Cellular response to the genotoxic insult: the question of threshold for genotoxic carcinogens

Abstract

Maintenance of cellular integrity is crucial for its physiological function, which is constantly threatened by DNA damage arising from numerous intrinsic and environmental sources. By transcribing the genetic information stored in the intact DNA, RNA polymerase II generates mRNA that instructs ribosomes to produce specific proteins. DNA damage leads to interruption of mRNA synthesis with the potential production of unstable transcriptions and proteins. DNA damage results in mutations with the possible consequence of cancer. To protect their integrity, eukaryotes including mammalian cells have developed different mechanisms including DNA damage response (DDR) to ensure protection of the genome. DDR identifies DNA lesions and, depending on the severity, triggers different responses. Mild DNA damage is normally managed by DNA repair; more severe or irreparable DNA damage triggers the induction of cell death programs such as apoptosis or necrosis. Although these defence mechanisms are increasingly being understood, the critical and rate limiting parameters and their dose–response to the insulting agent are poorly evaluated. From a toxicological point of view such information is essential to accept the existence of a threshold for genotoxic carcinogens. By describing the different cellular defence mechanisms and their regulation we intend to stress the need for such information to further evaluate the plausibility of a dose-dependent threshold mechanism of genotoxic carcinogens. Besides its scientific value, a better understanding of cellular defence and the onset of such counterbalancing reactions is of regulatory importance since a scientifically defendable threshold concept for genotoxic carcinogens will allow identification of the NOAEL and the derivation of health-based exposure limits.

---

Dr Greim is Emeritus Professor of Toxicology at the Technical University of Munich. His research experience is in drug metabolism, toxicokinetics, mechanisms of carcinogenic agents, and in vitro test systems. He has published over 500 papers and two text-books on Toxicology (Wiley). He has been a member or chair of numerous national and international scientific committees such as the MAK-Committee or SCHER of DG SANCO, Brussels. He is past-president of the German Society of Pharmacology and Toxicology and organized the International Congress of Pharmacology in Munich 1998. In 1996 he received the Arnold Lehman Award of the Society of Toxicology (USA).

Dr Albertini is a Research Professor of Pathology and Emeritus Professor of Medicine at the University of Vermont, USA. His research has been focused on somatic mutations in humans and animals. Dr Albertini developed the HPRT system for human studies and is evaluating PIGA mutations for this purpose. He is past-president of the US EMS, has authored or co-authored approximately 200 papers, has served on review and expert committees and consults on issues of Genetic Toxicology. Dr Albertini was clinically active in the areas of oncology, hematology and AIDS and has served as a Director of the Vermont Cancer Center from 1993 to 1995.

---

Introduction

There is clear evidence that genetic alterations are involved in the carcinogenic process,¹ and it is generally assumed that no threshold can be identified for genotoxic carcinogens. However, there is experimental evidence of no-observed effect-levels (NOELs) for carcinogenic and mutagenic effects in repeated dose studies in animals, and there are examples indicating that the shape of the dose–responses of DNA adducts and mutations differ.^2,3 The dose–response curves of mutations start at higher concentrations than those of the DNA-adducts. It is also evident that the dose–response for mutations will reach the background mutation frequency. This implies that, at very low doses, the mutation rate induced by a genotoxic carcinogen becomes indistinguishable from the background mutation frequency. Moreover, the array of cellular defence mechanisms against a genotoxic insult questions the assumption that a single event overcomes these barriers to cancer.

Accordingly, the Scientific Committee of the European Food and Safety Authority³ has concluded that based on the current understanding of cancer biology there are levels of exposure to substances, which are both DNA-reactive genotoxic and carcinogenic, below which the cancer incidence is not increased (biological thresholds). These conclusions have been evaluated in a joint opinion by the three scientific committees of the Directorate-General for Health & Consumers (DG SANCO) of the European Commission.⁴ They concluded that a dose for a specific compound may exist below which the compound does not induce a genotoxic effect. Moreover, in its Guidelines for Carcinogenic Risk Assessment, the US Environmental Protection Agency⁵ also indicates that direct, DNA-reactive carcinogens might act via a nonlinear mode of action, thereby supporting a nonlinear dose–response extrapolation, provided that the case for this can be demonstrated or expected.

Calabrese⁶ recently concluded that the concept of linearity in the dose–response of mutagenic effects is insufficiently supported. The concept is based on studies by the group of Stern on irradiated fruit flies. Although Spencer and Stern⁷ noted that it was not uncommon for control mutation rates to exceed those seen at 25 and 50 due to background variation, Uphoff and Stern⁸ in a summarizing evaluation of three critical studies concluded that the dose–response of irradiation is linear. Accordingly the US National Academy of Sciences⁹ concluded, ‘if we increase the radiation that reaches the reproductive glands by X percent, the number of mutations caused by radiation will also be increased by X percent’. One year later these conclusions were generalized to radiation induced cancer by the US National Committee of Radiation Protection and Measurements (NCRP); and in 1977 the US National Academy of Sciences Safe Drinking Water Committee extended this to all genotoxic chemical carcinogens (for references see ref. 6).

Since then numerous, mostly descriptive studies have been performed to support or question this general concept of a linear non-threshold low dose mutagenic and carcinogenic response for DNA-reactive carcinogens. In most cases in vitro or in vivo studies have determined the dose–response of genotoxic effects or the increase in tumour rates. Studies designed to evaluate a threshold effect showed that at low doses the effects did not differ from the control values and have been used to calculate a ‘virtual safe dose’, a ‘practical threshold’ or a ‘no-adverse-effect-level’.^10–14 Quantification of these effects is difficult, though a quantitative estimate has been made by Williams et al.¹³ Considering that the spontaneous DNA modification of a mammalian cell is about one lesion per 10⁶ bases, the lowest DNA adduct level of three genotoxic carcinogens was determined to be one adduct per 10⁸ nucleotides, which is far below the spontaneous DNA modifications and considered unlikely to be of biological significance.

Although such information adds to the assumption of a no effect level at low doses, this is challenged by the argument that the statistical power is insufficient to differentiate between the low dose effects and the controls. Even in their study using 40 800 trout in a ED001 tumour bioassay Bailey et al.¹⁵ concluded that the use of over 30 000 animals did not provide proof that a threshold was reached. Since it is nearly impossible to provide an unambiguous answer to the question whether such a threshold can be generally assumed by such experimental approaches, it seems to be more appropriate to evaluate the cellular response to the impact of DNA-reactive carcinogens, especially at low exposure.

