Genotoxicity results from the adverse effects of chemicals on DNA and the genetic processes of living cells.22 These effects can be in the form of point mutations in the DNA (mutagenicity), DNA strand breaks, or chromosome alterations. For a long time now there has been a strong association between genetic damage and carcinogenicity, especially in biological systems that have the requisite metabolic activation.
In vitro assays for the detection of gene mutations have been in use for over two decades, with the Ames assay23 being adopted as the de facto gold standard. The standard Ames assay consists of testing a compound against five different strains of S. typhimurium in both the presence and absence of S9 metabolic activation in concentrations up to
5000 mg per plate. An increase of greater than twofold in the number of revertant colonies per plate when compared with the control is considered a positive response for that compound. A single positive response for any of the strain/S9 conditions results in an overall positive mutagenicity call for the compound being tested.
While not considered 'high-throughput' by today's standards, the Ames assay is still regarded as one of the most rapid and reproducible assays available in toxicology. An analysis of the inter- and intralaboratory reproducibility of Salmonella test results yielded a strict positive-versus-negative concordance of 85%.24 The speed, cost, and ease of use of this assay has resulted in a large number of unique chemicals being tested and the subsequent publication of their results in the public domain.
This rich source of data combined with the regulatory implications of having an Ames-positive result make mutagenicity an obvious target for the development of in silico models to predict the outcome of this assay. Because of the relatively simple mechanisms by which point mutations can occur, this toxicological endpoint renders itself amenable to all types of model development. Indeed, there are many commercial systems that claim to be able to predict mutagenicity (see Section 5.39.4 on commercial systems for more details).
In the late 1980s and early 1990s, Ashby and Tennant and their co-workers published reviews of compounds tested in genotoxicity assays as part of the US National Toxicology Program (NTP).25-27 This analysis included an assessment of the relationship between chemical structure and its ability to cause genotoxicity. They identified that there were many chemical functional groups (Figure 3) that had a strong association with causing genetic toxicity.
These structural alerts have formed the basis of many predictive models for this endpoint, and been used extensively within the pharmaceutical and agrochemical industry to screen out and avoid potentially genotoxic compounds. The unfortunate consequence of this is that any training set based solely on pharmaceuticals or agrochemicals is going to be heavily biased toward Ames negatives, making the development of robust models difficult for these types of compounds. In a review of some 394 pharmaceuticals currently on the market, only 7.2% of compounds were reported to give a positive response in a bacterial mutation assay.28 Even taking into account that most Ames-positive compounds are not pursued as viable drug candidates, reports suggest that the percentage of compounds synthesized and tested by the industry that are in fact Ames-positive is only around 10%.14'29 In contrast, the data set assessed by Ashby and Tennant and their co-workers contained approximately 37% Ames positive compounds.
Given that there are substantial differences between the types of chemicals tested by the NTP and those synthesized by the pharmaceutical industry, models based solely on the NTP data set are unlikely to perform well against pharmaceuticals. This observation has been reported on numerous occasions in reference to the commercially available systems for toxicity prediction.14,28-30
Some in silico models for mutagenicity prediction take the overall Ames result as a simple binary classification in the development of the model (i.e., either positive or negative). However, other methods have sought to make the results of this assay a more continuous variable and hence amenable to statistical analysis. These methods have introduced the concept of mutagenic potency, and several methods for calculating mutagenic potency have been proposed, each with advantages and disadvantages:
• Method 1 uses the maximum increase in revertants as a measure of mutagenic potency, that is, the potency of a compound increases in relation to the maximum increase in revertants observed in the assay. This method provides an absolute measure of the potency, but does not take into account the potential for cytotoxicity that would limit the number of revertants at higher concentrations.
• Method 2 assumes that compounds that elicit positive responses in multiple-strain/S9 combinations are more potent mutagens than those that only show positive responses under a single condition. This method provides a more objective view of potency, but can become a subjective measure where compounds illicit a reproducible and robust high increase in the number of revertants but only under one or two specific conditions.
• Method 3 uses the drug concentration (e.g., mgper plate) that induces a 2.0-fold increase in the number of revertants, and therefore the more potent the mutagen the lower the concentration required to yield a positive response. Although this method offers the advantage that the concentration of compound is used, consideration of molar concentrations would further improve this approach. However, this method does not take into account the potential for cytotoxicity to impact on the number of revertants.
Often these methods will focus on particular classes of chemicals, developing mathematical relationships between the mutagenic potency and the chemical structures. One important class of potential mutagens - aromatic amines - has been the focus of attention for many researchers working in this area. This is primarily as a result of their suspected link with carcinogenicity through their presence in cooked foods, diesel fumes, and common environmental contaminants.
The metabolic activation of aromatic amines is complex (Figure 4). They can be converted to aromatic amides in a reaction catalyzed by an acetyl coenzyme A (CoA)-dependent acetylase. The acetylation phenotype varies among the population: persons with the rapid acetylator phenotype are at higher risk of colon cancer,32,33 whereas those who are slow acetylators are at increased risk of bladder cancer.34 This latter association may result from the fact that activation of aromatic amines by N-oxidation is a competing pathway for aromatic amine metabolism. Also, the N-hydroxylation products when protonated under the acid conditions in the urinary bladder form reactive electrophiles that bind covalently with DNA or proteins to produce macromolecular damage.
