: the need to turn to a biological approach, Regul. Toxicol. Pharmacol. 24 (1996) 45–59. [4] B.G. Short, V.L. Burnett, M.G. Cox, J.S. Bus, J.A. Swenberg, Site-specific renal cytotoxicity and cell proliferation in male rats exposed to petroleum hydrocarbons, Lab. Invest. 57 (1987) 564–577. [5] W.G. Flamm, L.D. Lehmann-McKeeman, The human relevance of the renal tumor-inducing potential of d-limonene in
214
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14] [15]
[16]
[17]
[18]
[19]
[20]
[21]
D.B. Clayson / Cancer Letters 117 (1997) 209–215 male rats: implications for risk assessment, Regul. Toxicol. Pharmacol. 13 (1991) 70–86. S.J. Borghoff, A.B. Miller, J.P. Bowen, J.A. Swenberg, Characteristics of chemical binding to a-2u-globulin in vitro: evaluating structure-activity relationships, Toxicol. Appl. Pharmacol. 107 (1991) 228–238. D.B. Clayson, L. Fishbein, S.M. Cohen, Effects of stones and other physical factors on the induction of rodent bladder cancer, Food Chem. Toxicol. 33 (1995) 771–784. N. Ito, S. Fukushima, A. Hagiwara, M. Shibata, T. Ogiso, Carcinogenicity of butylated hydroxyanisole in F344 rats, J. Natl. Cancer Inst. 70 (1983) 343–352. D.B. Clayson, F. Iverson, E.A. Nera, E. Lok, The importance of cellular proliferation induced by BHA and BHT, in: G.M. Williams, H. Sics, G.T. Baker, J.W. Erdman, C.J. Henry (Eds.), Antioxidants: Chemical, Physiological, Nutritional and Toxicological Aspects, Princetown Scientific, Princetown, NJ, 1992, pp. 231–242. J.A. Staffa, M.A. Mehlman (Eds.), Innovations in Carcinogen Risk Assessment (ED01 study). Pathotox, Park Forest South, IL, 1979. I. Berenblum, P. Shubik, The role of croton oil applications, associated with a single painting of a carcinogen, in tumour induction of the mouse’s skin, Br. J. Cancer 1 (1947) 379– 382. I. Berenblum, P. Shubik, A new, quantitative, approach to the study of the stages of chemical carcinogenesis in the mouse’s skin, Br. J. Cancer 1 (1947) 383–387. I. Berenblum, P. Shubik, The persistence of the latent tumour cells induced in the mouse’s skin by a single application of 9:10-dimethyl-1:2-benzanthracene, Br. J. Cancer 3 (1949) 383–387. S.M. Cohen, I.B. Ellwein, Genetic errors, cell proliferation and carcinogenesis, Cancer Res. 51 (1991) 6493–6505. E.C. Miller, J.A. Miller, H.A. Hartmann, N-Hydroxy-2-acetylaminofluorene: a metabolite of 2-acetylaminofluorene with increased carcinogenic activity in the rat, Cancer Res. 21 (1961) 815–824. E.C. Miller, J.A. Miller, M. Enomoto, The comparative carcinogenicity of 2-acetylaminofluorene and its N-hydroxy metabolite in mice, hamsters and guinea pigs, Cancer Res. 24 (1964) 2018–2031. J.R. DeBaun, E.C. Miller, J.A. Miller, N-Hydroxy-2-acetylaminofluorene sulfotransferase: its probable role in carcinogenesis and in protein-(methion-S-yl) binding in rat liver, Cancer Res. 30 (1970) 577–595. J.A. Miller, Carcinogenesis by chemicals: an overview (G.H.A. Clowes Memorial Lecture), Cancer Res. 30 (1970) 859–876. C.E. Searle (Ed.), Chemical Carcinogens, 2nd ed., ACS Monograph 182, American Chemical Society, Washington, DC, 1984. B. Halliwell, O.I. Aruoma, DNA-damage by oxygen derived species: its mechanism and measurement in mammalian systems, FEBS Lett. 281 (1991) 9–19. B.N. Ames, L.S. Gold, Endogenous mutagens and the causes of aging and cancer, Mutat. Res. 250 (1991) 3–16.
