Mutagenesis

Mutagenesis Advance Access originally published online on March 16, 2007
Mutagenesis 2007 22(3):161-175; doi:10.1093/mutage/gem006

This Article Abstract FREE Full Text (PDF) OA All Versions of this Article: 22/3/161    most recent gem006v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (2) Google Scholar Articles by Kirkland, D. Articles by Pfuhler, S Search for Related Content PubMed PubMed Citation Articles by Kirkland, D. Articles by Pfuhler, S Social Bookmarking          What's this?

© The Author 2007. Published by Oxford University Press. All rights reserved.
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions@oxfordjournals.org

In vitro approaches to develop weight of evidence (WoE) and mode of action (MoA) discussions with positive in vitro genotoxicity results

DJ Kirkland^*, M Aardema¹, N Banduhn², P Carmichael³, R Fautz⁴, J-R Meunier⁵ and S Pfuhler⁶

Covance Laboratories Limited, Otley Road, Harrogate HG3 1PY, UK ¹The Procter & Gamble Company, Miami Valley Laboratories, PO Box 538707, Cincinnati, OH 45253-8707, USA ²Henkel KGaA, Henkelstrasse 67, D-40191 Düsseldorf, Germany ³Unilever, Safety & Environment Assurance Centre, Sharnbrook MK44 1LQ, UK ⁴Kao Professional Salon Services (KPSS) GmbH, Pfungstaedter Strasse 92-100, D-64297 Darmstadt, Germany ⁵L'Oréal, Avenue Eugène Schuller, F-93600 Aulnay sous Bois Cedex, France ⁶Procter & Gamble, Rte de Chesalles 26, CH-1723 Marly, Switzerland

A recent analysis by Kirkland et al. [Kirkland, D., Aardema, M., Henderson, L. and Müller, L. (2005) Evaluation of the ability of a battery of 3 in vitro genotoxicity tests to discriminate rodent carcinogens and non-carcinogens. I. Sensitivity, specificity and relative predictivity. Mutat. Res. 584, 1–256] demonstrated an extremely high false positive rate for in vitro genotoxicity tests when compared with carcinogenicity in rodents. In many industries, decisions have to be made on the safety of new substances, and health risk to humans, without rodent carcinogenicity data being available. In such cases, the usual way to determine whether a positive in vitro genotoxicity result is relevant (i.e. indicates a hazard) for humans is to develop weight of evidence (WoE) or mode of action (MoA) arguments. These are based partly on further in vitro investigations, but usually rely heavily on tests for genotoxicity in one or more in vivo assays. However, for certain product types in the European Union, the use of animals for genotoxicity testing (as well as for other endpoints) will be prohibited within the next few years. Many different examples have been described that indicate DNA damage and genotoxic responses in vitro can arise through non-relevant in vitro events that are a result of the test systems and conditions used. The majority of these non-relevant in vitro events can be grouped under a category of ‘overload of normal physiology’ that would not be expected to occur in exposed humans. However, obtaining evidence in support of such MoAs is not easy, particularly for those industries prohibited from carrying out in vivo testing. It will become necessary to focus on in vitro studies to provide evidence of non-DNA, threshold or in vitro-specific processes and to discuss the potential for such genotoxic effects to occur in exposed humans. Toward this end, we surveyed the published literature for in vitro approaches that may be followed to determine whether a genotoxic effect observed in vitro will occur in humans. Unfortunately, many of the approaches we found are based on only a few published examples and validated approaches with consensus recommendations often do not exist. This analysis highlights the urgent need for developing consensus approaches that do not rely on animal studies for dealing with in vitro genotoxins.

	Introduction

Top Introduction Processes and properties... Bacterial mutation tests Mammalian cell tests Conclusions References

 
The safety assessment for most new chemical substances includes the requirement for an assessment of genotoxic potential. The following are examples of recent published guidance:

(i) International Conference on Harmonization guidances for testing of pharmaceuticals (2,3),

(ii) EU Technical Guidance Document for testing of industrial chemicals (4),

(iii) German BfR overview of strategies for testing of industrial chemicals (5),

(iv) UK Committee on Mutagenicity Guideline for testing of chemicals (6),

(v) Food and Drug Administration (FDA) Redbook (7),

(vi) Updated Recommended Strategy for Testing Oxidative Hair Dye Substances for their Potential Mutagenicity/Genotoxicity (8),

(vii) Recommended Mutagenicity/Genotoxicity Tests for the Safety Testing of Cosmetic Ingredients to be Included in the Annexes to Council Directive 76/768/EEC (9) and

(viii) FDA Guidance for Industry. Recommended Approaches to Integration of Genetic Toxicology Study Results (10).

No single test system is able to detect all genotoxic agents, and therefore a ‘battery’ of tests is usually required, initially in vitro with, in many cases, follow-up testing in vivo.

The most commonly used in vitro tests are

(i) a test for the induction of gene mutations in bacteria,

(ii) an in vitro mouse lymphoma tk locus assay and

(iii) an in vitro test for chromosome damaging potential in mammalian cells (metaphase analysis or, more recently, micronucleus test).

The sensitivity of genotoxicity tests has improved over the years such that it is now widely accepted that the chances of failing to detect genotoxic activity in a standard or extended battery of tests [e.g. as described in (3)] are negligible. However, a number of recent analyses (1,11,12) demonstrated an extremely high false positive rate for in vitro genotoxicity tests (in particular, mammalian cell tests) when compared with rodent carcinogenicity.

In most industries, decisions are made on the safety of new substances, and human health hazards have to be identified, without rodent carcinogenicity data being available. In such cases, the usual approach for determining whether a positive in vitro genotoxicity result indicates a hazard for humans is firstly to conduct in vivo genotoxicity assays to establish whether the genotoxic potential seen in vitro is manifested in vivo. The most widely used in vivo test is for chromosome damage using rodent erythropoietic cells (micronucleus test in bone marrow or blood or metaphase analysis of bone marrow cells). Other in vivo genotoxicity tests [e.g. DNA repair (unscheduled DNA synthesis, UDS), mutation in transgenic animals, tests for measurement of DNA adducts, DNA strand breaks including comet assay] may also be used. In vivo genotoxicity assays typically are negative but are often not sufficient by themselves to outweigh positive in vitro data. Additional supporting data, including evidence that the test chemical and/or metabolites reaches the in vivo target tissue, add strength to negative in vivo results, but add considerable cost and sometimes animal usage.

