DNA Reactivity as a Mode of Action and Its Relevance to Cancer Risk Assessment

Abstract

The ability of a chemical to induce mutations has long been a driver in the cancer risk assessment process. The default strategy has been that mutagenic chemicals demonstrate linear cancer dose responses, especially at low exposure levels. In the absence of additional confounding information, this is a reasonable approach, because risk assessment is appropriately considered as being protective of human health. The concept of mode of action has allowed for an opportunity to move off this default position; mutagenicity is now not considered as the driver but rather the mode of action is. In a more precise way, it is the set of key events that define a mode of action that is fundamental in defining the shape of a cancer dose response. A key event is an informative bioindicator of the cancer response and as such should be predictive of the tumor response, at least in a qualitative way. A clear example of the use of key events in cancer risk assessment is for DNA reactive chemicals. A series of such key events is initiated by the production of DNA damage in target cells from direct interaction of the chemical with DNA leading to the production of mutations by misreplication that results in  enhanced cell replication. This enhanced cell replication eventually leads to the development of preneoplastic cells and ultimately overt neoplasms. The response of each of these key events to dose of the chemical can inform the cancer dose–response curve shape. Thus, the dose–response curve for any DNA-reactive chemical can be predicted from knowledge of its mode of action and the behavior of the induced key events.

Introduction

This overview is intended to provide a discussion of how the cancer risk assessment process, especially as presented by the U.S. Environmental Protection Agency (EPA), has been enhanced over the past few years by relying to a much greater extent on the use of data on mechanisms of disease formation, thereby moving away from the fully pragmatic approach that had been the general practice. While the opportunities to put this enhancement into practice have remained somewhat limited to this point, the aim is clear. The EPA process is described in its 2005 Guidelines for Carcinogen Risk Assessment (EPA 2005). This set of guidelines builds upon the National Research Council (NRC) Report, Science and Judgment in Risk Assessment (NRC 1994) that refines the well-recognized framework of toxicity assessment originally proposed in the so-called Red Book, Risk Assessment in the Federal Government: Managing the Process (NRC 1983) to include: hazard identification and dose–response assessment; exposure assessment and emissions characterization; and risk characterization. The current EPA process is built on a framework of mode of action and key events together with consideration of human relevance. This short history is important for considering the quantitative assessment of risk that is generally required for risk management decisions.

It is useful to provide definitions of mode of action and key events as defined by the EPA (2005) to be able to appreciate more fully how non-cancer data can be used to inform cancer dose–response curves, the subject of this overview. The term mode of action (MoA) is defined as a sequence of key events and processes starting with interaction of an agent with a cell, proceeding through operational and anatomical changes, and resulting in cancer formation. A key event is defined as an empirically observable precursor step that is itself a necessary element of MoA or is a biologically based marker for such an element. Within this defined context, DNA reactivity is an MoA and formation of a DNA adduct can be a key event, if in turn it can be converted into a mutation by a replication error.

In general terms, there are several different MoAs that are considered as independent events although in all likelihood, it is going to be rare that only one prevails for any particular chemical. These broad classes of MoA include DNA reactivity, cytotoxicity and regenerative cell proliferation, mitogenicity, and receptor-mediated processes. The present review considers DNA reactivity as it relates to the application of MoA in a cancer risk assessment framework. A comprehensive review of the characteristics and activities of DNA-reactive chemicals can be found in Preston and Ross (2010).

DNA Reactivity as a Mode of Action and the Supporting Key Events

Preston and Williams (2005) developed a set of key events by which DNA-reactive chemicals can produce metastatic cancer. This set of key events was developed to guide the risk assessment for this group of DNA-reactive chemicals.