[22] D.B. Clayson, R. Mehta, F. Iverson, Oxidative DNA damage: the effect of certain genotoxic and operationally non-genotoxic carcinogens, Mutat. Res. 317 (1994) 23–42. [23] C. Von Sonntag, The Chemical Basis of Radiation Biology. Taylor and Francis, London, 1987. [24] Y. Kuchino, F. Mori, H. Kasai, H. Inoue, S. Iwai, K. Miura, E. Ohtsuka, S. Nishimura, Misreading of DNA templates containing 8-hydroxydeoxyguanosine at the modified base and at adjacent residues, Nature (London) 327 (1987) 77– 79. [25] R.A. Floyd, The role of 8-hydroxyguanosine in carcinogenesis, Carcinogenesis 11 (1990) 1447–1450. [26] D.B. Clayson, J.A.S. Pringle, The influence of a foreign body on the induction of tumours of the bladder epithelium of the mouse, Br. J. Cancer 20 (1966) 564–568. [27] J.L. Larson, D.C. Wolf, B.E. Butterworth, Induced cytotoxicity and cell proliferation in the hepatocarcinogenicity of chloroform in female B6C3F1 mice: comparison of administration by gavage in corn oil vs. ad libitum in drinking water, Fundam. Appl. Toxicol. 22 (1994) 90–102. [28] M.A. Pereira, Route of administration determines whether chloroform enhances or inhibits cell proliferation in the liver of B6C3F1 mice, Fundam. Appl. Toxicol. 23 (1994) 87–92. [29] P. Olsen, O. Meyer, N. Gille, G. Wurtzen, Carcinogenicity study on butylated hydroxytoluene (BHT) in Wistar rats exposed in utero, Food Chem. Toxicol. 24 (1986) 1– 12. [30] K. Inai, T. Kabuke, S. Nambu, T. Takemoto, E. Kou, H. Kishina, M. Fujihara, S. Yonhara, S. Suehiro, T. Tsuya, K. Horiuchi, S. Tokuoka, Hepatocellular tumorigenicity of butylated hydroxytoluene administered orally to B6C4F1 mice, Jpn. J. Cancer 70 (1988) 49–58. [31] E.M. Bombard, J.N. Bremmer, B.A. Herbold, Review of the mutagenicity/genotoxicity of butylated hydroxytoluene, Mutat. Res. 288 (1992) 187–200. [32] D. Briggs, E. Lok, E.A. Nera, K. Karpinski, D.B. Clayson, Short-term effects of butylated hydroxytoluene on Wistar rat liver, urinary bladder and thyroid gland, Cancer Lett. 46 (1989) 31–36. [33] E. Lok, R. Mehta, P. Jee, G. Laver, E.A. Nera, E. McMullen, D.B. Clayson, The effect of butylated hydroxytoluene on the growth of enzyme-altered foci in male Fischer 344 rat liver tissue, Carcinogenesis 16 (1995) 1071–1078. [34] E.V. Jensen, Estrogen receptors in hormone-dependent breast cancers, Cancer Res. 35 (1975) 2262–3364. [35] A. Poland, E. Glover, A.S. Konde, Stereospecific, high affinity binding of 2,3,7.8-tetrachloro-p-dibenzodioxin by hepatic cytosol. Evidence that the binding species is receptor for induction of arylhydrocarbon hydroxylase, J. Biol. Chem. 231 (1976) 4936–4946. [36] J.D. Tugwood, I. Issemann, R.G. Anderson, K.R. Bundell, W.L. McPheat, S. Green, The mouse peroxisome proliferator activated receptor recognizes a response element in the 5′ flanking sequence of the rat acyl CoA oxidase gene, EMBO J. 11 (1992) 433–439. [37] I. Issermann, S. Green, Activation of a member of the steroid
D.B. Clayson / Cancer Letters 117 (1997) 209–215 hormone receptor superfamily by peroxisome proliferators, Nature (London) 347 (1990) 645–650. [38] G.M. Williams, Chemicals with carcinogenic activity in the rodent liver. Cancer Lett. 117 (1997) this issue. [39] J.J.P. Gille, C.C.M. van Berkel, H. Joenje, Mutagenicity of
215
metabolic oxygen radicals in mammalian cell cultures, Carcinogenesis 15 (1994) 2695–2699. [40] J.S. Felton, K.W. Turteltaub, Accelerator mass spectrometry for measuring low dose carcinogen binding to DNA, Environ. Health Perspect. 102 (1994) 450–452.