In the European Union (EU), for cosmetic products and their ingredients, in vivo testing will be prohibited for all endpoints of acute toxicity (including genotoxicity) by 2009. In the absence of in vivo testing, it will become more difficult to develop convincing weight of evidence (WoE) arguments or to determine whether or not positive in vitro genotoxicity results are relevant for human exposures. Thus, it is important to consider the likely reasons why in vitro positives may not be indicative of a human hazard under normal conditions of exposure. In 1991, an International Commission for Protection against Environmental Mutagens and Carcinogens (ICPEMC) Task Force (13) described many indirect mechanisms that could result in genotoxicity and that such mechanisms would be expected to exhibit no-effect concentrations (NOECs) below which genotoxicity would not be induced. In recent times, there have been many more publications that have discussed both indirect mechanisms for genotoxins and also direct DNA-damaging mechanisms that may have a threshold (14–23). The likely reasons an in vitro effect may not occur in humans are summarized broadly as

(i) chemicals may exert their in vitro genotoxic effects via primary damage to a non-DNA target, and as such there would be an NOEC below which there would be no damage either to cellular target or DNA. Often such NOECs are not reached in vivo, and thus, the in vitro genotoxicity is not predictive of an in vivo (or human) hazard;

(ii) chemicals may induce damage by a process that is specific to the in vitro test system or conditions. Such a process will never occur in, and is irrelevant for, exposed humans and

(iii) chemicals (or their metabolites) may induce direct damage to DNA, but only at certain concentrations above a threshold defined, for example, by detoxification or other protective processes such as those that occur at extreme or non-physiological conditions. Often such conditions/thresholds are not reached in vivo, and thus, the in vitro genotoxicity is not predictive of an in vivo (or human) hazard.

Taking into consideration the above broad categories of mode of action (MoA), we searched the literature for specific examples and experimental approaches that can be undertaken in vitro to determine whether a positive genotoxicity result is occurring by one of these MoAs. Overall, our review indicates that non-DNA or threshold mechanisms are not easy to prove. Further, the NOEC or threshold concentration may change from system to system and from in vitro to in vivo. Thus, trying to establish safe levels of exposure in humans from in vitro data will be inaccurate. It should also be borne in mind that positive genotoxicity results may be the consequence of more than one MoA, and evidence of an apparent NOEC or threshold may obscure other DNA-damaging mechanisms.

Determination of whether a non-DNA, threshold or in vitro-specific MoA is involved may be enhanced by investigating whether interaction with DNA is involved, and if so at what concentrations. A chemical suspected of exerting its genotoxicity via interaction with a non-DNA target may produce reactive oxygen species (ROS) that could damage DNA (see later for Discussion) but should not be able to induce adducts in DNA. Alternatively it may intercalate, and therefore undergo a non-covalent interaction with DNA (24), which would also not lead to covalently bound adducts. Evidence supporting a non-DNA or in vitro-specific mechanism may be obtained by demonstrating the inability of a chemical to produce DNA adducts in the cells/bacteria where the genotoxic effects are seen. ¹⁴C- or ³H-labelled chemical with specific radioactivity as high as possible should be used, but this is expensive and may not give the level of sensitivity needed. Treatments should replicate those inducing the genotoxicity, but this may not always be possible, e.g. if a 24-h continuous treatment in the absence of S9 is required for a clastogenic positive, the cells may not tolerate such long exposures to high specificity radioactive chemical. If S9 was required for the genotoxic effect then it must be used in the adduct study, and concentrations that were genotoxic must be employed. Demonstrating absence of an effect is always difficult, but when the assay is made as sensitive as possible, and conditions that resulted in genotoxicity are replicated, then this can provide useful additional evidence. Some guidance on the parameters to follow can be found in Phillips et al. (25).

A more sensitive method to look for DNA adducts is ³²P post-labeling. However, this method is most effective in the detection of bulky adducts. If the adduct may not have been characterized previously, then method development will be required in order to select appropriate solvents for chromatography. Inevitably, when no adduct is predicted (because a non-DNA target is suspected), it is impossible to know that all relevant DNA digestion and solvent combinations have been used. Also, as ³²P post-labeling is (at least initially) a qualitative technique, providing convincing evidence of the absence of ‘spots’ on a thin layer chromatogram is not easy, and the absence of DNA adducts is not definitive proof that no direct interaction with or modification of DNA has occurred.

For convenience, the following discussions are organized according to test type, but the different MoAs are summarized according to type in Table I.

View this table: [in this window] [in a new window]   Table I. Summary of indirect, non-relevant or threshold mechanisms of genotoxicity

 

	Processes and properties applicable to all in vitro tests

Top Introduction Processes and properties... Bacterial mutation tests Mammalian cell tests Conclusions References

 
Rat S9-specific or enhanced effects
Although different carcinogens are activated by different cytochrome P450 (CYP) and non-CYP enzymes, there is almost universal use of a single metabolic activation system (liver S9 induced either by pre-treatment of rats with Aroclor 1254 or with phenobarbital/ß-naphthoflavone) for all in vitro genotoxicity tests. Metabolites produced by this S9 may be quite different from those produced by normal human liver metabolism. The induction by Aroclor 1254 leads to overrepresentation of the CYP 1A and 2B enzymes compared to other hepatic CYP forms, as shown in Table II.

View this table: [in this window] [in a new window]   Table II. The impact of Aroclor on the induction of various CYPs in comparison to normal rat and human livera (data kindly provided by H. R. Glatt, Potsdam)

 
If the rat liver S9 included in the standard Ames and mammalian cell tests is able to produce a metabolite that is not produced in humans, then a positive response with a chemical in these in vitro tests in the presence of S9 would be irrelevant for humans. If a rat-specific metabolite is suspected, one approach may be to compare human S9 with rat S9. Beaune et al. (27) compared rat and human S9 in the Ames test. Human S9 gave greater mutagenic responses than induced or uninduced rat S9 with 2-aminoanthracene and 2-aminofluorene, but the reverse was seen with aflatoxin-B1 and 3-methylcholanthrene. They concluded that ‘results with rat liver bioactivation cannot automatically be extrapolated systematically to humans’. Hakura et al. (28) also compared induced and uninduced rat liver S9 with human S9 in the Ames test. For most of the chemicals tested, there were some quantitative but no qualitative differences between four different human S9 samples and induced or uninduced rat S9. Of the four different human preparations, however, one had significantly higher P450 enzyme activity than the others did. Only this sample of human S9 gave a positive response with benzo[a]pyrene; the other three samples were negative or equivocal. The consistency of rat liver S9 comes mainly from the in-breeding of the rats used for its production. By contrast, human S9 must be expected to show wide inter-individual variability. Although pooled human S9 can be obtained, it is not likely to provide the consistency of response to standard mutagens as rat liver S9. Thus, although human S9 can be useful, it is not likely to produce reliable and consistent responses.

Another approach would be to test in cell lines genetically engineered to express normal rat or human P450 activity such as genetically engineered Chinese hamster V79 cells (29,30) or ER-181 cells (31), human AHH-1 or MCL-5 cells (32,33) and THLE cells (34).

The metabolic aspects of genotoxicity testing have recently been discussed by a workgroup of the International Workshops on Genotoxicity Tests (35). Although the emphasis was to ensure that false negatives are avoided (i.e. that human metabolites not formed by rodents are adequately tested), some useful advice on alternative metabolism systems is included therein.

