Mutagenicity to Salmonella, Drosophila and the mouse bone marrow of the human antineoplastic agent fotemustine: prediction of carcinogenic potency

Combinations of genotoxic tests for the evaluation of group 1 IARC carcinogens

Many of the known human carcinogens are potent genotoxins that are efficiently detected as carcinogens in human populations but certain types of compounds such as immunosuppressants, sex hormones, etc. act via non-genotoxic mechanism. The absence of genotoxicity and the diversity of modes of action of non-genotoxic carcinogens make predicting their carcinogenic potential extremely challenging. There is evidence that combinations of different short-term tests provide a better and efficient prediction of human genotoxic and non-genotoxic carcinogens. The purpose of this study is to summarize the in vivo and in vitro comet assay (CMT) results of group 1 carcinogens selected from the International Agency for Research on Cancer and to discuss the utility of the comet assay along with other genotoxic assays such as Ames, in vivo micronucleus (MN), and in vivo chromosomal aberration (CA) test. Of the 62 agents for which valid genotoxic data were available, 38 of 61 (62.3%) were Ames test positive, 42 of 60 (70%) were in vivo MN test positive and 36 of 45 (80%) were positive for the in vivo CA test. Higher sensitivity was seen in in vivo CMT (90%) and in vitro CMT (86.9%) assay. Combination of two tests has greater sensitivity than individual tests: in vivo MN + in vivo CA (88.6%); in vivo MN + in vivo CMT (92.5%); and in vivo MN + in vitro CMT (95.6%). Combinations of in vivo or in vitro CMT with other tests provided better sensitivity. In vivo CMT in combination with in vivo CA provided the highest sensitivity (96.7%).

CarcinoPred-EL: Novel models for predicting the carcinogenicity of chemicals using molecular fingerprints and ensemble learning methods

Carcinogenicity refers to a highly toxic end point of certain chemicals, and has become an important issue in the drug development process. In this study, three novel ensemble classification models, namely Ensemble SVM, Ensemble RF, and Ensemble XGBoost, were developed to predict carcinogenicity of chemicals using seven types of molecular fingerprints and three machine learning methods based on a dataset containing 1003 diverse compounds with rat carcinogenicity. Among these three models, Ensemble XGBoost is found to be the best, giving an average accuracy of 70.1 ± 2.9%, sensitivity of 67.0 ± 5.0%, and specificity of 73.1 ± 4.4% in five-fold cross-validation and an accuracy of 70.0%, sensitivity of 65.2%, and specificity of 76.5% in external validation. In comparison with some recent methods, the ensemble models outperform some machine learning-based approaches and yield equal accuracy and higher specificity but lower sensitivity than rule-based expert systems. It is also found that the ensemble models could be further improved if more data were available. As an application, the ensemble models are employed to discover potential carcinogens in the DrugBank database. The results indicate that the proposed models are helpful in predicting the carcinogenicity of chemicals. A web server called CarcinoPred-EL has been built for these models (http://ccsipb.lnu.edu.cn/toxicity/CarcinoPred-EL/).

A core in vitro genotoxicity battery comprising the Ames test plus the in vitro micronucleus test is sufficient to detect rodent carcinogens and in vivo genotoxins

In vitro genotoxicity testing needs to include tests in both bacterial and mammalian cells, and be able to detect gene mutations, chromosomal damage and aneuploidy. This may be achieved by a combination of the Ames test (detects gene mutations) and the in vitro micronucleus test (MNvit), since the latter detects both chromosomal aberrations and aneuploidy. In this paper we therefore present an analysis of an existing database of rodent carcinogens and a new database of in vivo genotoxins in terms of the in vitro genotoxicity tests needed to detect their in vivo activity. Published in vitro data from at least one test system (most were from the Ames test) were available for 557 carcinogens and 405 in vivo genotoxins. Because there are fewer publications on the MNvit than for other mammalian cell tests, and because the concordance between the MNvit and the in vitro chromosomal aberration (CAvit) test is so high for clastogenic activity, positive results in the CAvit test were taken as indicative of a positive result in the MNvit where there were no, or only inadequate data for the latter. Also, because Hprt and Tk loci both detect gene-mutation activity, a positive Hprt test was taken as indicative of a mouse-lymphoma Tk assay (MLA)-positive, where there were no data for the latter. Almost all of the 962 rodent carcinogens and in vivo genotoxins were detected by an in vitro battery comprising Ames+MNvit. An additional 11 carcinogens and six in vivo genotoxins would apparently be detected by the MLA, but many of these had not been tested in the MNvit or CAvit tests. Only four chemicals emerge as potentially being more readily detected in MLA than in Ames+MNvit--benzyl acetate, toluene, morphine and thiabendazole--and none of these are convincing cases to argue for the inclusion of the MLA in addition to Ames+MNvit. Thus, there is no convincing evidence that any genotoxic rodent carcinogens or in vivo genotoxins would remain undetected in an in vitro test battery consisting of Ames+MNvit.