Carcinogen risk assessment: a necessary dilemma? David B. Clayson P.O. Box 248, 3886 Carp Road, Carp, Ontario K0A 1L0, Canada Received 1 February 1997; accepted 28 February 1997
Abstract Carcinogen risk assessment is the process by which an attempt is made to estimate human risk due to carcinogens, from the results of animal studies. It is based upon a number of prudent default assumptions, that is, assumptions that cannot be proved scientifically because either the basic concept is philosophical in nature or because the amount of scientific evidence required is too costly to obtain even on a world-wide basis. Recently, scientific effort has shown that more and more examples have been described suggesting these examples do not behave in the way indicated by the default assumptions. Since carcinogen risk assessment processes were initiated, it has been demonstrated that cancer may arise by four or more different mechanisms. It is the purpose of this paper to enquire whether consideration of these basically different mechanisms may facilitate carcinogen risk assessment. 1997 Elsevier Science Ireland Ltd. Keywords: Carcinogen risk assessment; Default assumptions; Carcinogenesis mechanisms; Electrophile generation; Oxidative DNA stress; Cellular proliferation; DNA-protein ligands and DNA expression
1. Introduction The Delaney Clause of the United States Food and Drug Act, 1958 [1] specifically forbade the addition to food of any chemical that caused cancer in man or in animals. This clause contains two highly prudent default assumptions: first, that any agent found to be a carcinogen in animals will likewise be cancer-forming in humans, and second that there is no safe level of exposure for humans to any carcinogen. It effectively removed the ability of the United States Food and Drug Administration to make reasoned judgments on the relevance of any animal carcinogen to humans. Such a draconian measure could not be applied to other areas of the environment without serious damage to the national economy. The United States
Environmental Protection Agency and the Occupational Safety and Health Administration therefore decided to establish methods to calculate the risks to humans posed by exposure to various animal carcinogens. Two further default assumptions were necessary. First, as most animal bioassays are conducted at very high doses of test agent (the maximum tolerated dose, MTD) whereas human exposures are generally much lower, it was assumed that cancer formation would be linearly related to the level of carcinogen exposure. Second, because cost severely restricted the number of animals employed in any one bioassay, it was decided to use the upper confidence limit of the bioassay results rather than the actual experimental data. Barnard [2] has estimated that, in total, these default assumptions may introduce an
0304-3835/97/$17.00 1997 Elsevier Science Ireland Ltd. All rights reserved PII S0304- 3835 (97 )0 0232- 2
210
D.B. Clayson / Cancer Letters 117 (1997) 209–215
exaggeration of the effect of exposure of humans to an animal carcinogen of 16- to 10 800-fold. This is not very satisfactory. 1.1. Exceptions to the default assumptions Clayson and Iverson [3] discussed the exceptions to the current default assumptions in some detail. Only a few examples will be mentioned here. Not all animal carcinogens are effective in humans. This is well illustrated by the non-genotoxic male rat-specific kidney carcinogens [4–6]. The male rat, but not the female, nor humans nor mice, elaborates a specific low molecular weight alpha2u-globulin that forms ligands with a variety of substances, such as the flavoring agent dlimonene. These ligands are absorbed from the blood stream at a specific part of the renal tubules, where hyaline droplets and eventually tumors are formed. The lack of the specific globulin in humans makes it highly unlikely that kidney tumors will be formed. Again, rats and mice that develop bladder stone later develop bladder cancer. The human response to bladder stone is certainly much less marked than in rodents and may be due to stones forming as a result of bladder tumors rather than the tumors resulting from the presence of stone [7]. That is, non-genotoxic rodent bladder stone-formers are either non-carcinogenic in humans or, at the most, exceptionally weak tumor inducers. It is not true that no carcinogen exhibits a threshold or exposure level below which it is safe. This may be illustrated by the food-use phenolic antioxidant, butylated hydroxyanisole (BHA). BHA when administered at a level of 2% in the diet leads to rat forestomach carcinomas [8]. While BHA is antioxidant at lower exposure levels, it is prooxidant at higher levels. At the point at which this property changes (ca. 0.1%) there is a threshold involving not only tumorigenesis but also the induction of pathological lesions such as inflammation, irritation and cellular proliferation [9]. Dose-tumor response is not always linear. This is clearly illustrated by the costly ED01 study [10] in which a variety of levels of 2-acetylaminofluorene (AAF) were fed to a total of over 23 000 female BALB/C mice to determine the shape of the tumordose response curve to a tumor-response level of 1%. While the curve for liver tumors approximated to
linearity, that for urinary bladder tumors was more akin to an ice-hockey stick with little tumorigenicity at lower exposures to AAF but a steep rise thereafter. The concept of using upper confidence levels of induced tumors presents few problems if most of the animals develop tumors in response to the agent. It represents an attempt to compensate for the relatively few animals used in the bioassay. The result is catastrophic if there is no increase or only a minor increase in tumors. Let us suppose that a bioassay of 5% water in the diet had been performed and there was no increase in the number of tumors. Use of the upper confidence limit to determine a ‘safe’ level of water would lead to a recommended human consumption of water of only a few milliliters/day. As Barnard [2] points out, these default assumptions leave much to be desired. An attempt needs to be made to determine whether mechanistic considerations can help.