Phase 2 enzymes are essentially inactive in standard treatments with S9, as their cofactors are not added, unlike NADPH, the cofactor for CYPs. Yoshikawa et al. (36) showed that inactivation of styrene oxide as a bacterial mutagen was effected by glutathione S-transferase in the soluble part of the S9 fraction, but this was demonstrated by addition of glutathione to the incubations. Ordinarily, cofactors such as (i) glutathione, (ii) acetyl CoA (for acetylation), (iii) 3'-phosphoadenosine 5'-phosphosulphate (for sulphation) and (iv) uridine diphosphate glucuronic acid (for glucuronidation) are not added to the S9 mix. Although some will be present in liver preparations such as S9, detoxification by these processes will be limited. We may therefore envisage a threshold of detoxification in an in vitro system using S9 mix, which could be influenced by the addition of the above cofactors to the test system. Exceeding a detoxification threshold with S9 mix could produce some misleading results. For example Glatt et al. (37) demonstrated that activation of methylbenz[a]anthracenes to mutagens by intact hepatocytes correlated much better with carcinogenic potency than did activation by S9. In some situations, we may envisage that a positive result may arise in the presence of induced S9 mix because there is insufficient detoxification capability which, if Phase 2 cofactors were added, would not occur. There are certainly several examples in the literature of genotoxicity in vitro being dramatically inhibited by addition of Phase 2 cofactors (36,38), and one approach to analysing this mechanism may be to add various Phase 2 cofactors to the in vitro incubations. In vivo, where the Phase 1 and Phase 2 activities are more balanced, such chemicals may well produce negative genotoxicity results as the detoxification threshold is not breached.

DNA repair capability
The Salmonella and Escherichia coli bacteria used in the Ames test contain a variety of repair deficiencies. The commonly used Salmonella strains (except TA102) carry an uvrB mutation, which contains a deletion of a gene coding for a DNA excision repair system. Salmonella strains TA97, TA98, TA100 and TA102, and the most commonly used E. coli strains carry the pKM101 plasmid which enhances the error-prone DNA repair system in these organisms. These confer increased sensitivity to DNA-reactive chemicals and metabolites.

CHO-K1 and CHO-WBL cells, as are frequently used for in vitro chromosomal aberration or micronucleus studies, have a mutant p53 sequence, mutant protein and lack the G₁ checkpoint. CHL cells have a wild-type p53 DNA sequence, but expression of p53 protein is not regulated normally (39). L5178Y cells, as used in the mouse lymphoma tk mutation assay, also have a dysfunctional p53 protein (40), and V79 cells also lack a functional p53 protein (41). Such deficiencies mean that cells with damaged DNA (as a result of chemical treatment) will not stop at certain cell cycle checkpoints while the damage is repaired, and this leads to increased genomic instability. TK6 cells, which have functional p53, have shown similar tendency to CHO-WBL cells to give positive results with non-mutagenic chemicals and metabolic poisons (42). Several unpublished sources, and now a published paper (43), have indicated that analysis of chromosomal aberrations or micronuclei in human lymphocytes yields fewer positive results (and by implication fewer irrelevant positive results) than are seen in the common rodent cell lines. This may be explained in part by the fact that levels of p53 increase when lymphocytes begin to cycle after phytohaemagglutinin stimulation (44,45).

The above situations suggest that the levels and efficiency of DNA repair in vivo are likely to be higher than in many of the in vitro systems we use. Thus, cells in normal human tissues should be more effective at repairing DNA damage than can be achieved in bacteria or established cell lines in vitro. Based on this, there are clearly opportunities for genotoxic effects to be induced by chemicals in vitro that will not occur in humans in vivo. However, it is not easy to envisage experimental methods that could demonstrate the relevance (or lack of relevance) of an in vitro positive result based on DNA repair capability. Normal DNA repair capacity would be expected in primary cells, but apart from blood cells, animals would have to be sacrificed in order to provide the cultures.

Early passage cell lines may also preserve sufficient DNA repair and p53 functionality to avoid irrelevant positive results, but comparisons with established rodent cell lines have not been done. For the future, it may be desirable to develop and use new in vitro systems that demonstrate acceptable sensitivity to detect DNA-reactive, mutagenic carcinogens and in vivo genotoxins, but which have normal DNA repair and p53 functions so reducing the risk of irrelevant positives.

Instability of cells and culture conditions
The stability of the cells being used for in vitro genotoxicity testing is an important factor. The instability of rodent cell lines and its contribution to ‘spurious’ positive results have been questioned in the past (46), but no recent evaluations have been published. There are anecdotal stories that rodent cell lines (e.g. CHO, V79, CHL, L5178Y) in one laboratory are not the same as in another. Even different isolates of the same clone, with the same modal chromosome number, can have different plating efficiencies and different responses to genotoxins. Thus, it is highly likely that some cell lines, or isolates of cell lines, may be less able to detoxify chemicals or neutralize the effects of free radicals and ROS than others. It is also possible that some isolates may be less genomically stable than others and more susceptible to increased chromosomal aberrations in the presence of damage to a non-DNA target. It would be prudent to avoid the risk of such increased susceptibilities in the first place, but until side-by-side comparisons on different cell clones and isolates are conducted and published there are no data to guide selection. In the meantime, good cell culture practice would suggest it is wise to use cells of low passage and to check for absence of mycoplasma.

Halliwell et al. (2003) have published data indicating the potential of cell culture media to oxidize a wide range of chemicals (including flavonoids and thiols) to produce hydrogen peroxide (47,48). Hydrogen peroxide is a clastogen, and therefore high levels of peroxide produced as a result of oxidation by media could lead to chromosomal aberrations and small colony mouse lymphoma mutants. These responses would not be the result of the test chemical interacting with the cell, biotransformation by the cell or metabolism by the S9, but would be the result of generation of ROS through oxidation of the test compound by the culture medium. It is possible that the same chemicals may generate ROS in the aqueous environment of human tissues, but the p53- and DNA repair-competent cells there will be less susceptible to ROS-induced damage than rodent cell lines in vitro. Thus, positive genotoxic findings from ROS generated merely from reaction between culture medium and test chemical would overestimate the human hazard.

	Bacterial mutation tests

Top Introduction Processes and properties... Bacterial mutation tests Mammalian cell tests Conclusions References

 
Several analyses of correlations between results in the bacterial mutation test and rodent carcinogenicity (1,49,50) have shown that, of all the individual in vitro tests, the Ames test shows the lowest sensitivity in terms of predicting rodent carcinogenicity. This is not surprising considering the numbers of carcinogens that require the structure and function of a eukaryotic cell in order to demonstrate that activity. By contrast, the Ames test shows the highest specificity, i.e. the lowest frequency of false positive results with non-carcinogens. Thus, there are fewer examples of chemicals giving positive results in the Ames test via non-DNA, in vitro-specific or threshold MoAs than have been postulated for mammalian cell tests.

The following are processes that can indirectly lead to genotoxic responses in mammalian cells, but for which there is no evidence of similar effects in bacteria:

(i) low pH and high osmolality were shown not to result in increased mutation in Ames bacteria (51) and

(ii) high levels of cell killing in bacteria do not appear to actually lead to significant increases in mutation, although artefacts can arise. Highly toxic chemicals can lead to severe loss of background lawn, and the few surviving bacteria can multiply more times before exhausting the available histidine on the plates. Instead of a lawn of microcolonies, one sees a smaller number of discrete and visible colonies that resemble mutant colonies. However, if they are transferred to a histidine-free plate (by replica plating), they will not grow. They are sometimes therefore called ‘pseudo-mutants’. An alternative scenario is that as the few surviving bacteria undergo more divisions in the presence of the available histidine, there will be more spontaneous mutations (as each division has a defined probability of mutation) thus leading to a slightly higher frequency of spontaneous mutants. However, experience indicates this will usually be <2-fold.