Modulation of O6-alkylating agent induced clastogenicity by enhanced DNA repair capacity of bone marrow cells

The murine bone marrow micronucleus assay has been used to examine (1) the potentiation of fotemustine and streptozotocin induced-clastogenicity by the O6-alkylguanine-DNA alkyltransferase (ATase) inactivator O6-benzylguanine (O6-beG) and (2) the level of protection afforded against this potentiation by retrovirus-mediated expression of an O6-beG-resistant mutant of human ATase (haTPA/GA) in mouse bone marrow. Both fotemustine and streptozotocin induced significantly higher levels of micronucleated polychromatic erythrocytes (p < 0.001 for the highest doses studied) compared to those seen in vehicle-treated animals. The number of micronuclei produced by either agent was dramatically elevated by pretreatment with O6-beG (p < 0.001). Furthermore, in myeloablated mice reconstituted with bone marrow expressing the O6-beG-resistant hATPA/GA as a result of retroviral gene transfer, the frequency of micronucleus formation following exposure of mice to otherwise clastogenic doses of fotemustine or streptozotocin, in the presence of O6-beG, wash highly significantly reduced (p < 0.001 for both agents) relative to that in mock transduced controls. These data clearly implicate O6-chloroethyl- and O6-methylguanine as clastogenic lesions in vivo and establish ATase as a major protective mechanism operating to reduce the frequency of such damage. The potentiation of drug induced clastogenicity by O6-beG suggests that the clinical use of this inactivator in combination with O6-alkylating agents, could substantially increase the risk of therapy related malignancy. Nevertheless the use of hATPA/GA as a protective mechanism via gene therapy may overcome this risk.

Heritable and cancer risks of exposures to anticancer drugs: inter-species comparisons of covalent deoxyribonucleic acid-binding agents

In the past years, several methodologies were developed for potency ranking of genotoxic carcinogens and germ cell mutagens. In this paper, we analyzed six sub-classes of covalent deoxyribonucleic acid (DNA) binding antineoplastic drugs comprising a total of 37 chemicals and, in addition, four alkyl-epoxides, using four approaches for the ranking of genotoxic agents on a potency scale: the EPA/IARC genetic activity profile (GAP) database, the ICPEMC agent score system, and the analysis of qualitative and quantitative structure-activity and activity-activity relationships (SARs, AARs) between types of DNA modifications and genotoxic endpoints. Considerations of SARs and AARs focused entirely on in vivo data for mutagenicity in male germ cells (mouse, Drosophila), carcinogenicity (TD50s) and acute toxicity (LD50s) in rodents, whereas the former two approaches combined the entire database on in vivo and in vitro mutagenicity tests. The analysis shows that the understanding and prediction of rank positions of individual genotoxic agents requires information on their mechanism of action. Based on SARs and AARs, the covalent DNA binding antineoplastic drugs can be divided into three categories. Category 1 comprises mono-functional alkylating agents that primarily react with N7 and N3 moieties of purines in DNA. Efficient DNA repair is the major protective mechanism for their low and often not measurable genotoxic effects in repair-competent germ cells, and the need of high exposure doses for tumor induction in rodents. Due to cell type related differences in the efficiency of DNA repair, a strong target cell specificity in various species regarding the potency of these agents for adverse effects is found. Three of the four evaluation systems rank category 1 agents lower than those of the other two categories. Category 2 type mutagens produce O-alkyl adducts in DNA in addition to N-alkyl adducts. In general, certain O-alkyl DNA adducts appear to be slowly repaired, or even not at all, which make this kind of agents potent carcinogens and germ cell mutagens. Especially the inefficient repair of O-alkyl-pyrimidines causes the high mutational response of cells to these agents. Agents of this category give high potency scores in all four expert systems. The major determinant for the high rank positions on any scale of genotoxic of category 3 agents is their ability to induce primarily structural chromosomal changes. These agents are able to cross-link DNA. Their high intrinsic genotoxic potency appears to be related to the number of DNA cross-links per target dose unit they can induce. A confounding factor among category 3 agents is that often the genotoxic endpoints occur close to or at toxic levels, and that the width of the mutagenic dose range, i.e., the dose area between the lowest observed effect level and the LD50, is smaller (usually no more than 1 logarithmic unit) than for chemicals of the other two categories. For all three categories of genotoxic agents, strong correlations are observed between their carcinogenic potency, acute toxicity and germ cell specificity.