These key events are as follows:

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

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

3. Replication or repair errors from the damaged template

4. Mutations in genes critical for cancer development in replicating target cells

5. Enhanced cell proliferation

6. Additional mutations induced from DNA damage and replication

7. Clonal expansion of mutant cells

8. Development of preneoplastic lesions and neoplasms

9. Malignant behavior

In the past 10 years or so, considerable progress has been made in the understanding of the basis for cancer formation, particularly the molecular and cellular changes that are involved or are associated with the cancer process. It is of considerable value in this regard to be able to present a general view of cancer, which can provide a mechanistic view of all cancer types in all species. Such a view is presented by Hanahan and Weinberg (2000) initially in their Hallmarks of Cancer and in a somewhat expanded form in Hallmarks of Cancer: The Next Generation (Hanahan and Weinberg 2011). The authors originally described 6 so-called acquired characteristics of cancer that are obtained by a cell (perhaps in an undefined order) for its progression from a normal state to a metastatic cancer. In the most recent iteration, the list of acquired characteristics was increased to 10 based on new experimental information. These characteristics are genome instability and mutation, resisting cell death, deregulating cellular energetics, sustaining proliferative signaling, evading growth suppressors, avoiding immune destruction, enabling replicative immortality, tumor promoting inflammation, activating invasion and metastasis, and inducing angiogenesis. Not only can such characteristics be used in the development of directed cancer treatment but importantly for the present discussion, they can serve as the framework for identifying informative bioindicators of response—key events in the cancer mode of action. In order to acquire one or more characteristics, it is broadly necessary for a cell to acquire a specific mutation. Indeed, it has been known for some time that transformed cells are genetically unstable and contain a broad array of mutations, both gene and chromosomal (Stoler et al. 1999). However, it has been shown more recently that these genetic alterations can be broadly separated into “drivers” and “passengers” (Pleasance et al. 2010). Using an ultra high throughput sequencing approach, Greenman et al. (2007) resequenced 210 cancer cell genomes and concluded, based on a computational analysis, that driver mutations could be induced in 120 genes from the total 20,000 or so genes present in a human cell; the rest of the mutations were considered to be passengers. In terms of key events, the genetic target for critical mutations in target cells is presumed to be no larger than about 120 genes and for any particular tumor type is likely to be considerably smaller. With the advent of sophisticated molecular techniques, it is quite feasible to identify genetic alterations in exposed cells and to quantitate such mutations.

Why Considering Key Events Is Informative

At low (environmental) levels of exposure, because of the sensitivity of measurement, together with other uncertainties in dose and extrapolation model used, it is not possible through epidemiology studies or rodent bioassays to use direct measurement of cancer as the end point for establishing the shape of the response for assessing risk at low doses. For this purpose, it is necessary to use a surrogate measure of cancer. From the preceding section, it will be clear that the most informative surrogate is likely to be one that represents a key event in the cancer process. Such key events are bioindicators of the apical end point, here that is cancer. Bioindicators can be a qualitative informer of dose–response curve shape and, quite possibly, a quantitative predictor of cancer dose response. For this latter purpose, it is expected that a bioindicator of a key event that is more proximal to the apical end point itself is more likely to provide a quantitative estimate of cancer frequency for a given exposure scenario. In this same context, DNA adducts, even specific ones, are likely to be rather poor predictors of cancer dose–response curves; they have to be converted into mutations to become true key events and the relationship between induced DNA adducts and mutations is observed not to be a direct proportionality. Dose–response curves for DNA adducts are generally linear; dose–response curves for mutations and chromosomal mutations are frequently nonlinear when more complete dose ranges are used (Doak et al. 2009; Swenberg et al. 2008). The dose–response curve for mutations can best be viewed as a probability curve for the conversion (by replication errors) of specific promutagenic DNA adducts into gene or chromosomal mutations. To be truly informative, the key event is the conversion of a specific promutagenic DNA adduct in a target cell being converted through an error of replication into a critical driver mutation in a target cell. There is no information that directly informs this scenario, but suffice it to say that this probability by default will not be linear.

Cancer is a multistage process as depicted, for example, by Hanahan and Weinberg (2011) for which the set of key events are required. Thus, it is possible to view the cancer dose–response curve as being an integration of the individual probability curves for each of the key events essential for cancer development following exposure to DNA-reactive carcinogens. This statement will not hold true if one key event is limiting in which case it will determine the shape of the curve, quite possibly having an initial threshold for response (Boobis et al. 2009).

Does Knowledge of Mechanism Help Define the Dose–Response Curve?