Feeding effects
If the chemical substance being tested is able to release histidine into the culture medium (e.g. from a protein, peptide or extract from a living organism such as a plant), then the Salmonella bacteria on the treated plates will be able to undergo more divisions than on a control plate before exhausting the available histidine. As each division has a defined probability of mutation, the numbers of spontaneous mutants on treated plates (containing histidine released from test substance) will be higher than on control plates with limited histidine. This can look like a positive response (51,52). However, at the same time as the number of spontaneous mutants is increasing, the density of the background lawn also increases. In practice, therefore, the spontaneous mutants only increase 2-fold in the presence of additional histidine before the background lawn becomes so confluent that the mutant colonies cannot be distinguished.

This type of feeding effect is only of concern for complex mixtures or extracts that contain histidine, and will not be an issue when testing pure chemicals. However, it is clearly a physiological mechanism relevant for the auxotrophic Salmonella typhimurium bacteria used in the Ames test, and an increase in histidine levels in humans exposed to the test substance will not lead to genotoxicity. Genotoxicity seen only in bacteria as a result of this ‘feeding’ effect does not therefore signify a hazard for humans.

Interestingly, in our experience, increasing the amount of tryptophan on E. coli plates does not result in any notable increase in spontaneous mutants. It seems that the background lawn of E. coli becomes confluent before any increased occurrence of spontaneous mutants is noted.

Specific or enhanced metabolism in bacteria
Sodium azide is converted to azidoalanine by the enzyme O-acetylserine(thio)lyase in bacteria and is mutagenic in these organisms. In fact, sodium azide is one of the standard positive controls in Salmonella strains TA100 and TA1535 tested in the absence of metabolic activation. Mammalian cells express the enzyme, but azidoalanine has not been detected and so sodium azide was not mutagenic in mammalian cells (53,54). It also did not induce UDS in HeLa cells (55), DNA damage in Chinese hamster or human cells (54) or chromosomal aberrations in CHO cells, although an increase in sister chromatid exchanges (SCE) was reported (see http://ntp-server.niehs.nih.gov for results of US National Toxicology Program studies). It is not known whether O-acetylserine(thio)lyase is non-functional in mammalian cells or whether azidoalanine is formed but is rapidly converted to a non-genotoxic intermediate. No published data on in vivo mutagenicity, micronuclei or chromosomal aberrations could be found for sodium azide. However, it was not carcinogenic for F344 rats in a 2-year study conducted by the US National Toxicology Program (56). Based on this analysis, although sodium azide is a good mutagen in the Ames test, it would not appear to be a likely genotoxin in rodents in vivo or in humans. The mutagenicity of sodium azide in the Ames test therefore appears to be irrelevant for humans. If there are other chemicals that have been identified as bacterial-specific mutagens, it would be useful to see the data published.

Salmonella and E. coli bacteria used in the Ames test are very efficient at nitroreduction. Mammalian cells do possess nitroreductases but these are usually inhibited by oxygen, and under normal conditions of mammalian cell culture, or in normal oxygenated tissues in vivo, nitroreduction of a chemical is likely to be inefficient or even absent. For example, nitroreduction of 1-nitropyrene is inhibited by oxygen at representative intracellular concentrations (57). Thus, positive Ames test results with an aromatic nitro compound may not be predictive of genotoxicity in mammalian systems. The novel anti-epileptic drug AMP937, which has an aromatic nitro group, was shown to be clearly mutagenic in several Ames strains. However, it was devoid of genotoxic activity after oral dosing in a series of rigorous in vivo studies including DNA binding, transgenic mutation and comet assay in the gastrointestinal (GI) tract where bacterial nitroreduction would be expected to have maximum effect (58). Thus, even when such a chemical can reach locations (e.g. the GI tract) where anaerobic nitroreduction can occur, it may not result in genotoxicity. For substances administered topically (such as cosmetics) to humans where systemic exposure is low, and enterohepatic recirculation into the GI tract is therefore negligible, the chances of activation by bacterial nitroreduction are very small.

Not all aromatic nitro compounds are activated by nitroreduction, and care must be taken in interpreting Ames test results with this class of chemicals. For example, 1-nitropyrene is activated not only by nitroreduction under anaerobic conditions but also by ring oxidation under aerobic conditions. Consequently, 1-nitropyrene is mutagenic in mammalian cells under both anaerobic and aerobic (S9) conditions (59).

Bacteria are also efficient at azoreduction; however, the situation is not the same as with nitroreduction. At least two types of cytochrome P450 catalyse the microsomal reduction of azo dyes to dimethylaminoazobenzene. The P450s are not identified by name but by inducers and MoA. The first type is induced by clofibrate and is not inhibited by oxygen or carbon monoxide. This involves dyes with only electron-donating substituents. The second type is induced by phenobarbital, ß-naphthoflavone, isosafrole, pregnenolone-16--carbonitrile and clofibrate and is inhibited by oxygen and carbon monoxide. This involves dyes with electron-donating and -withdrawing substituents (60). Thus, azo dyes of the second type may be mutagenic in bacteria but not in aerobic conditions in mammalian cells and tissues. However, azoreduction can occur in cytosol and endoplasmic reticulum of skin cells from various species (61), although the actual enzymes responsible were not identified. As these authors comment, such ‘extensive azoreduction observed during percutaneous absorption may modulate the toxicities of these compounds’, and such modulation could therefore possibly lead to production of genotoxic species in the skin of humans. Thus, it is likely that a positive Ames test resulting from azoreduction may well be predictive of a human hazard.

	Mammalian cell tests

Top Introduction Processes and properties... Bacterial mutation tests Mammalian cell tests Conclusions References

 
Genotoxic responses in mammalian cells resulting from changes in pH and osmolality have been well documented and are only mentioned briefly in this commentary. There are many other situations which may lead to genotoxicity in mammalian cells in vitro either because of interaction with a non-DNA target or because there is a threshold mechanism underlying a DNA-reactive genotoxin. In such cases, if human exposures are below the levels that lead to DNA damage, there will be no genotoxic hazard for humans.

A large review of genotoxicity under extreme conditions (particularly in mammalian cells) was conducted by an ICPEMC Task Group (13). This group identified a number of mechanisms that could indirectly lead to genotoxicity (particularly chromosomal damage, but also mouse lymphoma tk mutations) in mammalian cells. These are summarized in Table III and discussed below.