We have published this evaluation previously¹⁶ and refer the reader to this book for further details. In this review we do not intend to describe all the details and specific regulatory mechanisms of a cell. Our attempt is to increase awareness of the cellular response to the genotoxic impact of reactants and to trigger further studies to better understand at which doses the cellular integrity becomes disturbed.

Key events on the progression of cancer initiated by genotoxic carcinogens

Preston and Williams¹⁷ have generalized the key events in tumour development resulting from genotoxic carcinogens. The earliest of these deal with mutation induction, i.e. the production of a non-reversible heritable alteration in the genetic information content of a somatic cell to initiate the process, thus defining the direct, DNA reactive mutagenic MOA. The first four of these events are:

1. Exposure of target cells to ultimate DNA-reactive and mutagenic species.

2. Reaction with DNA in target cells to produce DNA damage.

3. Misreplication on damaged DNA template or misrepair of DNA damage.

4. Mutations in critical genes in replicating target cells.

Although ultimate tumour production requires several additional key events, it is in these first four that host defence mechanisms may operate to introduce thresholds in the dose–response to mutation induction – the initiating event in carcinogenesis due to chemicals with a direct, DNA reactive mutagenic MOA.

Exposure of target cells

Toxicokinetic factors determine whether a genotoxic chemical can reach a target cell for tumour initiation. An example of where it does not is formaldehyde, which is a highly reactive endogenously produced aldehyde that avidly reacts with biomolecules such as proteins and DNA – a property that renders it genotoxicity. Consequently, it is mutagenic and carcinogenic at sites of contact in rodents following inhalation exposures.¹⁸ More recently IARC has concluded that formaldehyde is a human leukemogen,¹⁹ a conclusion, however, that lacks biological plausibility.

A body of cumulative evidence indicates that inhaled formaldehyde does not enter the systemic circulation in rats, monkeys or humans at exposure concentrations of 14 ppm, 6 ppm or 1.9 ppm, respectively.^20,21 The systematic fate of exogenous formaldehyde was further studied in Sprague-Dawley rats exposed by inhalation to formaldehyde labelled with the stable isotope (13C-FA).²² Stable isotopes allow differentiation between exogenously and endogenously produced formaldehyde. No 13C-FA could be detected in blood, with detection sensitivity sufficient to have measured as little as 1.5% of the endogenous FA concentrations if 13C-FA were present. This inability of exogenously administered FA to reach the blood was attributed to rapid detoxification and high tissue reactivity at the sites of contact.

Highly sensitive liquid chromatography–mass spectrometry (LC-MS)-selected reaction monitoring (MS-SRM) methods have also been used to study formation of exogenous formaldehyde–DNA adducts in rats following exposure to stable isotope labelled formaldehyde. These studies have unequivocally demonstrated that inhalation exposures do not produce exogenous DNA adducts in tissues distal from the sites of exposure, i.e. such adducts are formed in nasal epithelium but not in bone marrow.^23–25 At low exposure levels therefore, additional DNA damage from exogenous formaldehyde over background simply does not occur distally. A genotoxic carcinogen cannot produce an effect in a tissue that it does not reach.

Reaction with DNA in target cells to produce DNA damage

Several DNA-reactive genotoxic carcinogens are endogenously produced in mammalian cells. Exposure to these chemicals occurs in two ways: as a result of their endogenous production and from environmental sources, the latter including direct exposures to the agents themselves or to agents that are metabolized to the agent. In addition to formaldehyde, ethylene oxide, acetaldehyde, ethanol and isoprene are examples of endogenous mutagens²⁶ with vinyl acetate, which is metabolized to acetaldehyde, being an example of an agent that is transformed to an endogenous mutagen.

Cellular exposures to endogenous mutagens are inescapable. Cells cope with this by maintaining intra-cellular levels at homeostatic concentrations by detoxification. However, when external exposures are high, physiological concentrations may be exceeded, and adverse effects produced. An additional mutational load resulting from exogenous sources can be made manifest, but only when physiological intracellular concentrations are exceeded.

The highly sensitive LC-MS/MS-SRM methods described above are also being used to explore the role of exogenous exposures to mutagenic chemicals that are endogenously produced. Such studies have clearly demonstrated that ethylene oxide and acetaldehyde derived exogenous adduct levels are insignificant compared to the corresponding endogenous adduct levels at low external exposure concentrations.^27,28 Even for formaldehyde, where exogenous DNA mono-adducts and guanine–guanine cross-links were found in cells at the site of contact, their levels relative to endogenous adducts were insignificant or non-detectable at low exposure concentrations.²⁵ Detoxification mechanisms introduce non-linearity and even thresholds in the dose–response curve for endogenously produced genotoxic carcinogens. At low external exposure concentrations, cells are not required to deal with DNA damage over background because it does not occur.

The cellular response to the genotoxic impact

Key events beyond these first two in the MOA for tumours produced by genotoxic carcinogens involve the induction of mutations.¹⁷ Even here however, when cells must deal with DNA damage, there are mechanisms that strongly suggest non-linearity of the dose response curves.

Several mechanisms are involved in the cellular response to genotoxic chemicals:

• metabolic inactivation of the ultimate reactive compounds;

• shielding the genetic material by membranes or proteins against attack by the reactant;

• repair of DNA lesions;

• elimination of heavily damaged cells by apoptosis or necrosis.

There is ample literature on the first two points as well as consensus that metabolic inactivation diminishes the access of genotoxins to the critical cellular target. As described above, metabolic inactivation is a critical factor in maintaining homeostasis for endogenous carcinogens. Inactivation is also operative for purely exogenous genotoxic carcinogens. Since this is generally accepted and well described (see ref. 26, 29) we focus here on the other two mechanisms.

In general and depending on cell type, differentiation stage and age, the different repair systems ensure genome integrity. These include the array of DNA repair mechanisms (homologous recombination, non-homologous end joining, nucleotide excision repair, base excision repair and mismatch repair). Obviously as already described by Tong et al.³⁰ the period of DNA synthesis is the most sensitive phase of the cell cycle to the impact of chemical mutagens. As long as cells remain in the S/G2 phase the high-fidelity homologous recombination (HR) is preferred for repair of double-strand breaks. When proliferating, the error-prone non-homologous end joining (NHEJ) repair mechanism becomes active, which is the predominant mechanism during all cell cycle phases of mammalian cells. The DNA damage response system co-ordinates repair and cell cycle progression and decides the cell's fate by operating repair, cell cycle arrest, senescence, or cell death.

