Alternatives To The Carcinogenicity Bioassay- In Silico Methods, And The In Vitro And In Vivo Mutagenicity Assays.pdf

Review

Alternatives to the carcinogenicity bioassay: in silico methods, and the in vitro and in vivo mutagenicity assays

1.

Introduction

2.

Chemical carcinogenicity and mutagenicity: scientific

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

background 3.

Istituto Superiore di Sanita, Environment and Health Department, Viale Regina Elena 299, 00161 Rome, Italy

Experimental alternatives to the rodent bioassay

4.

Formalizing the chemistry of carcinogenicity

5.

Conclusions

6.

Expert opinion

Romualdo Benigni†, Cecilia Bossa, Olga Tcheremenskaia & Alessandro Giuliani

Importance of the field: Carcinogenicity and mutagenicity are toxicological end points posing considerable concern for human health. Due to the cost in animal lives, time and money, alternative approaches to the rodent bioassay were designed based on: i) identification of mutations and ii) structure--activity relationships. Areas covered in this review: Evidence on i) and ii) is summarized, covering 4 decades (1971 -- 2010). What the reader will gain: A comprehensive, state-of-the-art perspective on alternatives to the carcinogenicity bioassay. Take home message: Research to develop mutagenicity-based tests to predict carcinogenicity has generated useful results only for a limited area of the chemical space, that is, for the DNA-reactive chemicals (able to induce cancer, together with a wide spectrum of mutations). The most predictive mutagenicity-based assay is the Ames test. For non-DNA-reactive chemicals, that are Ames-negative and mutagenic in other in vitro assays (e.g., clastogenicity), no correlation with carcinogenicity is apparent. The knowledge on DNA reactivity permits the identification of genotoxic carcinogens with the same efficiency of the Ames test. Thus, a chemical mutagenic in Salmonella and/or with structural alerts should be seriously considered as a potential carcinogen. No reliable mutagenicity-based alternative tools are available to assess the risk of non-DNA-reactive chemicals. Keywords: genotoxicity, prediction, QSAR, structural alert, toxicology, tumorigenesis Expert Opin. Drug Metab. Toxicol. (2010) 6(7):809-819

1.

Introduction

More than other toxicity end points, carcinogenicity has been the subject of a long series of mechanistic investigations, and a considerable scientific background is thus available. Because the rodent bioassay -- the main experimental tool for assessing the carcinogenicity of chemicals -- involves a heavy cost in terms of animal lives, time and money [1,2], a plethora of short-term, alternative approaches were designed during the last decades, and considered as candidates for testing schemes and regulatory acceptance. Following the seminal work of the Millers [3,4] on electrophilic and pro-electrophilic carcinogens, distinguished contributions to the advancement and dissemination of knowledge in this field came from several investigators. Two main lines of research were followed: i) the search for cheaper and shorterterm alternatives (short-term tests, STTs) to the rodent bioassay. The large majority of these alternative approaches are based on the detection of genotoxic effects; and

10.1517/17425255.2010.486400 © 2010 Informa UK Ltd ISSN 1742-5255 All rights reserved: reproduction in whole or in part not permitted

809

Alternatives to the carcinogenicity bioassay

Article highlights. .

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

.

.

.

.

A short-term mutation-based, in vitro assay is set up by Bruce Ames, and results have shown it to be highly predictive of chemical carcinogens. The hypothesis that the relationship between the mutation and cancer is a general one and may be the basis for expanding the predictability of carcinogens, stimulated the scientific community to devise further mutagenicity assays that are able to cover the whole spectrum of mutations. Unfortunately, additional in vitro short-tem mutagenicity assays do not complement the Ames assay in the prediction of carcinogenicity. In addition, the in vivo mutagenicity assays, expected to filter excess in vitro positive results, give rise to too many false negatives. Recent attempts to exploit the modern omics technologies (in vitro) to predict in vivo toxicity do not look very promising. The Ames test and the structural alerts coding for DNA-reactivity predict efficiently the genotoxic carcinogens, whereas the predictability of non-genotoxic carcinogens is still limited. The predictions obtained by the structural alerts can be refined by the use of fine-tuned quantitative structure--activity relationships for congeneric classes of chemicals.