View this table: [in this window] [in a new window]   Table III. Some indirect mechanisms of genotoxicity from Scott et al. (13)

 
Any of the extreme conditions discussed below can just as easily occur in exposed humans, as in cultured cells in vitro, if sufficient exposures can be achieved. For indirect genotoxicity resulting from such extreme conditions, there will be NOECs below which the cellular conditions are sufficiently normal that no genotoxicity results. NOECs may vary with cell type and endpoint studied. If, for a given chemical that induces genotoxicity in vitro through an indirect mechanism, the NOEC is determined based on a sensitive cell type and endpoint, by use of appropriate safety factors, it may be possible to define safe exposures in humans. It may also be possible to define safe exposures in humans even if the MoA has not been established. Recent advice from the European Committee for Medicinal Products for Human Use (62) allows that, for new drug substances, humans can be exposed to genotoxic impurities, for which no threshold mechanism has been identified, at a level of 1.5 µg/day. The recommendation is based on the concept of the threshold of toxicological concern.

Where animal testing is permitted, the usual process to investigate a genotoxic agent suspected of acting via a non-DNA or threshold DNA MoA would be

(i) obtain evidence in vitro that an indirect (non-DNA) or threshold DNA MoA is most likely responsible for the genotoxicity (see Discussions below);

(ii) obtain results from rigorous in vivo tests in relevant tissues and measuring appropriate genotoxic endpoints. If positive, then obtain evidence that the same non-DNA or threshold DNA MoA applies and

(iii) determine the maximum exposures in vivo at which no genotoxic responses are induced, compare those exposures with anticipated human exposures and apply appropriate safety factors.

If, as will be the case for cosmetic ingredients in Europe, the relevant in vivo tests, and measures of exposure, are not permitted, establishing human safety will be much more challenging. It should still be worthwhile to seek evidence of indirect genotoxicity, and, in the most sensitive systems, to determine the NOEC. Extrapolating from these NOECs to safe human exposures may be expected to involve large safety factors, and there is currently no guidance as to what might be appropriate.

What follows are examples of disruption of normal cellular processes that the ICPEMC task force (13) identified as likely to lead indirectly to genotoxicity. In each case, we discuss how robust is the evidence that such disruption does lead indirectly to genotoxicity, how many examples have been published and, if a novel chemical is thought to exert its genotoxicity by such a disruptive process, what experimental approaches can be taken to obtain evidence. However good the evidence for a non-DNA or threshold DNA MoA is, the possibility of a ‘second mechanism’ of genotoxicity involving direct damage to DNA with no threshold must be considered.

Enzyme inhibition
Inhibition of DNA polymerases. The process of induction of chromosomal damage by inhibitors of DNA polymerases has been hypothesized as follows:

(i) if DNA repair replication during excision repair is inhibited, DNA strand breaks are accumulated. These lead to chromatid and chromosome gaps and deletions at the first mitosis and

(ii) the broken strands interact and become rearranged in the G1 phase to appear as dicentrics or rings at the second mitosis.

Several substances have been described that inhibit DNA polymerase (63) and induce genotoxicity, e.g

(i) cytosine arabinoside inhibits a variety of DNA- and RNA-related enzymes, although it primarily inhibits DNA synthesis (63,64). It induces chromosomal aberrations in cultured cells (65,66) and mutations in mouse lymphoma cells (67) in vitro and micronuclei in vivo (68). Its inhibition of DNA synthesis can be reversed by addition of deoxycytidine (63);

(ii) neocarzinostatin induced micronuclei in cultured cells (69) and was a weak inducer of DNA repair in bacteria (70);

(iii) phosphonoacetic acid is an inhibitor of polymerase , but the reports on its genotoxicity are contradictory. A report by Zhang et al. (71) indicated it could induce Ames mutation, SCE in CHO cells and micronuclei in mouse bone marrow, but there is also a report of failure of the disodium salt of phosphonoacetic acid to induce mutation in Ames bacteria or mouse lymphoma cells (72). However, these papers do not contain sufficient detail for thorough checking (one paper in Chinese, the other an abstract) and

(iv) aphidicolin is also an inhibitor of polymerase , and as well as enhancing the genotoxicity of other agents, can itself induce chromosomal aberrations in vitro (64,73,74).

For a novel chemical, we might speculate that predominant induction of gaps and deletions at the first mitosis, induction of endoreduplicated cells, cell cycle delay and sensitivity in the G2 phase of the cell cycle, which are typical of substances like aphidicolin (75), might indicate inhibition of DNA synthesis or repair as an indirect mechanism. We would also expect to see disruption of cell cycle kinetics with probable accumulation of S-phase cells. Supporting evidence could be obtained from assays to measure inhibition of DNA polymerase (76) and polymerase ß (77). Also, use of the alkaline elution method (78) might show that smaller fragments of cellular DNA accumulate with time after treatment. The comet assay may be helpful in providing supportive evidence, but is not likely to provide such useful quantitative data on DNA size fragments as the alkaline elution assay. Finally, modern methods such as screening for inhibition of replication protein A have been described and may be useful (79).

Inhibition of gyrases/topoisomerases. Some substances that kill bacteria by inhibition of DNA gyrase can also cause other changes both in bacteria and mammalian cells. Novobiocin, which is an inhibitor of gyrase subunit B (80), induced mutation in the standard Ames test and the VITOTOX assay (81) and activated poly(ADP-ribose)polymerase in chick embryo cells (82). Many fluoroquinolones that have been developed to treat bacterial infections by inhibition of gyrase function also inhibit mammalian topoisomerases and can induce genotoxicity in mammalian cells as a result (15,83).

Topoisomerases are enzymes that catalyse the transient breakage and rejoining of either one [topoisomerase (topo) I] or two (topo II) DNA strands, to allow one strand to pass through another and prevent unresolvable tangles during DNA replication. Inhibitors of topo I or II do not bind to DNA but seem to increase the number of strand breaks or prevent their resealing. There is also evidence that inhibitors of topo II may be able to intercalate DNA (84). These activities do not tend to produce point mutations but lead to chromosomal aberrations, recombination and aneuploidy. There are many publications on the genotoxicity of topoisomerase inhibitors [for reviews, see (85,86)], some of which are very potent, but, as the target is an enzyme, each is expected to have a threshold and evidence of this has been published by Lynch et al. (87).

If a novel chemical is suspected of causing genotoxicity by inhibition of mammalian topoisomerase, then direct evidence of binding to topo I or II should be obtained. It will then be important to distinguish whether inhibition of mammalian topoisomerase, and therefore genotoxicity, occurs at low concentrations [e.g. the anti-cancer drugs etoposide, amsacrine (m-AMSA)] or whether it occurs at sufficiently high concentrations (e.g. as with the fluoroquinolone antibiotics) that a margin of safety can be estimated. Absence of induction of DNA adducts, and evidence of intercalation (88) would provide useful additional support of an indirect mechanism of genotoxicity, but, as discussed earlier, absence of DNA adducts is difficult to determine.

Inhibition of kinases. Kinases are a new area of research for the pharmaceutical industry. Thus, there are few publications on inhibition of kinases and genotoxicity. CK2 was the first protein kinase identified and has been found to facilitate the repair of chromosomal DNA single-strand breaks (89). Inhibition of CK2 by chemical treatment would be expected to result in chromosomal aberrations. DNA-dependent protein kinase (DNA-PK) is required for efficient repair of DNA double-strand breaks (90) and protection of mammalian telomeres. Inhibition of DNA-PK phosphorylation with IC86621 led to large numbers of chromatid-type aberrations (telomere fusions). Protein tyrosine kinases (PTKs) play fundamental roles in signal transduction. It is not known how or whether deregulation of signal transduction pathways might lead to DNA damage, but two inhibitors of PTKs, tyrphostin 23 and tyrphostin 46, induced chromosomal aberrations and SCE in CHO cells (91).