Mechanisms are operative to protect against the tumour development resulting from genotoxic carcinogens even after mutation has occurred. The mutant cells may themselves be eliminated, thereby removing their progression to malignancy. Other factors such as post-translational modifications or gap-junction action maintain cellular homeostasis. These are described in greater detail below.

DNA repair

To protect their integrity, eukaryotes including mammalian cells have developed different mechanisms including the DNA damage response (DDR) to ensure protection of the genome. DDR identifies DNA lesions and, depending on the severity, leads to different responses (see ref. 31).

The base excision repair (BER) system

The BER system repairs small base lesions derived from oxidation and alkylation damage, which result from about 10⁴ damaging events per mammalian cell per day.³² The predominant cause of DNA damage is the constant production of reactive oxygen species.³³ These derive from mitochondria, cytochrome P450 dependent reactions, intracellular metabolism of xenobiotics and drugs, inflammation, ionizing radiation and ultraviolet light. The resulting H₂O₂ and ˙NO readily diffuse through the cell and are detoxified by enzymes such as glutathione reductase and catalase. Since both H₂O₂ and ROS are used in cellular signalling pathways, a cellular steady-state level is maintained, which is 10 nM in the case of H₂O₂. The most reactive ROS species is the ˙OH radical, which is responsible for the majority of oxidative DNA damage resulting in apurinic/apyrimidinic sites, single DNA strand breaks, sites of base damage.³⁴ Approximately 18 000 base losses or abasic sites per cell are formed by spontaneous depurination per day.³⁵ According to Nakamura and Swenberg³⁶ each cell contains about 100 000 endogenous DNA-lesions, of which 50 000 are abasic sites, whereas the steady state level is assumed to be about 500 base alterations per cell.¹

Most of the nuclear oxidative DNA lesions are repaired by the BER pathway.^37–39 This includes recognition of the lesion by an appropriate glycosylase, which cleaves the bond between the damaged base to the desoxyribose. At the resulting apurinic/apyrimidinic (AP) site, the DNA backbone is cleaved by a DNA AP endonuclease (APE), which creates a nick in the AP site with the 3′OH and 5′deoxyribose phosphate termini. The latter blocking end is removed by polymerase β, which then fills in the gap with the appropriate nucleotide. The remaining nick is sealed by a phosphodiester bond, which is catalysed by a DNA ligase III. Some DNA glycosilases are bifunctional and also cleave the DNA backbone. In this case the 3′ blocking end needs to be removed by APE1 exonuclease activity and polynucleotide kinase (PNK). More recently, it has been shown that the formation of BER complexes at the lesion site is regulated by the protein XRCC1 (X-ray cross complementing factor 1).

In mammals, 11 glycosilases have been identified, five of which are specific for oxidized bases. Each of these enzymes recognizes and repairs a limited number of lesions although their excision mechanism is quite common: it involves the removal of the damaged base from the DNA helix followed by a nucleophilic attack at the Cl′ of the target nucleotide promoting degradation.⁴⁰ Their specific functions and characteristics are described by Duclos et al.⁴¹ It is remarkable that the deficit of a single DNA glycosilase seems to be overcome by compensatory mechanisms involving redundant repair activities. This and other information indicate that, modified by additional proteins, mammalian BER enzymes interact with one another to increase repair efficiency. Double knockout mice turn to be tumour prone and by that establish the link between BER deficiency and tumorigenesis.

In mitochondria the DNA is associated with the inner membrane and closely located to the vicinity of the respiratory machinery, which constantly generates ROS. As a consequence the steady-state level of oxidative DNA damage is significantly higher than in nuclear DNA.⁴² Basically the same mechanisms including BER remove the DNA damage induced by ROS.⁴³

At the transcriptional side, transcription blocking DNA lesions induce the following cascade of cellular defence (see ref. 44). The blocked RNA polymerase II complexes trigger the activation of stress kinases, which induce stress responses via p53 leading to cell cycle arrest. RNA polymerase II complexes blocked at DNA lesions recruit BER enzymes to remove blocking DNA lesions via transcription coupled repair (TCR). Finally, the repaired DNA allows the transcription machinery to re-engage in mRNA synthesis. In the meantime the existing pool of mRNA may compensate for the loss of nascent RNA. Finally exhaustion of this pool affects synthesis of ‘short lived survival factors’, among them the cytoplasmic anti-apoptotic factor MCL-1 triggers apoptosis.

As described by Dianov et al.⁴⁵ the steady-state level of the BER system is tightly regulated and is linked to the amount of DNA lesions.⁴⁶ When the levels of BER exceed the level of DNA lesions, the excessive BER enzymes are degraded by proteasomal degradation. This is achieved by controlling the cytoplasmic pool of the three major BER enzymes XRCC1, Lig III and Pol b through targeted proteasomal activity. Enzymes targeted for degradation are marked with a chain of ubiquitin, which serves as a target for additional ubiquitin chain attachment. The poly-ubiquitylated enzymes become recognized by the 26S proteasome and are degraded. Accumulation of Pol b is regulated by the ARF (Alternative Reading Frame) protein, which is a well-known tumour suppressor gene (p14 in humans) and is released in response to DNA damage. Its release inhibits ubiquitin ligases and thereby increases Pol b availability and results in the increased rate of DNA repair.