2. Carcinogenic mechanisms The seminal observations of Berenblum and Shubik [11–13] provided the first firm experimental evidence that the induction of cancer consisted of two or more stages. They showed that many mainly benign tumors in mouse’s skin could be induced by a single limited administration of a potently active polycyclic aromatic hydrocarbon such as 7,12-dimethylbenz(a)anthracene, followed by repeated painting with croton oil. Neither the hydrocarbon nor the croton oil alone led to the induction of tumors on an appreciable scale. The use of two quite different agents made it clear that two different targets were involved in skin tumor formation; these were named ‘initiation’ and ‘promotion’. As is usual with important observations, Berenblum and Shubik’s work engendered much further study aimed at increasing the number of agents that were initiating or promoting in mouse’s skin, were able to induce similar effects in other tissues, and importantly to demonstrate firmly that initiated cells could lie dormant in their tissues of origin for extended periods. This effort is still continuing 50 years after the initial reports appeared. Growing appreciation of the carcinogenic process has led to the belief that initiation is due to mutations
D.B. Clayson / Cancer Letters 117 (1997) 209–215
in oncogenes and tumor suppressor genes. Cohen and Ellwein [14], amongst others, have suggested that cancer formation is, at its simplest, a cascade of specific mutations interspersed with cellular proliferation to increase the size of the clones of mutated cells. This does not exclude other possibly important factors such as modification of the immunological status of the host, but nevertheless provides a convenient and largely valid model on which to base a newer approach to risk assessment. 2.1. Initiation involving electrophile generation The initial molecularly-based mechanism of carcinogenesis was due to Drs. James and Elizabeth Miller and their colleagues, who investigated the metabolic activation of 2-acetylamino-fluorene (AAF). They first identified a new metabolite, the N-hydroxy derivative, which they demonstrated was a somewhat more potent carcinogen than the parent acetamide in several rodent species. However, the N-hydroxy derivative lacked the ability to interact with DNA in vitro, a property of AAF in the living animal. It was shown that in rat liver, a sulfotransferase enzyme was able to create the sulfate ester of N-hydroxy-2-AAF. This, instead of being more easily excreted than the Nhydroxy derivative, was highly unstable. The sulfate ion acted as a leaving group, liberating the highly reactive 2-acetylimino-fluorene moiety that interacted with a wide variety of cell constituents including DNA [15–18]. This demonstration engendered a great deal of further work that showed N-hydroxylation followed by conjugation by one of a variety of groups was a major route in aromatic amine carcinogenesis. It was also found that other carcinogens such as the polycyclic aromatic hydrocarbons, the nitrosamines, and agents such as the aflatoxins all induced cancer through their conversion to a reactive entity, an electrophile. Certain other carcinogens such as the nitrosamides, the sulfur and nitrogen mustards broke down spontaneously into electrophiles in biological media [19]. 2.2. Initiation involving oxygen (and nitrogen) radicals Radiation-induced cancer has been recognized to
211
be associated with radical formation for many years. Only recently has the possible importance of radical formation in chemically-induced cancer begun to be appreciated [20–22]. Fat metabolism provides a useful example of what is happening. During fat metabolism, a hydroxy radical is produced that is sufficiently stable to diffuse from its site of origin but is sufficiently reactive to interact with DNA. In addition, although hydrogen peroxide does not interact directly with DNA, if it is not degraded by catalase, it may react with transition metal ions including those derived from iron and copper to produce a highly reactive oxygen radical. This radical, if produced in close proximity to the DNA, like the less active radical mentioned previously, may produce one of about 20 different lesions on the DNA. These include single and double strand breaks, mono- and dihyroxylated bases, and so on [23]. Oxygen radical formation differs from electrophile generation in so far as it is a major naturally-occurring and ongoing process. Studies in acellular systems have shown that the most adequately studied oxygen radical-induced lesion, 8-hydroxy-2-deoxyguanosine, leads to mutation on DNA replication [24,25]. Ames and Gold [21] estimated that the DNA from an untreated adult rat liver contained an average of about 106 oxygen radical-induced lesions in each cell genome. Others have reported similarly high levels in other human and animal cells. After 66% partial hepatectomy in rats, the liver returns to near its normal size and functionality, indicating that these cells must be able to protect themselves efficiently from mutational events caused by these lesions (for references see [22]). This implies that the risks associated with these oxygen radical-induced lesions may differ at least quantitatively from those due to electrophile generators and that this factor needs attention in quantitative risk assessment. 