This box summarizes key points contained in the article.

different structures for which metabolism to electrophilic reactants had been demonstrated. Overall, this evidence led them to suggest ‘that most, if not all, chemical carcinogens either are, or are converted in vivo to, reactive electrophilic derivatives which combine with nucleophilic groups in crucial tissue components, such as nucleic acids and proteins’ [4]. After a number of decades, the hypothesis of the electrophilic reactivity of (many classes of) chemical carcinogens maintains its validity and has been incorporated into a more general theory on the chemical carcinogens. From the point of view of their mechanism of action, carcinogens are classified into: i) genotoxic carcinogens, which cause damage directly to DNA; many known mutagens are in this category, and often mutation is one of the first steps in the development of cancer [5] and ii) epigenetic carcinogens that do not bind covalently to DNA, do not directly cause DNA damage, and are usually negative in the standard mutagenicity assays [6]. Whereas the epigenetic carcinogens act through a large variety of different and specific mechanisms, the genotoxic carcinogens have the unifying feature that they are either electrophiles per se or can be activated to electrophilic reactive intermediates, as originally postulated by the Millers.

Experimental alternatives to the rodent bioassay

3.

The Salmonella typhimurium or Ames test After Millers’ work, a crucial task faced by scientists was the generation of cheaper and shorter-term alternatives to the rodent bioassay that was, and largely still is, the main tool of the research on chemical carcinogens. Bruce Ames created a series of genetically-engineered Salmonella typhimurium bacterial strains, each strain being sensitive to specific classes of chemical carcinogens (e.g., alkylating, intercalating). The Salmonella, or Ames test, is an in vitro model of chemical carcinogenicity --as understood in the Millers’ time -- and consists of a range of bacterial strains that together are sensitive to a large array of DNA damaging agents [7-10]. Because most of the known carcinogens at that time acted through genotoxic mechanisms, the activity of carcinogens as mutagens to Salmonella almost always seems plausible within the context of the Millers’ hypothesis [11]. The application of the Ames test to large numbers of chemicals has shown that this assay has a high positive predictivity for the chemical carcinogens: Ames-test mutagens have a high probability (around 80%) of being also carcinogenic in rodents, whereas a negative result has no discriminatory value (a chemical negative in Salmonella has the same probability of being either a noncarcinogen or a nongenotoxic carcinogen) [9,12]. It should be emphasized that the excellent performance of the Ames test in the identification of carcinogens derives directly from its brilliant scientific design: it has been purposely constructed as to mimic in vitro the rodent bioassay, with each strain engineered to be sensitive to a 3.1

ii) the formalization of the structure--activity relationships (SAR) concepts that assisted the Millers in understanding the mechanisms of chemical carcinogenicity. In this paper, the two lines of research are summarized -- through both published and original results -- and it is shown how the combined consideration of the two arms of research contributes to a better understanding of the state of the art in the field of alternatives to the carcinogenicity experiments. This novel, integrated view of the field is obtained by exploiting a wide range of tools made available by the recent progress of in silico approaches, ranging from quantitative structure--activity relationships (QSAR) to expert systems that code for mechanisms of chemical carcinogenicity, and recently developed chemical relational databases.

Chemical carcinogenicity and mutagenicity: scientific background 2.

Historically, the electrophilic theory of chemical carcinogenesis developed by James and Elizabeth Miller [3,4] enabled the activity of the large majority of animal carcinogens known by the 1970s to be tentatively rationalized. Historically, the Millers first noted the electrophilicity of carcinogenic alkylating agents. Since then, a number of alkylating agents were found to be carcinogenic, and these chemicals were also electrophilic as administered. In addition, the Millers were much impressed by the variety of chemical carcinogens of rather 810