Role of ATM and ATR in the repair of DNA double strand breaks, replication blocking lesions and cross links and arrest of cell cycling

DNA double strand breaks (also induced by ROS), replication blocking lesions and DNA inter-strand crosslinks are lethal DNA lesions and induce a cascade of mechanisms for their elimination. Whereas double strand breaks (DSB) are sensed by ATM (ataxia-telangiectasia mutated) protein, ATR (Rad3-related) protein is activated by single strand breaks (SSB) and gapped DNA, which results from stalled DNA replication forks.⁴⁷

ATM and ATR regulate three crucial cellular processes: DNA repair, cell cycling and apoptosis. ATM is a protein kinase, which is mainly activated by double strand breaks. It binds to several proteins of DNA binding capabilities to activate their exo-and endonuclease activities. ATR – also a kinase – is activated by single strand breaks and gapped DNA, which is generated at stalled DNA-replication forks. Recruited to replication blocking DNA adducts by ATRIP (ATR-interacting protein) the complex binds to RPA (replication protein A). During normal DNA replication the latter prevents DNA from winding back or from forming secondary structures. In the case of replication blocking the single stranded DNA–RPA interaction persists and attracts the ATR–ATRIP complex, which leads to phosphorylation of several proteins and finally to homologous recombination to eliminate the replication blockage. Both processes lead to the phosphorylation of histone 2AX. The phosphorylated histone can be visualized by immunofluorescence microscopy and is used to quantify double strand break induction and repair after exposure to genotoxic agents.

The cell-cycle checkpoint pathway mediated by the stress-activated protein kinases p38 mitogen-activated protein kinase (MAPK) and its substrate MAPKAP kinase-2 (MK2) is critical for arresting the cell cycle in the case of DNA crosslinks and DNA strand breaks.^48–50 Importantly, both the ATR-Chk1 pathway and the p38-MK2 pathway are required for effective cell-cycle arrest in the absence of p53.⁵¹ Under these conditions, cytoplasmic MK2 triggers a cell-cycle checkpoint through the post-transcriptional regulation of gene expression by modulating the function of RNA-binding proteins. MK2 phosphorylates the RNA-binding protein hnRNPA0, inducing its association with, and stabilization of, the mRNA of Gadd45α, the cyclin-dependent kinase inhibitor.⁵¹ In addition, MK2 induces miR-34c in response to DNA damage in cells that lack p53. MiR-34c then represses the translation of c-Myc to promote S-phase arrest.⁵²

ATR and ATM also interfere with cell cycling by phosphorylation of Chk1 and 2 (checkpoint kinases 1 and 2) and p53. ATR phosphorylates Chk1 after the formation of stalled DNA replication forks whereas ATM phosphorylates Chk2 upon formation of double strand breaks. Both phosphorylate p53, which prevents its proteasomal degradation.

The mechanism by which the tumour suppressor gene p53 affects cell cycling and stimulates the DNA repair machinery after activation by DNA damage has already been described by Hartwell and Kastan in 1994.⁵³ Triggering the expression of p21, which inhibits the formation of the Cyclin E-Cdk2 complex, the ATM and ATR mediated stabilization of p53 prevents the entry of cells into the S-phase, thus slowing down cell cycling and DNA replication. In addition p53 binds to ERCC3, one of several excision repair molecules that together identify and remove damaged segments from DNA.

ATM and ATR activate G1 cell cycle arrest by the phosphorylation of the phosphatase Cdc25a (cell division cycle 25 homolog A), which regulates the phosphorylation status of Cdk2. Inhibition of the phosphatase activates Cdk2, which prevents entry of the cell into the S-phase. ATM and ATR also regulate G2/M cell cycle arrest. Triggering the phosphorylation of Cdc25c, this phosphatase binds to a protein and is transported out of the nucleus and thereby becomes unable to dephosphorylate Cdk1 (cyclin dependent kinase 1), which finally results in G2/M arrest.⁵⁴

The crucial role of p53 in the regulation of cell proliferation and the progress to cancer is further emphasised by Brighenti et al.⁵⁵ Inflammatory IL-6 enhances proliferation and hinders apoptosis by p53 down regulation with the consequence of reduced E-cadherin expression resulting in increased cell invasiveness and decreased response to cytotoxic stress. IL-6 stimulates c-Myc mRNA translation, which up-regulates rRNA with the consequence of MDM2 mediated proteasomal degradation of p53.

Interestingly, the oxoguanine DNA glycosilase (OGG1), which initiates repair of 8-oxoG, is phosphorylated by Cdk4, a serine–threonine kinase.⁵⁶ The cyclin dependent kinases inhibit progression of cell cycling through the G1–S checkpoint to allow more time for DNA repair. Whereas Cdk2 activity is reduced upon H₂O₂ treatment of human diploid fibroblasts, Cdk4 activity does not change, so that this specific cyclin kinase is still available to phosphorylate and up-regulate OGG1 activity.

Different elimination rates for DNA adducts

Goggin et al.⁵⁷ determined the persistence and repair of several types of butadiene-specific DNA adducts, which results from the metabolic activation of butadiene to 1,2,3,4-di-epoxybutane (DEB). These lesions were detected in tissues of laboratory rodents exposed to butadiene (BD) by inhalation. The half-lives of the most abundant crosslinks, bis-N7G-butadiene, in mouse liver, kidney and lungs were 2.3–2.4 days, 4.6–5.7 days and 4.9 days, respectively. The in vitro half-lives of bis-N7G-butadiene were 3.5 days (S,S isomer) and 4.0 days (meso isomer) due to their spontaneous depurination. In contrast, tissue concentrations of the minor DEB adducts, N7G-N1A-BD and 1,N6-HMHP-dA, remained essentially unchanged during the course of the experiment, with an estimated half-life of 36–42 days. No differences were observed between DEB–DNA adduct levels in BD-treated wild-type mice and the corresponding animals deficient in methyl purine glycosylase or the Xpa gene. This indicates that DEB-induced N7G-N1A-BD and 1,N6-HMHP-dA adducts persist in vivo, potentially contributing to mutations and cancer observed as a result of butadiene exposure. Although the study clearly demonstrates that specific DNA adducts may be persistent or repaired slowly, the butadiene exposures were high and the effect of lower doses was not measured. This at least suggests that the high dose has overwhelmed repair capacities.

It also has been postulated that DNA repair in healthy control animals is sufficient to repair the DNA adducts formed by acute exposures to DNA-reactive chemicals, whereas long-term exposure may reduce the repair capacity by modifying the expression of DNA repair proteins (e.g.ref. 58). This may result from impaired DNA repair proteins because of mutations caused by the slow accumulation of DNA adducts over a long period of time.⁵⁹ To determine the role of oxidative DNA damage in the toxicity of polychlorinated biphenyls, Jeong et al.⁶⁰ showed that single exposure of rats to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) or a mixture of polychlorinated biphenyls (PCBs) did not significantly increase the formation of DNA M1dG adducts in the liver and brain. However, chronic exposure of single PCBs resulted in a dose dependent increase in the number of M1dG adducts, which have been further increased by a mixture of the two PCBs (126 and 153) tested. Unfortunately neither the concentrations of the PCBs in the target tissues nor the NOEL of the effects have been determined, so that it remains unclear whether they are a consequence of the higher target doses or the result of suppressed DNA repair capacity.