2.3. DNA lesions and cellular mutation It should not be imagined that every induced DNA lesion will lead to a cancer mutation. Some lesions will so poison the cell as to lead to cytotoxicity. Others will be faithfully repaired by the appropriate DNA repair enzymes. Mutations may occur as the result of base mispairing if the cell and its DNA replicates before the DNA lesion is repaired.
212
D.B. Clayson / Cancer Letters 117 (1997) 209–215
2.4. Cellular proliferation The importance of cellular proliferation in carcinogenesis has not received the attention it merits because of the amount of labor involved in counting individual cells in tissue slides. Hopefully, this demand on labor will be overcome in the near future by the use of computer-based image analysis. Electrophile-generating and possibly oxygen radical-forming carcinogens induce proliferative cell regeneration in response to cytotoxicity and possibly through their ability to interfere with apoptosis, which is the mechanism of programmed cell death that keeps tissue mass in proportion to the needs of the total organism. In some cases, induction of cellular proliferation appears to be the major effect of a particular carcinogen. This is certainly the case with non-genotoxic agents leading to stone formation in rodent bladders and thence to bladder cancer [7,26]. A naturally occurring or surgically implanted stone may increase the rate of cell turnover many times. Increased cell proliferation has also been reported with a number of rodent liver carcinogens, such as chloroform [27,28]. There is a need to determine the nature of the initiating stimulus before carcinogenesis risk assessment is attempted. Initiation in such cases could occur naturally in a tissue (genetic inheritance), could result from naturally occurring carcinogens or accidental exposure to another carcinogen, or to non-faithful DNA replication. In my view, the former is marginally more likely especially in animals that demonstrate a significant background tumor yield. A further complication in the study of cellular proliferation is that initiated cells at different stages of malignancy may differ in their response to a particular agent. This is well illustrated by the non-genotoxic liver carcinogen, butylated hydroxytoluene (BHT) [29–31]. When administered acutely to rats, BHT induced a relatively minor, time-limited increase in cell turnover. This relapsed to below the levels observed in the untreated liver tissue after 4 days. However, BHT had quite a major effect when administered to rats with enzyme-altered foci in their livers [32,33]. Attempts to derive dose-response relationships for cellular proliferation to the relatively low exposure levels required for carcinogen risk assessment for
humans is seldom attempted because of the labor involved. The derivation of computer-based image analysis is essential for this purpose. 2.5. Ligand formation and DNA interaction In some cases, application of a carcinogen leads to physical interaction with a specific protein and the ligand combining with a specific site on the DNA to modify its expression. This was first demonstrated by Jensen [34] with estrogen. The ligand DNA interaction led to enhanced cellular proliferation in estrogenresponsive tissues such as the mammary gland. Other examples include 2,3,7,8-tetrachloro-para-dibenzodioxin [35] and the peroxisome proliferators [36,37]. Where possible, it is probably better to consider such agents in terms of the more general tissular changes that they engender (e.g. cell proliferation) than to try to assess the risk they pose in terms of either ligand formation or DNA interaction. It has been suggested that agents that act in this way will de facto exhibit a threshold or no effect level because of the need to attain a certain critical concentration of ligand before DNA expression is altered. This, like the additional assumption that all non-genotoxic carcinogens may be thresholded, is not tenable. Estrogen, for example, is continually controlling the rate of cellular proliferation in the mammary gland and in other responsive tissues. Addition of any level of excess estrogen should have the potential to modify tissue behavior. Thresholds require real evidence, not assumptions, before they can be accepted.