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

Benigni, Bossa, Tcheremenskaia & Giuliani

known type of carcinogens. However, the correlation between the Ames test and carcinogenicity is valid only at the level of yes/no activity, whereas mutagenic and carcinogenic potencies are uncorrelated [13]: thus, only the (first?) rate-limiting step of the chemical--macromolecules interaction is in common in the two systems. Looking for further mutagenicity assays to complement the Ames test

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

3.2

Because the scientific community was convinced that the correlation between chemical carcinogenicity and mutation was a general one and that it was possible to increase the above correlation by considering also genetic events different from those at the basis of the Ames test (i.e., base substitutions and deletions/additions), further mutation assays based on other events, namely, structural chromosome aberrations (breaks and rearrangements) and numerical chromosome aberrations (loss or gain of chromosomes, defined as aneuploidy) were developed. In addition, the use of tests based on cultured cells of mammalian origin was explored as well [14]. The idea of the complementarity between STTs (different for their genetic end point and the phylogenetic characterization) has been substantially accepted in the various regulatory settings. Normally three (or in some cases two) in vitro tests are required by regulatory authorities, namely a test for induction of gene mutation in bacteria, a test for induction of gene mutations in mammalian cells and a test for induction of chromosomal aberrations in mammalian cells. A corollary is that positive genotoxicity in vitro should be confirmed, in a second phase, in an in vivo genotoxicity assay in order to avoid the possibility of false positive results (i.e., positive results due to artifacts of the in vitro tests), and a positive in vivo genotoxicity test may be sufficient to identify a chemical as a potential human carcinogen [15-18] (see also the Technical Guidance Documents of the European Chemicals Agency: http://guidance.echa.europa.eu/docs/guidance_document/ information_requirements_r7a_en.pdf?vers=20_08_08). However, even though several of the ‘additional’ genotoxicity assays have gained a large popularity and have been adopted in various regulatory settings, rigorous comparative studies have failed to demonstrate a correlation with and predictive ability in respect to rodent carcinogenicity [12,19-22]. This point is examined more in detail below. The relationships between carcinogenicity and the additional mutagenicity assays

3.3

It should be emphasized that, even though an increasing number of nongenotoxic carcinogens are observed among the chemicals put into the market in recent times, a statistically significant relationship between the results of the two assays in the available database of chemicals tested with both is clearly apparent. Table 1 reports the concordance between rodent carcinogenicity and Ames test results, based on the

data in the large database ISSCAN v3a, which is characterized by the high quality of the biological calls [23]. ISSCAN is freely available from the website of the Istituto Superiore di Sanita: http://www.iss.it/ampp/dati/cont.php?id=233&lang=1&tipo=7. In this large set of results, it appears clearly that the Ames test has a high positive predictivity for carcinogenicity: 311 out of 387 Salmonella mutagens (i.e., 80%) are actually carcinogenic (Chi-square = 92.6696, p < 0.0001). On the contrary, the evidence on statistically significant relationships between the rodent bioassay results and the ‘additional’ genotoxicity assays adopted in regulatory settings was never demonstrated. A clear-cut experiment was performed by the National Toxicology Program (NTP): 114 chemicals, with rodent carcinogenicity results, were tested in four basic in vitro genotoxicity assays that differed both in terms of types of cells and of genetic end point: Ames test (bacteria, gene mutation); chromosomal aberrations in Chinese hamster ovary (CHO) cells (mammalian cells, chromosomal effects); mutation in mouse lymphoma cells (mammalian cells, gene mutation) and sister chromatid exchanges in CHO cells (mammalian cells, chromosomal effects). Table 2 summarizes the main results of the NTP comparative study. The main outcome was that only the Ames test has a strong statistically significant association with the rodent bioassay. Chromosomal Aberrations in CHO cells has a weaker correlation (but it was not complementary to Salmonella in a battery approach), whereas the two other tests have no correlation at all [12,24]. As a matter of fact, in the NTP database it is possible to distinguish: i) a group of chemicals positive in all four in vitro assays, most of these chemicals being also carcinogens; ii) another group of chemicals negative in all the assays and iii) chemicals negative in the Ames test and positive in one or all the three other assays: these chemicals are mostly noncarcinogens. Thus, whereas the Ames test has a strong positive predictivity for the rodent bioassay (chemicals positive in Salmonella have a high probability of being also carcinogenic), the three other in vitro genotoxicity assays are prone -- to a large extent -- to generate false positive predictions of carcinogenicity [19,21]. The absence of complementarity between the Ames test and, for example, the mouse lymphoma mutation assay in identifying the chemical carcinogens is detailed more clearly in Table 3. The data refer to the NTP chemicals negative in the Ames test: out of 77 Ames negatives, there are similar numbers of rodent carcinogens (45%) and noncarcinogens (55%). Table 3 shows that the mouse lymphoma assay does not identify efficiently either the carcinogens or the noncarcinogens (as a matter of fact, the majority of noncarcinogens are positive in the assay). As a result, there is no statistically significant correlation between rodent carcinogenicity and the mouse lymphoma assay (chi-square = 0.25; p = 0.62), which is not able to identify correctly the carcinogens missed by the Ames test.