Due to the lack of publications, there is little guidance on how to obtain evidence of indirect genotoxicity due to kinase inhibition. It seems likely that approaches similar to those used for suspected topoisomerase inhibitors would be appropriate and should help identify if a threshold exists. Absence of formation of DNA adducts would provide supporting information, but with limitations, as discussed in the Introduction.

Imbalance of DNA precursors
Deoxyribonucleotide pool imbalances can lead to aberrant DNA replication or repair which in turn leads to a multitude of genetic effects including mutation, recombination, strand breakage, chromosomal aberrations and loss, micronuclei and SCE (92–94). Chemicals shown to disrupt nucleotide triphosphate levels include methotrexate (and other anti-folate drugs including sulphonamides), hydroxyurea and FUdR. Chemicals causing disruption of nucleotide pools would likely lead to genotoxicity without any interaction between the chemical itself and DNA. Concentrations of chemicals below the NOEC for nucleotide pool disruption would therefore not be genotoxic. As long as these concentrations are not reached in humans, there will be no hazard to exposed individuals.

It should be possible to check whether any observed genotoxicity is due entirely to nucleotide pool imbalance by restoring the pools to a balanced state, when the genotoxic responses should return to normal. For example, chromosome aberrations induced by excessive thymidine are reduced by simultaneous addition of deoxycytidine, as described by Kunz (92). However, it should be noted that the clastogenicity may only be partially reduced rather than restored completely to control levels, and this may make conclusions on mechanisms difficult. As discussed before (Introduction), absence of DNA binding is not definitive but may provide useful additional evidence in support of an indirect, non-DNA MoA.

Energy depletion
A number of substances are toxic via mechanisms that uncouple oxidative phosphorylation. Many do not display genotoxic properties and are unable to induce UDS (e.g. flavoskyrin, xenoclauxin, desacetylduclauxin and flavoglaucin) (95–98). Some genotoxic chemicals (e.g. various chlorophenols) (98) have the ability to uncouple oxidative phosphorylation (mitochondrial respiration), although it is not clear whether this MoA is entirely responsible for the genotoxic effects in these cases. However, the picture with other inhibitors of microbial respiration is confused. The dihydroxyanthraquinone alizarin uncoupled oxidative phosphorylation but did not induce UDS whereas 1-hydroxyanthraquinone and chrysazin induced UDS but did not uncouple oxidative phosphorylation, and the closely related quinizarin and anthrarufin did neither (99).

2,4-Dinitrophenol also uncouples oxidative phosphorylation (100,101), is cytotoxic and induces chromosomal aberrations (102). However, it is highly likely that the genotoxicity does not occur as a result of energy depletion, but via an arylnitrenium ion whose activity would be enhanced by the electron-donating properties of the ring substituents (102).

Cyanide is also an uncoupler of oxidative phosphorylation. It does not induce mutations (103) but does induce DNA double-strand breaks and DNA fragmentation in mammalian cells (104,105) at high levels of cytotoxicity, suggesting a non-DNA MoA.

Thus, although energy depletion was previously identified as an indirect mechanism of genotoxicity, our review of the evidence suggests there is not a simple correlation. Perhaps, energy depletion is one of several consequences for a number of chemicals. There are very few (possibly only one) examples of substances believed to induce genotoxicity solely as an indirect consequence of inhibition of oxidative phosphorylation. If desired, experimental evidence of uncoupling oxidative phosphorylation can be obtained as described by Rahn et al. (101).

Metabolic overload
Reactive metabolites may induce DNA damage at high concentrations but be detoxified or conjugated and rendered harmless at low concentrations. Such substances are DNA reactive but with a threshold. At high concentrations of test chemical, positive results may arise both in vitro and in vivo, but normal levels of human exposure may be below the threshold.

Probably, the best example of metabolic (or metabolite) overload is paracetamol (acetaminophen). Paracetamol inhibits ribonucleotide reductase, and this may contribute to its genotoxicity, especially in vitro, but it is also metabolized to a quinone imine [N-acetyl-p-benzoquinone imine (NAPQI)] which is electrophilic and can bind to DNA. Paracetamol produces chromosomal aberrations in vitro and in bone marrow in vivo, but only at high doses. NAPQI is ordinarily conjugated to glutathione, which prevents it reacting with DNA, and as long as glutathione is available, DNA adducts and chromosomal aberrations are not formed. If the glutathione-conjugating capacity of the cells or tissues is exceeded, then genotoxicity occurs. A thorough review (106) of all the available data with paracetamol concluded there was a threshold of conjugation, which was exceeded in those tests producing positive results, but which was sufficiently higher than normal human exposures to predict that the drug was safe for continued human use.

We can speculate there will be other situations where detoxification of a DNA-reactive chemical or metabolite is overloaded, particularly in the in vitro test systems where such high concentrations can be used. Evidence for such an explanation would need to be gained from determination of the conjugation threshold, and possible absence of both DNA adducts and genotoxic responses below this threshold. Expected human exposures would need to be compared with this threshold, and appropriate safety factors employed, in order to determine safety for humans without the need for in vivo testing.

The following situations also involve direct damage to DNA but with an expected threshold below which there is no genotoxic hazard.

Production of active oxygen species. ROS induce DNA damage in vitro and in vivo, can cause cancer and are also involved in ageing (107,108). Many chemicals are believed to induce damage to DNA via production of ROS. These chemicals range from pesticides, solvents such as chloroform, carbon tetrachloride and phenols, catechols and catecholamines, oestrogens, metals and even benzo[a]pyrene. Oxidative stress can be caused by the depletion of glutathione and other antioxidant defences (109). It has often been suggested that intact mammalian tissues generally have higher glutathione levels and better antioxidant defence mechanisms than cells in culture (in particular long-established cell lines). However, no published data have been found to support this. If this suggestion is true, it is expected that compounds that induce genotoxicity via ROS mechanisms in vitro will be less likely to do so in vivo, or higher doses will be required. In such cases, chemicals inducing genotoxicity via production of ROS would damage DNA directly but would be expected to have a threshold.

Evidence of ROS activity in vitro may be obtained by study of the effects of antioxidants such as glutathione peroxidase (110), catalase (111) and superoxide dismutase (112), vitamins and various phytochemicals (113) such as polyphenols and catechins. However, our experiences suggest that ROS-induced genotoxicity is rarely reduced completely to background levels by co-incubation with antioxidants. This could indicate that chemicals that produce ROS are likely to be genotoxic via other mechanisms. However, it is also possible that antioxidants applied extracellularly are not very effective at counteracting or preventing intracellular damage. Evidence of oxidative damage to DNA may also be obtained by using the comet assay with formamidopyrimidine-DNA-glycosylase (FPG) modification. FPG is an enzyme that preferentially cleaves oxidized bases from duplex DNA. It therefore increases the migration of DNA containing oxidative damage compared with single-cell gels in the absence of FPG (114–117). An increase in DNA migration (e.g. comet tail moment) of 2-fold or more in the presence of FPG, when compared without FPG, would indicate significant sites of oxidative damage.