The answer to this question is a limited, yes. The simplest approach for investigating how the mechanism of induction of mutations can aid in defining the shape of the cancer dose–response curve is to use a radiation-based approach of “hit theory” (preferably, track theory). An extensive experience on the study of mutations and chromosome alterations by ionizing radiation has led to the general conclusion that these responses can be induced by one or two ionization tracks (Ottolenghi, Ballarini, and Merzagora 1999; Steel 1996). The consequence is that gene mutations (requiring only one DNA lesion for their formation) are induced by a single ionization track and will increase as a linear function of dose. In contrast, chromosome alterations require two DNA lesions for their formation and since these can be induced by one or two ionization tracks, the outcome will be a combined linear (one track)-quadratic (two track) dose–response curve. This approach can be applied to the induction of gene mutations and chromosome alterations by DNA-reactive chemicals for which a “hit” is equivalent to a DNA modification. Mutations induced by DNA reactive chemicals can be either gene (point) mutations involving the alteration or deletion of one or two bases or chromosome mutations that involve deletion of an extensive length of DNA or the exchange of large chromosomal sections. Based on the principles of hit theory, chemicals produce gene mutations by a one “hit’ process and chromosomal alterations by two “hits.” Thus, the dose–response curve for total mutations will be a combination of one hit and two hit events, giving an overall shape of linear-quadratic (Y = aD + bD ²), where Y is the yield of mutations, D is dose, and a and b are the linear and dose-squared coefficients, respectively. In addition, for any specific chemical, the actual form of the dose–response curve will be a feature of the relative proportion of gene mutations and chromosomal mutations that it induces. For example, the higher the proportion of chromosomal mutations, the more nonlinear the curve will be. Thus, knowledge of the mutational spectrum for a particular DNA-reactive chemical will be informative for dose–response considerations. Also, given that mutations are an integral component of the carcinogenic process, it is reasonable to assume that knowledge of mutation spectrum can provide information pertinent to the dose–response curve for cancer induced by a particular chemical. Clearly, there can also be an impact on dose–response curve shape from a number of different host factors based, in part, upon how these respond at different dose levels. It is not yet feasible to use the data for mutations to develop a quantitative estimate of cancer frequency, although it is possible to use key event data for setting a point of departure for extrapolation to low dose responses (EPA 2005).

What Additional Information Is Needed?

The preceding sections demonstrate just how much we already know about the mechanism of induction of mutations, the key events involved in carcinogenesis induced by DNA-reactive chemicals, and the role of mutations as key events. What is needed is more information on the specific detail of key events for a specific tumor type induced by a specific chemical. For example, it is well known that a parent DNA-reactive chemical can be metabolized into a number of reactive and nonreactive metabolites and that these can variously induce a range of types of DNA alteration (e.g., covalent DNA adducts, DNA single and double-strand breaks, and DNA-protein crosslinks). Also, within any one of these classes, there are likely to be a number of subtypes involving specific DNA bases or structures. Measurement of total DNA alterations (e.g., total DNA adducts) can be used as a measure of dose. However, the measure that is needed is effective dose, whereby only those DNA adducts that are promutagenic are considered. This measure of effective dose can be related to a response by knowing the rate of conversion of the promutagenic DNA adducts into mutations. At this time, such relationships are generally not known although it is frequently known which DNA adducts are promutatgenic for a specific chemical or chemical class (see Preston and Ross, 2010, for a more detailed discussion). However, the construction of specifically modified DNA sequences can aid in estimating the conversion frequencies of DNA adducts to mutations (e.g., Minko et al. 2008), although the technique has only been used to a very limited extent. It is possible also to compare mutation spectrum to DNA adducted sites to establish a relationship between a specific DNA adduct and its conversion into a mutation (Pfeifer and Besaratinia 2009; Besaratinia et al. 2009). It is necessary to conduct additional research along similar lines to more fully investigate these DNA adduct/mutation relationships for a broader range of DNA-reactive chemicals and their induced DNA adduct types.

There also remains a need to more fully define the driver mutations for cancer if these are to be used in a quantitative framework. Along these lines, a recent study has identified an approach to investigate epigenetic driver mutations as distinguished from passenger ones using epigenomic analysis of DNA methylation patterns (Kalari and Pfeifer 2010). The general approach can be used for investigating driver mutations for cancer and their induction following exposure to DNA-reactive carcinogens. Much research has already addressed the issue of how mode of action can be incorporated into cancer risk assessments but, as is always the case, much remains to be done.