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

811

Alternatives to the carcinogenicity bioassay

Table 1. Carcinogenicity rodent bioassay versus Ames test results. Carcinogenicity

Negatives Positives Total

Ames test Negatives

Positives

Total

227 209 436

76 311 387

303 520 823

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

The concordance between the two assay results was calculated on the basis of data collected in the ISSCAN v3a database (see details in the text).

Table 2. Carcinogenicity prediction by STTs in a NTP comparative exercise. Chi-square (p values) STT Ames test Chromosomal aberrations (CHO cells) SCE (CHO cells) Mouse lymphoma cells mutation STTs batteries Ames + chromosomal aberrations Ames + mouse lymphoma mutation

< 0.0001 0.011 0.277 0.305 0.0004 0.120

The table summarizes the main results of a NTP comparative exercise on the prediction of rodent carcinogenicity [12]. The statistics were re-calculated by us, based on the original data. CHO: Chinese hamster ovary; NTP: National Toxicology Program; SCE: Sister chromatid exchange; STT: Short-term test.

Table 3. MLY mutation assay versus carcinogenicity in a database of Ames test negatives. Carcinogenicity

Negatives Positives Total

MLY Negatives

Positives

Total

18 17 35

24 18 42

42 35 77

In order to assess the ability of the MLY mutation assay to complement the Ames test in predicting rodent carcinogenicity, the concordance between the two systems was calculated based on the subset of chemicals resulted to be Ames-negative in the NTP comparative exercise database. MLY: Mouse lymphoma; NTP: National Toxicology Program.

3.4 Looking for in vivo confirmation of in vitro mutagens

Another area that still requires elucidation and improvement is that of the in vivo genotoxicity assays as confirmation of in vitro positive genotoxicity results. The underlying rationale is that in vitro mutagenicity results should be checked in systems where the absorption, distribution, metabolism and excretion characteristics of mammals are operating, thus, 812