Thus, there may be few chemicals that induce DNA damage solely by an ROS mechanism, and even though normal human tissues may be expected to be more tolerant of ROS than established cell lines in culture, it will be difficult to define thresholds and to estimate safe levels of human exposure.

Lipid peroxidation. Lipid peroxidation is another mechanism by which ROS may be generated. This process has been implicated in the genotoxicity of asbestos, metals such as nickel, polyhalogenated biphenyls and phosphine (118–122). The main DNA-damaging products of lipid peroxidation are believed to be peroxyl radicals and electrophilic aldehydes (123). There are also examples (e.g. ricin) (124) of lipid peroxidation being induced in vivo and resulting in DNA breakage.

It has been reported that antioxidants can reduce the DNA damage induced by lipid peroxidants such as nickel (121) but there are few published examples. As 8-hydroxy-deoxyguanosine is involved in DNA damage resulting from lipid peroxidation (123), the FPG modification of the comet assay, as described and referenced above, may provide useful evidence in support of this mechanism.

As mentioned above, it has often been suggested that intact mammalian tissues generally have higher glutathione levels and better antioxidant defence mechanisms than cells in culture (in particular long-established cell lines). However, we can find no published data to support this. If it is true, it is expected that compounds that induce genotoxicity via lipid peroxidation mechanisms in vitro will be less likely to do so in vivo, or higher doses will be required.

Sulphydryl depletion. The most significant sulphydryl molecule that protects cells against the DNA-damaging effects of chemicals is glutathione. There are numerous published examples of induction of DNA damage being accompanied by glutathione depletion, e.g. acrolein, chloropropanones, some acrylate esters, 1,3-butadiene, isobutene, various metals and dichlorvos (125–130). Diethylmaleate is also able to deplete glutathione and induces mutations in mouse lymphoma cells (131). In some cases, the depletion of glutathione will be secondary to, and will therefore enhance, a direct DNA-mediated mechanism. For example, depletion of glutathione by acrolein enhances formaldehyde genotoxicity (125) and depletion of non-protein sulphydryl content by dichlorvos enhances DNA strand breaks induced by pro-oxidant ferrous iron or diethylnitrosamine (130). However, in other cases, the depletion of glutathione will allow genotoxic species to reach DNA-damaging levels that would not be reached in cells/tissues with higher (normal) levels of glutathione.

If a novel chemical is suspected of inducing genotoxicity via sulphydryl depletion, it should be possible to obtain some supporting evidence by:

(i) looking for different responses in different cell types with different constitutive levels of glutathione or

(ii) adding glutathione to treated cultures to see whether the genotoxic response is reduced, moves to higher concentrations or is eliminated.

Nuclease release from lysosomes
In 1987, Bradley et al. (132) demonstrated that the lysosomal detergent, N-dodecyl imidazole (NDI) was a weak inducer of DNA double-strand breaks and also induced chromosomal aberrations in CHO cells at cytotoxic concentrations. The authors postulated that the DNA damage was due to release of nucleases from damaged lysosomes, and that if this occurred in the presence of a DNA-reactive chemical, the genotoxic/carcinogenic effects would be enhanced. However, sufficient evidence was presented to show that significant frequencies of chromosomal aberrations (12–18% cells with aberrations) were induced by NDI alone, and where confluence of the cell monolayer was reduced only by 40% or so. Several other agents have been described as inducing lysosomal breakdown (2,4-dinitrophenol, fibres such as of silica or asbestos and dieldrin) (133–135). However, there are no studies to demonstrate whether any DNA damage induced by these agents could be due solely to lysosomal breakdown or whether (as with glutathione depletion) lysosomal damage would merely enhance a direct DNA-mediated effect.

As a non-DNA target, there should be an NOEC for genotoxicity caused by lysosomal breakdown. The only published method we can find for determining lysosomal damage is by measuring leakage of lucifer yellow CH (132). However, as this involves growing cells on cover slips, it is presumably not possible (or much more difficult) to measure lysosomal damage in mouse lymphoma cells or human lymphocytes.

There have been no new publications exemplifying genotoxicity as a result of lysosomal breakdown in nearly 20 years. With so few published examples, we conclude that it will be very difficult to obtain convincing evidence that lysosomal breakdown is the cause of a genotoxic response and that avoiding concentrations that do not break lysosomes will not pose a human hazard.

Inhibition of DNA/protein synthesis
Cycloheximide is a potent inhibitor of protein synthesis but also induces mutations in mouse lymphoma cells (131). Sodium fluoride induces a variety of genotoxic effects including chromosomal aberrations, SCE, UDS, transformation (136–138) and mutation (139,140). Some of these occur at concentrations that raised the osmolality of the medium, and we know this can result indirectly in genotoxic effects (13). However, it also inhibits DNA/protein synthesis, which is also implicated in its cytotoxicity (140), and is another mechanism leading to genotoxicity (75). Sodium arsenite also induces chromosomal aberrations and inhibits protein synthesis (141). In many cases, inhibition of protein synthesis may occur coincidentally with a genotoxic response under cytotoxic conditions rather than the cause of it. The absence of simple associations is further exemplified by the fact that diamminesilver tetraborate (142), extracts of Fusarium moniliforme contaminated corn (143) and ochratoxin A (144) all inhibit protein synthesis, but only the Fusarium toxin, zearalenone, has been reported to induce genotoxicity (chromosomal aberrations), and both ochratoxin A and zearalenone are rodent carcinogens.

If inhibition of protein synthesis is the sole cause of DNA damage, then, as a non-DNA target, there should be an NOEC below which there is no genotoxic response. Methods for demonstrating inhibition of protein synthesis (e.g. uptake of tritiated leucine) are well documented in textbooks, so obtaining evidence of an effect on protein synthesis should be straightforward. However, it is unlikely that inhibition of protein synthesis will be the sole cause of DNA damage in many cases.

Protein denaturation
Ashby et al. (145) proposed that oxidative denaturation of cellular proteins by hypochlorous acid might explain the in vitro clastogenicity (but negative in Ames test and in vivo) of calcium hypochlorite and N-chloropiperidine. Denaturation of enzymes involved in DNA synthesis or repair, or denaturation of chromatin proteins, could clearly lead indirectly to DNA damage, but these were not specifically studied by these authors. No other examples of protein denaturation leading indirectly to DNA damage have been found, and hence there is no clear guidance on the best ways to obtain evidence of this MoA. If exposure of humans to chemicals such as calcium hypochlorite and N-chloropiperidine is below the NOEC at which protein denaturation occurs, there should be no human genotoxic hazard.

Ionic imbalance
Severe ionic imbalance will usually lead either to change in pH or increase in osmolality. As both of these are controlled during in vitro studies, there should be no ‘false positives’ that would need follow-up in vivo. This is therefore not discussed further here.