3. Carcinogenesis mechanisms and risk assessment The preceding brief account of carcinogenesis mechanisms, which has relied heavily on an initiation/promotion or mutation/cellular proliferation model, makes it abundantly clear that a single format for the assessment of human risk from animal carcinogens will seldom be satisfactory. Reference to Williams [38] shows that, possibly, an even more complex approach will eventually be required than that suggested here. Because of the need to use the maximum tolerated dose (MTD) in rodent bioassays, the carcinogenicity data by itself is not adequate for risk assessment. Because of the relatively imprecise
D.B. Clayson / Cancer Letters 117 (1997) 209–215
methods by which the MTD is estimated, such a dose may be quite toxic to the host tissues leading to the suppression of tumor formation or, alternatively, may exhibit an enhancing or promotional effect. That is, at such high levels, the upper reaches of the dose-carcinogen response curve may be severely distorted making it virtually useless for extrapolation to the estimated levels of human exposure. The first need in assessing the risk associated with a novel carcinogen is to determine whether or not it is genotoxic. At this time positive genotoxicity data is a highly efficient way in which to identify electrophile generators. The same does not appear to be as true for agents leading to oxidative stress. Attempts are currently in progress to devise methodology to recognize agents leading to the formation of oxygen stress [39]. Such methods will require extensive validation before they can be used with confidence. Felton and Turteltaub [40] described an accelerator mass spectrometric method to measure the amount of DNA adducts formed after the administration of low levels of 2-acetylaminofluorene to rats. If this approach can be transferred from the research to the routine laboratory, it promises a far better indication of the probable shape of the dose cancer-response curve for electrophile generators than the presently used prudent but unscientific default assumptions. Such an approach makes the apparently reasonable assumption that DNA adduct formation for electrophile generators is proportional to tumor formation. Such an approach is not yet possible for agents leading to cancer through the induction of oxygen radicals. Methods need to be elaborated and validated. It should not be assumed, in the interim, that these agents will mimic electrophile generators. As already noted oxygen-induced DNA damage is a high level ongoing process in mammals and this may seriously affect the shape of the carcinogen dose response curve. It may be speculated, at this time, that substantial increases in oxygen radical formation will be necessary to overcome the cell’s apparent ability to defend itself against the natural occurrence of these lesions. However, the fact that radiation-induced cancer involves radical formation and no threshold has been suggested for radiation-induced tumors means that such a speculation must be regarded with great caution. The non-genotoxic carcinogens present even
213
greater difficulties because of the labor involved in quite primitive assessments of the dose response curve for induced cellular proliferation. It is anticipated that novel techniques, possibly computerbased image analysis, will be required to obtain meaningful exposure response curves and, where they exist, threshold values for most individual non-genotoxic carcinogens. A further problem arises that while, with the necessary effort, the approaches suggested here will yield useful data in animal studies, similar procedures in humans are unethical. To complete an effective approach to carcinogenesis risk assessment, it will also be necessary to devise methodology, largely in vitro, to determine similar parameters in humans. Without this not inconsiderable effort, carcinogenesis risk assessment will continue to be grossly inadequate and in my opinion a waste of valuable time and effort.
Acknowledgements I wish to thank Dr. Philippe Shubik most warmly for many personal kindnesses during my career including (1) inviting me to join his group in Omaha, Nebraska to continue working in chemical carcinogenesis, (2) explaining to me the nuances of international panel meetings so that I might participate more effectively, and (3) above all, for his cooperation and encouragement in founding and managing Cancer Letters for over 20 years. I must also thank Drs. W.G. Flamm and F. Iverson for reading this manuscript and making helpful comments.