Mutagenicity of anticancer drugs in mammalian germ cells

The evidence for mammalian germ cell mutagenicity induced by anticancer drugs is summarized. Primary attention is paid to the three major mouse germ cell mutagenicity tests- the dominant lethal, heritable translocation, and morphological specific locus tests- from which most germ cell mutagenicity data historically have been obtained. Of the 21 anticancer drugs reviewed, 16 have been tested in one or more of these three tests; with all 16 tested in the most common germ cell test, the male dominant lethal test, and 9 of the 16 also tested in the female dominant lethal test. The patterns of germ cell stage specificity for most of the anticancer drugs are similar, and generally resemble the patterns seen with other types of chemicals; however, some of the patterns are unique. For example, 2 of the 8 chemicals shown to induce dominant lethal mutations in female oocytes, do not induce dominant lethal mutations in male germ cells (adriamycin and platinol). Ten of the 16 chemicals tested in the dominant lethal test were positive in post-meiotic stages (spermatids through mature sperm), and seven also induced reciprocal translocations and/or specific locus mutations in post-meiotic stages. This propensity to induce mutations in post-meiotic stages has been observed with most mutagens. However, 5 of the anticancer drugs also induced dominant lethal mutations in spermatocytes (meiotic prophase cells) and one of them, 6-mercaptopurine, uniquely induced dominant lethal mutations exclusively in preleptotene spermatocytes. Finally, three of the anticancer drugs (melphalan, mitomycin C, procarbazine) are members of a very select group of chemicals shown to induce specific locus mutations in spermatogonial stem cells of mice. The implications for human risk are discussed.

DNA damage and repair in mutagenesis and carcinogenesis: implications of structure-activity relationships for cross-species extrapolation

Previous studies on structure-activity relationships (SARs) between types of DNA modifications and tumour incidence revealed linear positive relationships between the log TD50 estimates and s-values for a series of mostly monofunctional alkylating agents. The overall objective of this STEP project was to further elucidate the mechanistic principles underlying these correlations, because detailed knowledge on mechanisms underlying the formation of genotoxic damage is an absolute necessity for establishing guidance values for exposures to genotoxic agents. The analysis included: (1) the re-calculation and further extension of TD50 values in mmol/kg body weight for chemicals carcinogenic in rodents. This part further included the checking up data for Swain-Scott s-values and the use of the covalent binding index (CBI); (2) the elaboration of genetic toxicity including an analysis of induced mutation spectra in specific genes at the DNA level, i.e., the vermilion gene of Drosophila, a plasmid system (pX2 assay) and the HPRT gene in cultured mammalian cells (CHO-9); and (3) the measurement of specific DNA alkylation adducts in animal models (mouse, rat, hamster) and mammalian cells in culture. The analysis of mechanisms controlling the expression of mammalian DNA repair genes (alkyltransferases, glycosylases) as a function of the cell type, differentiation stage, and cellular microenvironment in mammalian cells. The 3 classes of genotoxic carcinogens selected for the project were: (1) chemicals forming monoalkyl adducts upon interaction with DNA; (2) genotoxins capable of forming DNA etheno-adducts; and (3) N-substituted aryl compounds forming covalent adducts at the C8 position of guanine in DNA. In general, clear SARs and AARs (activity-activity relationships) between physiochemical parameters (s-values, O6/N7-alkylguanine ratios, CBI), carcinogenic potency in rodents and several descriptors of genotoxic activity in germ cells (mouse, Drosophila) became apparent when the following descriptors were used: TD50 estimates (lifetime doses expressed in mg/kg b.wt. or mmol/kg b.wt.) from cancer bioassays in rodents; the degree of germ-cell specificity, i.e., the ability of a genotoxic agent to induce mutations in practically all cell stages of the male germ-cell cycle of Drosophila (this project) and the mouse (literature search), as opposed to a more specific response in postmeiotic stages of both species; the Mexr-/Mexr+ hypermutability ratio, determined in a repair assay utilizing Drosophila germ cells; mutation spectra induced at single loci (the 7 loci used in the specific-locus test of the mouse (published data), and the vermilion gene of Drosophila); and doubling doses (DD) in mg/kg (mmol/kg) for specific locus test results on mice. By and large, the TD50 values, the inverse of which can be considered as measures of carcinogenic potency, were shown to be predictable from knowledge of the in vivo doses associated with the absorbed amounts of the investigated alkylators and with the second-order constant, kc, reaction at a critical nucleophilic strength, nc. For alkylating agents kc can be expressed as the second-order rate constant for hydrolysis, kH2O, and the substrate constant s:kH2OTD50 is a function of a certain accumulated degree of alkylation, here given as the (average) daily increment, ac, for 2 years exposure of the rodents. The TD*50 in mmol/kg x day) could then be written: [formula: see text] This expression would be valid for monofunctional alkylators provided the reactive species are uncharged. This is the case for most SN2 reagents. Although it appears possible to predict carcinogenic potency from measured in vivo doses and from detailed knowledge of reaction-kinetic parameter values, it is at present not possible to quantify the uncertainty of such predictions. One main reason for this is the complication due to uneven distribution in the body, with effects on the dose in target tissues. The estimation can be impro