Apoptosis

Apoptosis—programmed cell death—represents the primary mechanism by which, in cases of heavy DNA damage, the DNA damage response leads to elimination of mutant cells from the tissues. The mechanisms that trigger apoptosis are described by Kaina et al.⁴⁷ and Ljungman.⁴⁴ In short, the extrinsic (receptor 1 based) mechanism activates ATM and ATR, which stabilizes p53. The role of the activated p53 is to induce cell cycle arrest, promote DNA repair or to trigger cells to undergo apoptosis or necrosis. Interaction of the activated p53 with the FasR/FADD receptor releases caspase-8 to activate the caspase-8 dependent apoptosis pathway. The intrinsic mitochondrial pathway is triggered by activation of p53, which activates Bax, and inactivation of Bcl-2. Both pathways increase the leakiness of mitochondria, causing release of cytochrome c from the mitochondria. Cytochrome c release induces assembly of Apaf monomers to the apoptosome resulting in the release of caspase-9, which activates caspase-3 and caspase-7. Both degrade the inhibition of caspase-activated DNAse (ICAD) and the activated DNAse cleaves the DNA into nucleosomal fragments. There is interaction between the two pathways and more details about the activation of proteins that lead to apoptosis are now understood. For example in normal cells Rb is activated and blocks de-amidation of bcl-xL, inhibiting NOXA and PUMA capabilities of Bax-activation, cytochrome c release and caspase activation. DNA damage activates p53 which up-regulates CDK-inhibitor p21 and the pro-apoptotic proteins NOXA and PUMA. In tumour cells without Rb-activation, DNA damage activates Bcl-de-amidation and releases the block of Bax leading to apoptosis.⁶¹ AP proteins inhibit caspases. Smac co-released with cytochrome c releases this inhibition.⁶²

Necrosis

In response to more severe and irreparable DNA damage, the cellular response shifts towards additional cell death programmes including mitotic catastrophe, autophagy, which might be followed by apoptosis and/or necrosis—the latter is characterized by swelling of the cell, loss of membrane integrity and the release of cytoplasmic compartments.⁶³ Upon sensing the damage a multiple network of signal transduction pathways triggers this cellular response through activation of a cascade of phosphorylation events, which initiate a number of cellular responses such as cell cycle transition, DNA replication, DNA repair and cell death. DNA damage causes hyper-activation of PARP-1, which reduces mitochondrial NAD and ATP levels. Cells with depleted ATP levels undergo necrosis instead of apoptosis because the latter requires ATP. Thus, the cellular energy status seems to decide whether severe DNA damage leads to apoptosis or necrosis.

Post-translational regulation

The three major functional processes of the DNA repair especially the BER system (recognition of the lesion and strand scission, gap tailoring, DNA synthesis and ligation) are coordinated via the XRCC1/ligase III and PARP1 scaffold proteins—essential in maintaining genomic integrity. Phosphorylation, acetylation, sumoylation, ubiquitylation and methylation of the various BER proteins can modify repair activities (for reviews see ref. 32, 34). Triggered by oxidative stress, these post-translational modifications activate and, in this way, adjust the DNA repair capacity to the actual cellular environment.

Apurinic/apyridimic endonuclease (APE) plays a major role in the repair of oxidative DNA damage like apurinic/apyridimic sites, base damage (e.g. 8-oxoguanine) and single strand breaks. As reviewed by Almeida and Sobol,³² its activity is mediated and modified by a number of reactions like acetylation, sumoylation, mono- or poly-ribosylation, mono- and poly-ubiquitylation or methylation. These post-translational modifications alter the binding characteristics, turnover rates, sub-cellular localization and/or overall efficacy of the target proteins. It is assumed that a correct modification improves the function of the protein to meet the individual cellular requirements.

The role of ATR and ATM in the coordination of protein complex formation in the BER pathway including phosphorylation of Chk1 and 2 (checkpoint kinases 1 and 2) and p53, which are involved in the control of cell cycling, has been discussed above.

The epigenetic regulation of gene expression by methylation of histones, which allows chromosomal regions to switch between on and off status,⁶⁴ indicates the further mechanism by which cells may react to a genotoxic insult. Among others the methylation process itself is controlled by Argonaute proteins,⁶⁵ which segregate into two clades, the Ago clade and the Piwi clade. Ago clade proteins complex with microRNAs and small interfering RNAs derived from double-strand RNA precursors. The microRNA–Ago complexes reduce the translation and stability of protein coding mRNAs, which results in a regulatory network that has an impact on about 30% of all genes.

In the case of the Ape gene acetylation, phosphorylation and ubiquitylation do not affect the activity of the enzyme. These reactions seem to modify the expression of the gene.³⁴

The role of the gap junction intercellular communication (GJIC) in maintenance of cellular homeostasis

GJIC plays a dominant role in the pre-transcriptional regulation of the cell. As described by Vinken et al.,⁶⁶ GJIC is a key player in the control of the cellular life cycle by providing an essential pathway for the intercellular exchange of small and hydrophilic molecules. It is assumed that numerous physiological processes are driven by substances that are exchanged via these channels contributing to the maintenance of tissue homeostasis. Gap junctions arise from the docking of two hemi-channels of adjacent cells, each of which are composed of six connexin units. Since DNA methylation maintains gene silencing, reduced connexion expression is associated with a high DNA methylation content of the connexion promoter in malignant cells of human and rodent cancer. Whereas DNA-reactive genotoxic carcinogens usually do not affect GJIC, non-genotoxic carcinogens, with several exceptions, decrease GJIC, which results in a disturbance of the equilibrium between cell growth and cell death. The extent to which GJIC-impairment is counterbalanced e.g. by increased connexin activation, at low exposure concentrations to non-genotoxic carcinogens, remains to be elucidated.