References [1] United States Food and Drug Act, Public Law 85–929, 85th Congress H.R. 13254, September 5th, 1958. [2] R.C. Barnard, Risk assessment. The default assumption controversy, Regul. Toxicol. Pharmacol. 21 (1995) 431–438. [3] D.B. Clayson, F. Iverson, Cancer risk assessment at the crossroads: the need to turn to a biological approach, Regul. Toxicol. Pharmacol. 24 (1996) 45–59. [4] B.G. Short, V.L. Burnett, M.G. Cox, J.S. Bus, J.A. Swenberg, Site-specific renal cytotoxicity and cell proliferation in male rats exposed to petroleum hydrocarbons, Lab. Invest. 57 (1987) 564–577. [5] W.G. Flamm, L.D. Lehmann-McKeeman, The human relevance of the renal tumor-inducing potential of d-limonene in
214
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14] [15]
[16]
[17]
[18]
[19]
[20]
[21]
D.B. Clayson / Cancer Letters 117 (1997) 209–215 male rats: implications for risk assessment, Regul. Toxicol. Pharmacol. 13 (1991) 70–86. S.J. Borghoff, A.B. Miller, J.P. Bowen, J.A. Swenberg, Characteristics of chemical binding to a-2u-globulin in vitro: evaluating structure-activity relationships, Toxicol. Appl. Pharmacol. 107 (1991) 228–238. D.B. Clayson, L. Fishbein, S.M. Cohen, Effects of stones and other physical factors on the induction of rodent bladder cancer, Food Chem. Toxicol. 33 (1995) 771–784. N. Ito, S. Fukushima, A. Hagiwara, M. Shibata, T. Ogiso, Carcinogenicity of butylated hydroxyanisole in F344 rats, J. Natl. Cancer Inst. 70 (1983) 343–352. D.B. Clayson, F. Iverson, E.A. Nera, E. Lok, The importance of cellular proliferation induced by BHA and BHT, in: G.M. Williams, H. Sics, G.T. Baker, J.W. Erdman, C.J. Henry (Eds.), Antioxidants: Chemical, Physiological, Nutritional and Toxicological Aspects, Princetown Scientific, Princetown, NJ, 1992, pp. 231–242. J.A. Staffa, M.A. Mehlman (Eds.), Innovations in Carcinogen Risk Assessment (ED01 study). Pathotox, Park Forest South, IL, 1979. I. Berenblum, P. Shubik, The role of croton oil applications, associated with a single painting of a carcinogen, in tumour induction of the mouse’s skin, Br. J. Cancer 1 (1947) 379– 382. I. Berenblum, P. Shubik, A new, quantitative, approach to the study of the stages of chemical carcinogenesis in the mouse’s skin, Br. J. Cancer 1 (1947) 383–387. I. Berenblum, P. Shubik, The persistence of the latent tumour cells induced in the mouse’s skin by a single application of 9:10-dimethyl-1:2-benzanthracene, Br. J. Cancer 3 (1949) 383–387. S.M. Cohen, I.B. Ellwein, Genetic errors, cell proliferation and carcinogenesis, Cancer Res. 51 (1991) 6493–6505. E.C. Miller, J.A. Miller, H.A. Hartmann, N-Hydroxy-2-acetylaminofluorene: a metabolite of 2-acetylaminofluorene with increased carcinogenic activity in the rat, Cancer Res. 21 (1961) 815–824. E.C. Miller, J.A. Miller, M. Enomoto, The comparative carcinogenicity of 2-acetylaminofluorene and its N-hydroxy metabolite in mice, hamsters and guinea pigs, Cancer Res. 24 (1964) 2018–2031. J.R. DeBaun, E.C. Miller, J.A. Miller, N-Hydroxy-2-acetylaminofluorene sulfotransferase: its probable role in carcinogenesis and in protein-(methion-S-yl) binding in rat liver, Cancer Res. 30 (1970) 577–595. J.A. Miller, Carcinogenesis by chemicals: an overview (G.H.A. Clowes Memorial Lecture), Cancer Res. 30 (1970) 859–876. C.E. Searle (Ed.), Chemical Carcinogens, 2nd ed., ACS Monograph 182, American Chemical Society, Washington, DC, 1984. B. Halliwell, O.I. Aruoma, DNA-damage by oxygen derived species: its mechanism and measurement in mammalian systems, FEBS Lett. 281 (1991) 9–19. B.N. Ames, L.S. Gold, Endogenous mutagens and the causes of aging and cancer, Mutat. Res. 250 (1991) 3–16.