providing a more realistic experimental setting than the ‘simple’ in vitro assays. The most commonly used in vivo genotoxicity assay to confirm positive in vitro results is a test for the detection of damage to chromosomes or the mitotic apparatus, namely the mammalian erythrocyte micronucleus test in mice. There is growing evidence that this assay is quite insensitive and responds negatively to various carcinogens that are mutagenic in the Salmonella assay [25]. Table 4 reports the relationship between carcinogenicity and in vivo micronucleus results for a set of chemicals positive in Salmonella. It should be noted that these in vitro positives are the ideal candidates for the in vivo follow-up recommended by various regulatory schemes. The table shows that the in vivo micronucleus test is not able to discriminate adequately between carcinogens and noncarcinogens in the set of Salmonella mutagens. It should be emphasized that the in vivo micronucleus assay shows the same inability to discriminate in the set of chemicals positive in the in vitro chromosomal aberrations test, which is a closer genetic end point [25]. Thus, whereas a positive in vivo micronucleus is predictive of carcinogenicity (but only confirms the positivity in vitro), a negative result does not rule out a carcinogenic potential and should be investigated further (perhaps with other in vivo assays, if necessary). To this aim, several attempts are going on, for example, consideration of other assays such as the Comet assay in vivo or the extension of the micronucleus assay to a larger range of organs [17]. New approaches to alternative methods As pointed out in the previous section, many problems are still open in the area of predicting carcinogenicity with experimental systems. As a matter of fact, the difficulty of predicting in vivo assay results with in vitro systems is common to most of the toxicological end points [26]. Recently, new impetus to the research on alternative toxicological methods has been given by the ToxCast project [27-29]. The pathway-based screening paradigm adopted by ToxCast is radically different from traditional in vitro testing. Basically, the perturbations provoked by chemicals to biochemical and biological pathways supposed to be critical to toxicity are analyzed. Such perturbations are studied in isolated systems in vitro (both cell-free and cell-based), with the use of modern highthroughput screening omics techniques. ToxCast Phase I considered 320 chemicals (309 unique structures, mainly agrochemicals) for which in vivo toxicity data (carcinogenicity, developmental and reproductive toxicity) were already available: > 500 omics assays were run on the chemicals in order to check the agreement between animal toxicity and omics assays. We analyzed the recently released results of ToxCast Phase I, with special attention to the in vitro/ in vivo relationship (Benigni et al., http://www.epa.gov/ NCCT/toxcast/files/summit/ToxcastDataSummit_Poster_ Benigni%20May2009.ppt). It appears that the correlation in vitro/in vivo is extremely low, or absent, depending on the in vivo toxicity end point considered. This evidence is in 3.5

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

Benigni, Bossa, Tcheremenskaia & Giuliani

Table 4. Carcinogenicity vs in vivo micronucleus in a database of Ames mutagens. In vivo micronucleus

Carcinogenicity

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

Negatives Positives Total

Negatives

Positives

Total

7 34 41

2 32 34

9 66 75

The relationship between the two systems is calculated based on a subset of chemicals positive in the Ames test. In most testing strategies and regulatory schemes, these in vitro positives are the natural candidates for in vivo follow-up. Details on the analysis and data are in [25].

agreement with findings from other fields, ranging from research on drug design with intensive use of omics technologies [30,31] to more traditional research on alternative tests for regulatory purposes [26]: isolated systems in vitro -- when perturbed by chemicals -- respond in a way largely different from how they respond when they are integrated into whole organisms. From a practical point of view, the above evidence indicates that we are still quite far from being able to replace the classical animal toxicity assays with stand-alone in vitro alternatives.

Formalizing the chemistry of carcinogenicity

4.

The structural alerts In parallel to the research on short-term experimental alternatives to the rodent bioassay, the mechanistic knowledge on chemical carcinogens was used to establish SAR. An important contribution came from John Ashby, which resulted in the definition and compilation of structural alerts (SA) following the electrophilicity theory of the Millers [11,32]. The SAs for carcinogenicity are defined as molecular functional groups or substructures that are mechanistically linked to the carcinogenic activity of the chemicals. Thus, the SAs code for chemical classes potentially able to cause cancer. Because the attack on and the modification of DNA is considered to be a main step in the mechanism of action of many carcinogens (i.e., the so-called genotoxic carcinogens), the SAs relative to such classes of carcinogens are also valid for the Salmonella mutagenicity end point. It should be emphasized that SAs, together with being an important scientific systematization of the knowledge on mechanisms of toxic action, hold a special place in predictive toxicology. This knowledge is routinely used in SAR assessment in the regulatory context (see, e.g., the mechanistically-based reasoning as presented in [33]). In addition, the SAs are at the basis of popular commercial (e.g., DEREK, by Lhasa Ltd) and noncommercial software systems (e.g., Oncologic, by US Environmental Protection Agency http://www.epa.gov/oppt/newchems/tools/oncologic.htm). 4.1