Examples for dose-dependent responses to the genotoxic impact

The information presented so far provides information that the cell possesses mechanisms to counterbalance the genotoxic impact of reactants. Although this is of scientific value, it is of little help to toxicology unless data on the dose response of the different cellular reactions are made available as exemplified subsequently.

When studying cyproterone acetate (CPA) in the transgenic Big Blue™ rat, the dose-dependence of the DNA adduct levels showed a linear increase between doses of 25 and 75 mg kg⁻¹, whereas the mutation frequency was similar to the controls at 25 and 50 mg kg⁻¹.⁶⁷ The linear dose response started at 75 mg kg⁻¹. The authors assumed that to express the mutations, an additional effect of CPA operating at high doses only—probably the mitogenic activity—is required. In previous studies, Schulte-Hermann et al.⁶⁸ showed an increase in the hepatic RNA and DNA synthesis, mitotic rate and liver growth after six consecutive oral doses of CPA between 40 and 100 mg kg⁻¹. This indicates that as long as there is no increase in cell proliferation, DNA adducts and possibly even mutations remain silent.

In evaluating the experimental and human data of vinyl chloride, Rozman et al.⁶⁹ reported that plotting the data on the logarithmic scale showed a linear dose–response, which allowed comparison of the doses used in the animal studies with the much lower human exposures. The logarithmic scale also showed a clear separation of the DNA adducts from the tumour data. The dose–response of the DNA adducts parallels the tumour dose–response data at lower doses and continues to even lower doses than the tumour data.

Gocke et al.⁷⁰ provided evidence for different dose–responses of DNA and protein alkylations and mutations. After treatment of CD1 mice and Muta™ mice with ethylmethane sulfonate (EMS), NOELs for the induction of clastogenic effects (polychromated micronuclei) in CD1 mice were 80 mg kg⁻¹ body weight per day and of around 25 mg kg⁻¹ for the induction of mutations in bone marrow cells in the Muta-mice, whereas the dose–response curves for the induction of ethyl adduct levels of terminal valine of globin in peripheral blood and of DNA in both test animals were linear to the lowest dose tested.

Similarly, low levels of methylmethansulfonate (MMS) also showed clearly no observed genotoxic effect levels (NOGELs) in vitro and in vivo, which are associated with up-regulation of the suicide enzyme O⁶MeG–DNA methyltransferase (MGMT).⁷¹ More recently Thomas et al.⁷² investigated the dose–response of mutations induced by the direct alkylator methylnitrosourea (MNU) in human lymphoblastoid cells. The NOGEL of 0.0075 μg ml⁻¹ was shifted to the left on the dose axis upon inactivation of the MGMT. Similar to Doak et al.⁷¹ the frequencies of mutants have been quantified by the hypoxanthine (guanine) phosphoribosyltransferase (H(G)PRT) assay. The available information on the dose response and NOGELs of EMS and MMS has been further evaluated by Gollapudi et al.⁷³

Swenberg et al.² studied the relationship between macromolecular adducts and mutations induced by DNA-reactive genotoxic carcinogens. The general conclusion is that DNA adducts as biomarkers of exposure extrapolate down to zero, whereas biomarkers of effect such as mutations can only be interpolated back to the background number of mutations. The likely explanation for this difference is that, at high exposures, the biology that results in mutagenesis is driven by DNA damage resulting from the chemical exposure. At very low exposures, the biology that results in mutagenesis is driven by endogenous DNA damage.

Conclusions

Maintenance of cellular integrity, which is crucial for physiological function, is constantly threatened by DNA damage arising from numerous intrinsic and environmental sources. This damage may result in DNA replication errors or DNA damage arising from endogenously formed DNA damaging chemicals such as ethylene oxide, endogenous formation of ROS, spontaneous depurination or errors of DNA polymerase. Superimposed is DNA damage resulting from exogenous agents such as ionizing radiation, viruses or the different DNA-reactive chemical agents. However, several mechanisms counterbalance interaction with the DNA and DNA damage. These include:

• Toxico-kinetic factors may operate to prevent a genotoxic chemical from ever reaching target cells.

• Intracellular detoxification mechanisms provide protection from the mutagenic consequences of endogenously produced DNA-reactive carcinogens and keep them at physiological concentrations by detoxifying enzymes, which also applies for exogenous genotoxins as long as they do not overwhelm detoxifying capacities.

• To protect their integrity, mammalian cells have developed mechanisms to ensure protection of the genome. Mild DNA damage is normally managed by DNA repair including impairment of cell cycling to enhance repair efficiency. More severe or irreparable DNA damage triggers the induction of cell death programmes such as apoptosis or necrosis.

• The background steady-state level of DNA damage is about 500 base alterations per cell, which corresponds to about one lesion per 10⁶ bases and is much lower than the approximately 18 000 base losses or abasic sites per cell formed each day. This illustrates a remarkable efficiency of the DNA repair mechanism. Assuming that there is no specific qualitative difference between the background and the chemical induced DNA damage, a small additional amount of DNA damage should be equally repaired efficiently and would not lead to an increase in mutations.

Considering this weight of evidence, it is unlikely that low exposure to genotoxic chemicals, which insignificantly affect the background rate of DNA damage, leads to impairment of cellular homeostasis.

Generally, repair is regulated by a number of pathways with overlapping specificities and becomes even more effective by activation of the complex regulatory system that includes cell cycle arrest and apoptosis. However, quantitative differences in the repair of lesions in different cells need to be considered. Due to the variety of specialized cells in different growth states and in different organs, which either rapidly proliferate while others are terminally differentiated and/or are senescent, their sensitivities to particular genotoxins vary widely. Moreover, damage can occur in sequences of nucleotides that can form non-canonical DNA structures, within which the lesions may be refractory to repair; other damage may be hidden in tightly packed transcriptionally silent chromatin and are inaccessible to repair. In any case persistent DNA lesions lead to subsequent p53 activation and by that to cell cycle arrest, up-regulated repair and finally apoptosis, which reduce the probability of trans-lesion synthesis and mutagenesis.