[22] D.B. Clayson, R. Mehta, F. Iverson, Oxidative DNA damage: the effect of certain genotoxic and operationally non-genotoxic carcinogens, Mutat. Res. 317 (1994) 23–42. [23] C. Von Sonntag, The Chemical Basis of Radiation Biology. Taylor and Francis, London, 1987. [24] Y. Kuchino, F. Mori, H. Kasai, H. Inoue, S. Iwai, K. Miura, E. Ohtsuka, S. Nishimura, Misreading of DNA templates containing 8-hydroxydeoxyguanosine at the modified base and at adjacent residues, Nature (London) 327 (1987) 77– 79. [25] R.A. Floyd, The role of 8-hydroxyguanosine in carcinogenesis, Carcinogenesis 11 (1990) 1447–1450. [26] D.B. Clayson, J.A.S. Pringle, The influence of a foreign body on the induction of tumours of the bladder epithelium of the mouse, Br. J. Cancer 20 (1966) 564–568. [27] J.L. Larson, D.C. Wolf, B.E. Butterworth, Induced cytotoxicity and cell proliferation in the hepatocarcinogenicity of chloroform in female B6C3F1 mice: comparison of administration by gavage in corn oil vs. ad libitum in drinking water, Fundam. Appl. Toxicol. 22 (1994) 90–102. [28] M.A. Pereira, Route of administration determines whether chloroform enhances or inhibits cell proliferation in the liver of B6C3F1 mice, Fundam. Appl. Toxicol. 23 (1994) 87–92. [29] P. Olsen, O. Meyer, N. Gille, G. Wurtzen, Carcinogenicity study on butylated hydroxytoluene (BHT) in Wistar rats exposed in utero, Food Chem. Toxicol. 24 (1986) 1– 12. [30] K. Inai, T. Kabuke, S. Nambu, T. Takemoto, E. Kou, H. Kishina, M. Fujihara, S. Yonhara, S. Suehiro, T. Tsuya, K. Horiuchi, S. Tokuoka, Hepatocellular tumorigenicity of butylated hydroxytoluene administered orally to B6C4F1 mice, Jpn. J. Cancer 70 (1988) 49–58. [31] E.M. Bombard, J.N. Bremmer, B.A. Herbold, Review of the mutagenicity/genotoxicity of butylated hydroxytoluene, Mutat. Res. 288 (1992) 187–200. [32] D. Briggs, E. Lok, E.A. Nera, K. Karpinski, D.B. Clayson, Short-term effects of butylated hydroxytoluene on Wistar rat liver, urinary bladder and thyroid gland, Cancer Lett. 46 (1989) 31–36. [33] E. Lok, R. Mehta, P. Jee, G. Laver, E.A. Nera, E. McMullen, D.B. Clayson, The effect of butylated hydroxytoluene on the growth of enzyme-altered foci in male Fischer 344 rat liver tissue, Carcinogenesis 16 (1995) 1071–1078. [34] E.V. Jensen, Estrogen receptors in hormone-dependent breast cancers, Cancer Res. 35 (1975) 2262–3364. [35] A. Poland, E. Glover, A.S. Konde, Stereospecific, high affinity binding of 2,3,7.8-tetrachloro-p-dibenzodioxin by hepatic cytosol. Evidence that the binding species is receptor for induction of arylhydrocarbon hydroxylase, J. Biol. Chem. 231 (1976) 4936–4946. [36] J.D. Tugwood, I. Issemann, R.G. Anderson, K.R. Bundell, W.L. McPheat, S. Green, The mouse peroxisome proliferator activated receptor recognizes a response element in the 5′ flanking sequence of the rat acyl CoA oxidase gene, EMBO J. 11 (1992) 433–439. [37] I. Issermann, S. Green, Activation of a member of the steroid
D.B. Clayson / Cancer Letters 117 (1997) 209–215 hormone receptor superfamily by peroxisome proliferators, Nature (London) 347 (1990) 645–650. [38] G.M. Williams, Chemicals with carcinogenic activity in the rodent liver. Cancer Lett. 117 (1997) this issue. [39] J.J.P. Gille, C.C.M. van Berkel, H. Joenje, Mutagenicity of
215
metabolic oxygen radicals in mammalian cell cultures, Carcinogenesis 15 (1994) 2695–2699. [40] J.S. Felton, K.W. Turteltaub, Accelerator mass spectrometry for measuring low dose carcinogen binding to DNA, Environ. Health Perspect. 102 (1994) 450–452.