Thus, the SAs represent a coarse-grain approach to (Q)SAR applications, useful for, example, in identifying mechanistic and regulatory classes of chemicals, and for setting priorities. In this sense, they are opposite and complementary to the fine tuned QSAR models for congeneric classes of chemicals [34]. In terms of practical implementation of the SAs, the role of the modulating factors should be recalled as well. Each SA code for a chemical class (e.g., aromatic amine) is potentially toxic. But not all aromatic amines are carcinogenic or mutagenic. Depending of the characteristics of the rest of the molecule (e.g., the presence of bulky substituents close to the nitrogen), the potential effect of the reactive group may be diminished or completely eliminated. Thus, the knowledge of modulating factors is an integral part of the implementation of the SAs for risk assessment and contributes to substantially increasing their specificity [35]. Following the SA compilation by Ashby, other authors tried to update the list of chemical classes whose ability to induce cancer (and mutations) is recognized (e.g., [10,36]). These more recent compilations incorporate also a number of SAs for nongenotoxic mechanisms of carcinogenesis. Structural alerts and their predictive power In this paper, the compilation of SAs recently implemented into the expert system Toxtree 1.60 is used for a number of analyses [35]. Toxtree (http://ecb.jrc.ec.europa.eu/qsar/qsartools/index.php?c=TOXTREE) is an open-source, freely available software application that places chemicals into categories and predicts various kinds of toxic effect by applying various decision tree approaches. Toxtree was developed by IdeaConsult Ltd (Sofia, Bulgaria) under the terms of an ECB contract [37]. Figure 1 (in the form of a radius of curvature (ROC) graph) displays the agreement between the SAs for carcinogenicity/ mutagenicity implemented in Toxtree, and the two toxicological end points of carcinogenicity and Ames mutagenicity (based on the chemicals contained in the ISSCAN v3a database). An ROC graph has 1 minus specificity (false positive rate) on the X-axis and sensitivity (true positive rate) on the Y-axis. Figure 1 shows that the SAs have both high sensitivity and specificity for the Ames test (overall accuracy = 0.79), while having a lower agreement with carcinogenicity (overall accuracy = 0.70). For obvious reasons, in the analysis of the mutagenicity results only the SAs for genotoxic action mechanisms were considered. Thus, the SAs originally developed for carcinogenicity agree better with Ames mutagenicity than with carcinogenicity itself. This is explained by the fact that most of the carcinogens known when the SAs and the Ames test were originally developed were genotoxic carcinogens, and so both models are tailored over the same type of mechanistic knowledge on action mechanisms. Because several carcinogens known today are nongenotoxic, the SAs correlate with the 4.2

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

813

Alternatives to the carcinogenicity bioassay

ISSCANv3a: SA-BB

0.6

of uncertainty similar to (or lower than) that of good experimental tests. An additional observation is that the SAs with high positive predictivity for carcinogenicity do not belong to specific areas of toxic mechanisms/pathways, but span different categories, and include both direct and indirect-acting agents.

0.4

4.3

1.0 Ames test True positive rate

0.8 Rodent carcinogenicity

The overlapping information from biology (Ames test) and chemistry (structural alerts)

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

0.2 0.0 0.0

0.2

0.4

0.6

0.8

1.0

False positive rate

Figure 1. Concordance of the structural alerts with the Ames test and rodent carcinogenicity.

Ames test results better than with the database of chemical carcinogens available today. It should be emphasized that the SAs predictive ability for Salmonella mutagenicity (accuracy = 0.79) is of the same order of magnitude of the experimental variability of the test itself (inter-laboratory reproducibility reported to be 80 -85%; [38]). This implies that: i) the assessment of chemical mutagenicity through the Ames test and through the SAs have similar reliability; ii) the knowledge on the mechanisms of Salmonella mutagenicity is largely exhaustive and well represented by the SAs; and iii) as a consequence of ii), both the presence and absence of SAs in a molecule have predictive power. On the contrary, the overall predictivity of the SAs for carcinogenicity is lower: the lack of SAs in a chemical is not indicative -- in any way -- of lack of potential toxicity, whereas the presence of SAs is strongly suggestive of potential carcinogenicity. This is better detailed in Table 5, which reports the positive predictivity of the individual SAs. Positive predictivity is the probability that a chemical is carcinogenic when it contains a given SA; it is calculated as: ratio of the number of positive chemicals with a given SA to the number of all chemicals with the same SA. The calculations in Table 5 are based on the carcinogenicity data contained in the ISSCAN v3a database. By considering only SA classes with > 10 chemicals in the database, Table 5 shows that the presence of SAs is highly predictive of the carcinogenic potential of various classes, for example: i) the alkylating agents N-nitroso compounds (SA_2), quinones (SA_12), hydrazines (SA_13), and esters of sulfonic or phosphonic acid (SA_21); ii) the heterocyclic PAH intercalating agents (SA_19); and iii) the arylaminoDNA-adducts forming aromatic amines (SA_28). It should be remarked that the positive predictivity of the selected SAs is higher than 80%, which corresponds to a level 814