In most studies the efficiency and adaptability of the cellular defence mechanisms and the dose dependency of their responses to the genotoxic challenge are insufficiently described. This hampers identification of a true (or absolute) threshold for the genotoxic compound investigated. Precise information as to the level of an exogenous DNA-reactive genotoxic chemical that does not overwhelm cellular defence is urgently needed. This requires dose–response evaluation of the impact of these genotoxic agents on DNA-adduct formation and the regulation of the cellular defence. Only a few recent communications present such information (e.g.ref. 24, 28, 72, 73). Such studies may include molecular combing techniques to assess changes in the rates of replication for specific cancer genes to determine possible changes in response to mutagenic carcinogens.⁷⁴

Based on this evaluation and as discussed by Ames,^75,76 the relevance of high dose animal experiments when there is low exposure to humans is questionable. In such cases the high experimental exposure may overwhelm any cellular defence mechanism including cytotoxicity with compensatory increased cell cycling, a situation that most likely does not occur at the much lower human exposure. Any extrapolation from the dose response of such effects to the lower human exposure is error prone. So far, these high dose experiments have been the basis for classification and labelling, which does not consider the obvious differences of effects at high doses in animal experiments and the usually much lower human exposure.

References

1. R. J. Preston, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
2. J. A. Swenberg, E. Fryar-Tita, Y. C. Jeong, G. Boysen, T. Starr, V. E. Walker and R. J. Albertini, Chem. Res. Toxicol., 2008, 21, 253–265.
3. E. Gocke and M. Wall, Toxicol. Lett., 2009, 190, 298–302.
4. European Commission, Risk Assessment Methodologies and Approaches for Genotoxic and Carcinogenic Substances, Health & Consumer Protection Directorate-General, 2009.
5. Guidelines for Carcinogenic Risk Assessment, US EPA Environmental Protection Agency, Washington DC, 2005.
6. E. J. Calabrese, Environ. Mol. Mutagen., 2011, 52, 595–606.
7. W. P. Spencer and C. Stern, Genetics, 1948, 33, 43–74.
8. D. E. Uphoff and C. Stern, Science, 1949, 109, 609–610.
9. US National Academy of Sciences, Biological Effects of Atomic Radiation, Washington DC, 1956.
10. A. Maekawa, H. Onodera, Y. Matsushima, T. Nagaoka, A. Todate, M. Shibutani, Y. Kodama and Y. Hayashi, Jpn. J. Cancer Res., 1989, 80, 632–636.
11. A. Maekawa, T. Ogiu, C. Matsuoka, H. Onodera, K. Furuta, Y. Kurokawa, M. Takahashi, T. Kokubo, H. Tanigawa and Y. Hayashi, et al. , Gan. Gann., 1984, 75, 117–125.
12. G. M. Williams, M. J. Iatropoulos and A. M. Jeffrey, Toxicol. Pathol., 2000, 28, 388–395.
13. G. M. Williams, M. J. Latropoulos and A. M. Jeffrey, J. Toxicol. Pathol., 2005, 18, 69–77.
14. G. M. Williams, M. J. Latropoulos and A. M. Jeffrey, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
15. G. S. Bailey, A. P. Reddy, C. B. Pereira, U. Harttig, W. Baird, J. M. Spitsbergen, J. D. Hendricks, G. A. Orner, D. E. Williams and J. A. Swenberg, Chem. Res. Toxicol., 2009, 22, 1264–1276.
16. H. Greim and R. J. Albertini, The cellular response to the genotoxic impact: The question of threshold for genotoxic carcinogens, RSC Publishing, 2012.
17. R. J. Preston and G. M. Williams, Crit. Rev. Toxicol., 2005, 35, 673–683.
18. IARC, IARC Monographs on the Evaluation of the Carcinogenic Risks to Humans, Volume 88, Formaldehyde, IARC, WHO, Lyon, 2006.
19. IARC, IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol. 100, A Review of Human Carcinogens Part F: Chemical Agents and Related Occupations, IARC, WHO, Lyon, 2012.
20. M. Casanova, H. D. Heck, J. I. Everitt, W. W. Harrington Jr. and J. A. Popp, Food Chem. Toxicol., 1988, 26, 715–716.
21. H. D. Heck, M. Casanova-Schmitz, P. B. Dodd, E. N. Schachter, T. J. Witek and T. Tosun, Am. Ind. Hyg. Assoc. J., 1985, 46, 1–3.
22. A. J. Kleinnijenhuis, Y. C. Staal, E. Duistermaat, R. Engel and R. A. Woutersen, Food Chem. Toxicol., 2013, 52, 105–112.
23. K. Lu, L. B. Collins, H. Ru, E. Bermudez and J. A. Swenberg, Toxicol. Sci., 2010, 116, 441–451.
24. K. Lu, B. Moeller, M. Doyle-Eisele, J. McDonald and J. A. Swenberg, Chem. Res. Toxicol., 2011, 24, 159–161.
25. B. C. Moeller, K. Lu, M. Doyle-Eisele, J. McDonald, A. Gigliotti and J. A. Swenberg, Chem. Res. Toxicol., 2011, 24, 162–164.
26. H. M. Bolt, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
27. D. A. Marsden, D. J. Jones, R. G. Britton, T. Ognibene, E. Ubick, G. E. Johnson, P. B. Farmer and K. Brown, Cancer Res., 2009, 69, 3052–3059.
28. B. C. Moeller, L. Recio, A. Green, W. Sun, F. A. Wright, W. M. Bodnar and J. A. Swenberg, Toxicol. Sci., 2013, 133, 1–12.
29. W. Dekant and A. Mally, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
30. C. Tong, M. Fazio and G. M. Williams, Proc. Natl. Acad. Sci. U. S. A., 1980, 77, 7377–7379.
31. P. C. Hanawalt, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