Other mechanisms
In addition to the mechanisms listed in Scott et al. (13) which have been discussed above, other processes that can indirectly lead to genotoxic responses, but where an NOEC may exist, are considered below.

Aneugens. It is now widely accepted that aneugens exhibit thresholds (17). The interaction with a non-DNA target (tubulin synthesis or polymerization, the spindle or the kinetochore), and the fact that these targets can apparently experience some degree of damage while functioning normally, means there are levels of exposure (the thresholds) below which no genotoxic effects are seen. Determination of the NOEC is, however, not a simple task. Initial indications of an aneugenic mechanism may come from observation of induced polyploidy or micronuclei. Evidence that only an aneugenic mechanism is involved would need to be obtained by demonstrating that the micronuclei predominantly contained whole chromosomes (or at least centromeres, considered indicative of whole chromosome loss). There are no criteria for what proportions of centromere-positive micronuclei should be induced in order to conclude an aneugenic mechanism. However, in our experience and in the published literature (146), control human lymphocyte cultures contain 30% centromere-positive cells whereas aneugen-treated cultures show 50–90% centromere-positive cells. However, the most sensitive measure of aneuploidy appears to be non-disjunction (using fluorescent in situ hybridization on whole chromosomes in binucleate cells) (147–149) where the threshold for non-disjunction has generally been found to be lower than the threshold for micronucleus induction.

Cytotoxicity. Scott et al. (13) indicated that high levels of cell killing may indirectly lead to DNA damage, but did not have convincing evidence. Kirkland (150) presented data that indicated clastogens only positive at high levels of cytotoxicity in vitro were less likely to be positive in vivo than clastogens positive at low levels of cytotoxicity and are therefore less hazardous for humans. This view was reinforced by Hilliard et al. (42) and an industry survey reported by Galloway (151). Several of the published guidelines for genotoxicity testing referred to in the Introduction accept the possibility that genotoxicity occurring only at highly cytotoxic concentrations may not be indicative of human hazard. However, it is important to understand that if very low concentrations are both cytotoxic and genotoxic, then these may be reached, and cause deleterious effects, in humans.

To demonstrate that a high level of cell killing may be responsible for a genotoxic effect requires high quality (genotoxicity and cytotoxicity data) from very carefully designed studies. Most important is that experiments are performed with closely spaced doses (as recommended for chromosomal aberration and mouse lymphoma tests) (150,152) and that negative genotoxic responses are seen at moderate-to-high levels of toxicity, close to the (usually) single dose point giving a positive response. The shape of the response curve is often said to resemble a ‘hockey stick’ or ‘broken stick’. The method of measuring cytotoxicity is also important. Mitotic index can be influenced by cell cycle disturbance as well as by cell lethality, and can result in selection of different concentrations for testing than with other measures (153). More recently, population doubling has also been shown to avoid some cytotoxic clastogens by leading to selection of lower concentrations for test than by other measures (154).

When arguing that a genotoxic response is due to high levels of cell killing, the shape of the dose–response curve will not be sufficient on its own. Additional evidence such as absence of DNA adducts under genotoxic conditions would be very useful. Galloway et al. (64) have shown a good correlation between early inhibition of DNA synthesis and clastogenicity due to non-DNA mechanisms (such as high cytotoxicity), and demonstration of this property should provide useful additional evidence.

Unexpected activity. In many cases, there may be no clue, from the structure or intended use of the chemical, what intracellular targets may be disrupted at the high concentrations used routinely in in vitro testing. It is quite possible that chemicals intended for household or cosmetic use may interfere with topoisomerase or kinase function, DNA synthesis or repair, spindle function, etc. in mammalian cells at high in vitro concentrations. These are non-DNA modes of action likely to exhibit NOECs that have been discussed in the context of the genotoxicity of pharmaceuticals (15,23), but which may also apply to chemicals intended for non-pharmaceutical use. It may be informative to screen any chemical for interaction with molecular and sub-cellular pharmacological targets using commercial screens. This may identify a potential non-DNA target worthy of investigation and add to the WoE.

	Conclusions

Top Introduction Processes and properties... Bacterial mutation tests Mammalian cell tests Conclusions References

 
As summarized in Table I, there are a number of accepted non-DNA, in vitro-specific and threshold DNA MoAs by which a chemical may demonstrate a genotoxic effect that is either not relevant for humans or has an NOEC. We have discussed these in detail in the preceding paragraphs and explained how evidence may be obtained to support any such MoA. In many cases, however, there will be no suggestions from the intended use of the chemical what kind of molecular or cellular targets may be affected. Because chemicals can demonstrate unexpected activity at the high concentrations used in bacterial and mammalian cell tests in vitro, it may be worthwhile to screen for activity against molecular targets using commercially available pharmacology screens. This would at least identify some approaches to follow and could provide useful WoE to results from ‘wet’ investigations.

Clear evidence that the suspected non-DNA, in vitro-specific or threshold DNA MoA is the only one responsible for the genotoxicity would be very useful, but in many cases will not be forthcoming. Chemicals and their metabolites are likely to affect multiple targets, although none of them may involve DNA directly. The collection of evidence is made more difficult by the fact there are few published examples of the indirect genotoxic effects of damage to some non-DNA targets. Thus, it is not possible to determine with confidence the extent of testing (numbers of replicates, sampling times, observations, etc.) required to provide convincing evidence of such modes of indirect genotoxic action.

Where (as in the near future in the EU) legislation prevents follow-up testing in animal models, such that confirmation of the expected negative results, or further evidence of an NOEC or threshold, cannot be established, only in vitro data will be available to determine human hazard. For those industries affected by such legislation, investigations into MoAs will not be based on extensively validated methods and it will not be easy to provide robust and convincing data on human hazards from in vitro positive results.

The majority of non-DNA, in vitro-specific or threshold DNA MoAs shown in Table I can be grouped under a category of ‘overload of normal cellular physiology’. As there are no well-validated methods from which to obtain evidence of such MoAs, it may be more important to avoid overload in the first place. In some cases, the mechanistic studies may usefully be supplemented with data showing that the in vitro genotoxic response was not associated with formation of DNA adducts, although demonstration of an absence of DNA adducts is not easy and is not definitive proof that DNA was not modified. Whether it will be possible to obtain sufficient convincing evidence only from in vitro investigations to determine lack of human hazard is questionable.

Finally, the future of in vitro testing may lie with radically new assays rather than supplementary investigations to interpret the relevance of results from existing assays. If new assays with improved prediction of human hazard were available, there would be fewer in vitro positive results to investigate.

	Acknowledgments

 
The work presented in this publication was initiated by the European Cosmetic Toiletry and Perfumery Association (COLIPA) task force genotoxicity and carcinogenicity, and partly sponsored by the European cosmetic industry association (COLIPA). NB. The views expressed by the above authors have been determined by an independent scientific analysis of the available data. Although the cosmetic industry association (COLIPA, Brussels) provided some financial assistance in order that this review could be undertaken, this association had no influence on the views expressed herein.

	Notes

 
^* To whom correspondence should be addressed. Tel: +44 1423 848401; Fax: +44 1423 848983; Email: david.kirkland{at}covance.com

	References