Unexpected genetic toxicity to rodents of the N',N'-dimethyl analogues of MNU and ENU

Lijinsky and his colleagues have reported that the N',N'-dimethyl analogues of ENU and MNU [N',N'-dimethyl-N-ethyl-N-nitrosourea (DMENU) and trimethylnitrosourea (TMNU), respectively] are carcinogenic to rats despite their extreme hydrolytic stability which would reduce or preclude generation of alkylating species analogous to those formed upon hydrolysis of ENU and MNU. Lijinsky and his colleagues were unable to rationalize those activities of DMENU and TMNU despite extensive experimentation. We therefore decided to study this problem further. Whichever mode is accepted for the generation of electrophilic/mutagenic/carcinogenic reactive species from ENU and MNU, blocking of the free-NH2 group with methyl groups (-NMe2) should ablate or abolish activity. Consistent with this DMENU and TMNU gave negative results in the NBP alkylation test while the parent compounds gave an instantaneous deep blue coloration. Studies of the rate of hydrolysis of these four compounds revealed ENU and MNU to have half-lives of 8 min, while the alkylated analogues (DMENU and TMNU) had half-lives of 25 and 41 days, respectively. Hydrolysis of ENU and MNU, to yield the alkylating species, proceeds either via proton abstraction from the -NH2 group or by attack by water on the carbon of the carbonyl group. Methylation will inhibit both of these pathways, the first absolutely (no -NH2 protons) and the second partially, via steric inhibition. The slow hydrolysis observed for DMENU and TMNU suggests that the latter route of hydrolysis is applicable. Studies with strain TA1535 of Salmonella typhimurium (without S9 mix) confirmed the potent mutagenic activity for ENU and MNU (approximately 300-fold increase in revertants at 2,000 micrograms/plate and approximately 180-fold increase in revertants at 150 micrograms/plate respectively). In contrast, the methylated analogues showed only weak mutagenic activity (approximately 3-fold) at approximately 100-fold higher dose-levels. Addition of S9 mix did not affect the mutagenicity of DMENU or TMNU. To this point, hypothesis and data coincide. ENU and MNU are potent micronucleus-inducing agents to the mouse bone marrow, and given the above data, it was expected that DMENU and TMNU would show weak or no activity in that assay. In fact, the methylated analogues were as effective as ENU and MNU as clastogens to the mouse bone marrow. Four possible reasons for this conflict of theory and data are explored. The speculative explanation we favour for these effects is that the net alkylation of bone marrow DNA is the same for all four chemicals. With ENU and MNU, most of the alkylating activity is dissipated by rapid hydrolysis. Thus, only a small fraction of the administered dose survives to alkylate the bone marrow. Due to the enhanced stability of the methyl analogues most of the delivered dose will reach the bone marrow. However, because of their lower intrinsic reactivity, only a small fraction of the target dose will alkylate the bone marrow DNA during the time window of the experiment. If these opposing influences happen to balance out, the essentially identical bone marrow genetic toxicity for the four chemicals could be explained.