32. K. H. Almeida and R. W. Sobol, DNA Repair, 2007, 6, 695–711.
33. P. Jezek and L. Hlavata, Int. J. Biochem. Cell Biol., 2005, 37, 2478–2503.
34. C. S. Busso, M. W. Lake and T. Izumi, Cell. Mol. Life Sci., 2010, 67, 3609–3620.
35. T. Lindahl and D. E. Barnes, Cold Spring Harbor Symp. Quant. Biol., 2000, 65, 127–133.
36. J. Nakamura and J. A. Swenberg, Cancer Res., 1999, 59, 2522–2526.
37. T. Izumi, L. R. Wiederhold, G. Roy, R. Roy, A. Jaiswal, K. K. Bhakat, S. Mitra and T. K. Hazra, Toxicology, 2003, 193, 43–65.
38. J. C. Fromme and G. L. Verdine, Adv. Protein Chem., 2004, 69, 1–41.
39. D. E. Barnes and T. Lindahl, Annu. Rev. Genet., 2004, 38, 445–476.
40. J. T. Stivers and Y. L. Jiang, Chem. Rev., 2003, 103, 2729–2759.
41. S. Duclos, S. Doublié and S. S. Wallace, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
42. M. L. Hamilton, Z. Guo, C. D. Fuller, H. Van Remmen, W. F. Ward, S. N. Austad, D. A. Troyer, I. Thompson and A. Richardson, Nucleic Acids Res., 2001, 29, 2117–2126.
43. P. Liu and B. Demple, Environ. Mol. Mutagen., 2010, 51, 417–426.
44. M. Ljungman, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
45. G. L. Dianov, G. Orlando and J. L. Parsons, in The Cellular Response to the Genotoxic Insult, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
46. J. L. Parsons, P. S. Tait, D. Finch, I. I. Dianova, S. L. Allinson and G. L. Dianov, Mol. Cell, 2008, 29, 477–487.
47. B. Kaina, M. Christmann, M. Eich and W. P. Roos, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
48. I. A. Manke, A. Nguyen, D. Lim, M. Q. Stewart, A. E. Elia and M. B. Yaffe, Mol. Cell, 2005, 17, 37–48.
49. H. C. Reinhardt, A. S. Aslanian, J. A. Lees and M. B. Yaffe, Cancer Cell, 2007, 11, 175–189.
50. S. Morandell, H. C. Reinhardt, I. G. Cannell, J. S. Kim, D. M. Ruf, T. Mitra, A. D. Couvillon, T. Jacks and M. B. Yaffe, Cell Rep., 2013, 5, 868–877.
51. H. C. Reinhardt, P. Hasskamp, I. Schmedding, S. Morandell, M. A. van Vugt, X. Wang, R. Linding, S. E. Ong, D. Weaver, S. A. Carr and M. B. Yaffe, Mol. Cell, 2010, 40, 34–49.
52. I. G. Cannell, Y. W. Kong, S. J. Johnston, M. L. Chen, H. M. Collins, H. C. Dobbyn, A. Elia, T. R. Kress, M. Dickens, M. J. Clemens, D. M. Heery, M. Gaestel, M. Eilers, A. E. Willis and M. Bushell, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 5375–5380.
53. L. H. Hartwell and M. B. Kastan, Science, 1994, 266, 1821–1828.
54. S. N. Dalal, C. M. Schweitzer, J. Gan and J. A. DeCaprio, Mol. Cell. Biol., 1999, 19, 4465–4479.
55. E. Brighenti, C. Calabrese, G. Liguori, F. A. Giannone, D. Trere, L. Montanaro and M. Derenzini, Oncogene, 2014, 33, 4396–4406.
56. J. Hu, S. Z. Imam, K. Hashiguchi, N. C. de Souza-Pinto and V. A. Bohr, Nucleic Acids Res., 2005, 33, 3271–3282.
57. M. Goggin, D. Sangaraju, V. E. Walker, J. Wickliffe, J. A. Swenberg and N. Tretyakova, Chem. Res. Toxicol., 2011, 24, 809–817.
58. C. Y. Chan, P. M. Kim and L. M. Winn, Toxicol. Sci., 2004, 81, 849–855.
59. M. E. Wyde, V. A. Wong, A. H. Kim, G. W. Lucier and N. J. Walker, Chem. Res. Toxicol., 2001, 14, 849–855.
60. Y. C. Jeong, N. J. Walker, D. E. Burgin, G. Kissling, M. Gupta, L. Kupper, L. S. Birnbaum and J. A. Swenberg, Free Radicals Biol. Med., 2008, 45, 585–591.
61. C. Li and C. B. Thompson, Science, 2002, 298, 1346–1347.
62. D. W. Nicholson and N. A. Thornberry, Science, 2003, 299, 214–215.
63. O. Surova and B. Zhivotovsky, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
64. A. G. Rivenbark and B. D. Strahl, Science, 2007, 318, 403–404.
65. A. A. Aravin, G. J. Hannon and J. Brennecke, Science, 2007, 318, 761–764.
66. M. Vinken, V. Rogiers and T. Vanhaecke, in The cellular Response to the genotoxic insult. The question of threshold for genotoxic carcinogens, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
67. O. Krebs, B. Schafer, T. Wolff, D. Oesterle, E. Deml, M. Sund and J. Favor, Carcinogenesis, 1998, 19, 241–245.
68. R. Schulte-Hermann, V. Hoffman, W. Parzefall, M. Kallenbach, A. Gerhardt and J. Schuppler, Chem.-Biol. Interact., 1980, 31, 287–300.
69. K. K. Rozman, L. Kerecsen, M. K. Viluksela, D. Osterle, E. Deml, M. Viluksela, B. U. Stahl, H. Greim and J. Doull, Drug Metab. Rev., 1996, 28, 29–52.
70. E. Gocke, T. Singer and L. Müller, in The cellular response to the genotoxic insult, ed. H. Greim and R. J. Albertini, RSC Publishing, 2012.
71. S. H. Doak, K. Brusehafer, E. Dudley, E. Quick, G. Johnson, R. P. Newton and G. J. Jenkins, Mutat. Res., 2008, 648, 9–14.
72. A. D. Thomas, G. J. Jenkins, B. Kaina, O. G. Bodger, K. H. Tomaszowski, P. D. Lewis, S. H. Doak and G. E. Johnson, Toxicol. Sci., 2013, 132, 87–95.
73. B. B. Gollapudi, G. E. Johnson, L. G. Hernandez, L. H. Pottenger, K. L. Dearfield, A. M. Jeffrey, E. Julien, J. H. Kim, D. P. Lovell, J. T. Macgregor, M. M. Moore, J. van Benthem, P. A. White, E. Zeiger and V. Thybaud, Environ. Mol. Mutagen., 2013, 54, 8–18.
74. J. Herrick and A. Bensimon, Methods Mol. Biol., 2009, 521, 71–101.
75. B. N. Ames, M. Profet and L. S. Gold, Proc. Natl. Acad. Sci. U. S. A., 1990, 87, 7782–7786.
76. B. N. Ames, M. Profet and L. S. Gold, Proc. Natl. Acad. Sci. U. S. A., 1990, 87, 7777–7781.

---