An initial activation step for both aromatic amines and amides is N-oxidation by CYP1A2. This cytochrome P450 is also responsible for the 3-demethylation of 1,3,7-trimethylxanthine (i.e., caffeine), the distribution of metabolic phenotypes in the population, as well as the disposition of an individual with respect to CYP1A2 metabolism, is relatively easy to determine.35 The reaction of N-hydroxy-arylamines with DNA appears to be acid catalyzed, but they can be further activated by either an acetyl CoA-dependent O-acetylase or a 3'-phosphoadenosine-5'phosphosulfate-dependent O-sulfotransferase. The N-arylhydroxamic acids, which arise from the acetylation of N-hydroxy-arylamines or N-hydroxylation of aromatic amides, are not electrophilic; therefore, they require further activation. The predominant pathway for this occurs through acetyltransferase-catalyzed rearrangement to a reactive N-acetoxy-arylamine. Sulfotransferase catalysis results in the formation of N-sulfonyloxy-arylamides. This complex pathway results in two major adduct types: amides (i.e., acetylated) and amines (i.e., nonacetylated).
The heterocyclic amines are formed during the preparation of cooked food, primarily from the pyrolysis ( > 150 °C) of amino acids, creatinine, and glucose. Heterocyclic amines have been recognized as food mutagens,36 and they have been shown to form DNA adducts and cause liver tumors in primates.37 Compared with other carcinogens, their metabolism is less well understood, but N-hydroxylation is considered to be a necessary step. Because they are similar in structure to the aromatic amines, it is not surprising that they too can be activated by CYP1A2. The N-hydroxy metabolites of 3-amino-1-methyl-5H-pyrido[4,3-b]indole (Trp-P-1), 2-amino-6-methyldipyrido[1,2-a:39,29-d]imida-zole (Glu-P-1), and 2-amino-3-methyl-imidazo-[4,5-f]quinoline (IQ) can react directly with DNA. Unlike aromatic amines, however, this reaction is not facilitated by acid pH. Enzymic O-esterification of the N-hydroxy metabolite is important in the activation of these food mutagens, and the N-hydroxy metabolite is also a good substrate for transacetylases. This suggests a possible role for these chemicals in the etiology of colorectal cancer in combination with the rapid acetylator phenotype.
Covalent binding to DNA
Mutagenicity and/or carcinogenicity
Figure 4 Metabolic activation pathways of aromatic amines.38
Through the use of various QSAR techniques, the mutagenic potency of aromatic amines has been correlated to a number of structural descriptors including lipophilicity as represented by the log P value of a chemical,38 the relative stability of the nitrenium ion,39 and the HOMO/LUMO calculations for class. However, upon further investigation, these methods, while providing an acceptable mechanism for predicting mutagenic potency, were poor at predicting the activity of nonmutagenic aromatic amines. Work conducted by Benigni et al4 showed that both electronic and steric factors at the amine site also play a role in the crucial metabolic activation of this class of compounds. For a comprehensive review of QSARs for aromatic amines, refer to Benigni et al41
In contrast to mutagenicity, very few in silico models have been developed to predict the potential of chemicals to cause chromosomal damage. Chromosomal damage can take various forms, including structural aberrations such as clastogenicity, and/or numerical aberrations, including aneuploidy.42 There are many assays developed for the detection of these endpoints, including the mouse lymphoma assay, in vitro cytogenetics assays using human lymphocyte cell lines, and in vivo micronucleus assays. Due to their low throughput and higher cost, these assays are often run only when compounds have already been shown to be negative in the Ames assay. Some companies also use higher-throughput screening assays, such as the in vitro micronucleus assay, to eliminate potential bad actors earlier in the drug discovery process. Often these company-generated data remain within corporate databases, and as such have not been subjected to the same level of scrutiny as that for Ames mutagens.
Some data on chromosome damaging agents have been published by Ishidate et al.4 and, more recently, by Ojiama et al4 Despite this, only a few attempts have been made to develop predictive models for this endpoint. One reason for this could be the diversity and complexity of mechanisms that are involved in chromosomal damage. Another possibility is that comparing results from the different assays makes combining data from these difficult and can cause problems when applying statistical algorithms to develop QSAR models.
Serra et al.45 report some success using topological descriptors and six-descriptor k-nearest-neighbor model to predict structural chromosomal aberrations. This model was built using a training set of some 245 chemicals, and was capable of predicting a test set comprised of 37 chemicals - 26 negatives and 11 positives - with an overall accuracy of 86.5%. Use of support vector machine algorithms46 on the same training and test sets yielded models with an overall accuracy of 83.8%.
Other QSAR models for chromosomal aberrations have been developed by Tafazoli et al.,47 but these were focused on a specific chemical class of compounds, namely short-chain chlorinated hydrocarbons. As such, the application of this model to noncongeneric data sets is inappropriate.
Was this article helpful?
Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.