Bone marrow micronucleus assay: a review of the mouse stocks used and their published mean spontaneous micronucleus frequencies

We have examined published negative control data from 581 papers on micronucleated bone marrow polychromatic erythrocytes (mnPCE) for differences in mean frequency and the frequency distribution profile among the mouse stocks used with the bone marrow micronucleus assay. For the 55 mouse stocks with published micronucleus assay data, the overall mean frequency is 1.95 mnPCE/1,000 PCE (1.95 mnPCE/1,000); for the 13 stocks most commonly used in the assay, it is 1.88 mnPCE/1,000. During the last 5 years, the mnPCE rate for these 13 major stocks has been 1.74 mnPCE/1,000. This current mean frequency is a substantial decrease from the mean of 3.07 mnPCE/1,000 observed for these 13 stocks for data published prior to 1981. Of the major stocks, the highest mean mnPCE negative control frequencies were observed for MS/Ae > BALB/c > C57Bl/6, and the lowest for CD-1 < Swiss Webster. We note that hybrid mouse stocks appear to have lower and less variable negative control frequencies than either of their parent strains and that the negative control frequency for some progeny stocks have diverged significantly from that of the parent stocks. Overall mean negative control frequencies appear to be correlated with breadth of the frequency distribution profile of published mean negative control values. Furthermore, a possible correlation between negative control frequency in the micronucleus assay and sensitivity to clastogens of different mouse strains may be indicated. The databases generated here allow us to define a range of norms for both the historical mean frequency and individual experimental mean frequencies for most stocks, but in particular, for the more commonly used mouse stocks. Our analysis, for the most part, bears out the recommendation of the first Gene-Tox Report on the micronucleus assay that the historical negative control frequency for a mouse stock should fall between 1 and 3 mnPCE/1,000. Eighty-six percent of the most commonly used mouse stocks have historical mean frequencies within this range. Though individual experimental mean values would not necessarily be expected to fall within the 1-3.00 mnPCE/1,000 range, 65.3% of the 2,327 published negative control values do, and 83.5% are < 3 mnPCE/1,000. The frequency with which an individual experimental mean value lies outside the 1.00 to 3.00 mnPCE/1,000 range differs among stocks and appears related to the mouse mean frequency. We suggest that the recommended range for historical mean frequency be extended slightly, to approximately 3.4 mnPCE/1,000, to accommodate some commonly used strains with overall mean negative control frequencies just above 3.00 mnPCE/1,000.(ABSTRACT TRUNCATED AT 400 WORDS)

Genetic method for pre-classification of genotoxins into monofunctional or cross-linking agents

To characterize environmental carcinogens, there is a need to distinguish monofunctional genotoxic agents from those having cross-linking potential, because chemicals which can cross-link DNA are among the most potent carcinogens in rodents [Barbin and Bartsch, 1989] and humans [Allen et al., 1988; Kaldor et al., 1988]. Here we provide a genetic method for a pre-classification of genotoxins with respect to their functionality--monofunctional versus cross-linking. The procedure is based on the determination of relative clastogenic efficiency by a two-endpoint comparison in Drosophila: (i) induction of chromosome loss (CL), (ii) incidence of recessive lethal mutations (RL). Analysis of CL/RL ratios of 53 genotoxins, all mutagens in Drosophila, permitted distinction of 45 into two major categories: (i) 21 monofunctional agents with CL/RL indices generally < or = 1; (ii) 24 agents with ratios > 2 exhibiting DNA cross-linking properties. Within the group of monofunctional agents, CL/RL ratios tend to be low for SN1 agents, i.e., for N-ethyl-N-nitrosourea, N-ethyl-N'-nitro-N-nitrosoguanidine, and for N-nitrosodiethylamine. With cross-linking agents, the number of reactive groups appeared of minor importance as bi-, tri-, and tetrafunctional agents showed no significant differences in their CL/RL indices. Among 8 chemicals which could not be grouped into one of the two categories are two (adriamycin, daunomycin) regarded as intercalating agents. It is concluded that this two-endpoint analysis in Drosophila has prognostic value and can assist in the characterization of genotoxic agents with unknown mode of action.