Given the preponderant role of the Ames test as in vitro predictor of rodent carcinogenicity, it is of interest to analyze whether its predictivity can be improved by the combination with tools such as the SAs. Figure 2 displays the result of such an analysis. It appears that the Ames test and the SAs in Toxtree have a similar overall performance, with the Ames test being more specific and the SAs being more sensitive. The combination of both in a battery improves very little the overall sensitivity, with an equivalent loss of specificity. This is a typical effect happening when two variables carrying the same type of information are combined. As a matter of fact, historically the Ames and the SAs were both tailored over the knowledge on the carcinogenicity mechanisms available when the Millers performed their research, and so they can be considered as different, but parallel models of the same phenomenon. Perhaps, the Ames test has an unquestionable superiority only when applied to the study of mixtures of unknown composition. QSAR models Whereas the presence of an SA in a molecule defines only a potential for being carcinogenic or mutagenic, the actual modulation of this potential depends on a series of factors (e.g., molecular mass, physical state, solubility, degree of chemical reactivity, pattern of substituents) which vary within each individual class of compounds. To a certain extent, such modulating factors can be approximated by minor, contextdependent SAs (e.g., the substituents in the different positions of the skeleton of a chemical with a primary SA). A powerful generalization is provided by the QSAR analysis, which is based on a limited number of physical chemical or structural properties with general relevance, and produces a mathematical model of the chemical determinants of the biological activity. The physical chemical properties of interest for the biological activity are hydrophobic, electronic and steric effects [34]. Building and evaluating a QSAR model is a complex procedure, usually consisting of a series of steps that can be performed according to different approaches. The selection of the approach depends on both the specific data in the hands of the investigators and on their preferences (and availability of tools). Good introductory guides on how to obtain a QSAR model can be found, for example, in [34,39]. In brief, at the beginning of the investigation one has, for a set of compounds, the values of the observed biological activity and the chemical structures. The first step of the analysis is the 4.4

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

Benigni, Bossa, Tcheremenskaia & Giuliani

Table 5. Structural alerts for carcinogenicity: positive predictivity.

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

Chemicals with SA

Actives

Acylating, direct acting SA_1 1 1 SA_15 5 3 Alkylating, direct acting SA_2 13 12 SA_3 2 2 SA_5 10 10 SA_6 4 4 SA_7 28 20 SA_8 75 51 SA_9 1 1 SA_10 55 37 SA_11 9 8 SA_12 15 13 Alkylating, indirect acting SA_4 6 6 SA_13 68 57 SA_14 8 8 SA_16 8 7 SA_21 120 105 SA_22 5 3 SA_23 5 5 SA_24 3 3 Intercalating and DNA-adducts forming, indirect acting SA_18 14 11 SA_19 14 12 SA_30 6 5 Aminoaryl DNA-adducts forming, indirect acting SA_25 3 3 SA_26 3 2 SA_27 88 63 SA_28 107 87 SA_28bis 15 10 SA_28ter 17 13 SA_29 26 20 Nongenotoxic SA_17 24 13 SA_20 18 14 SA_31a 16 5 SA_31b 10 8 SA_31c 4 2

% 100 60 92 100 100 100 71 68 100 67 89 87 100 84 100 88 88 60 100 100 79 86 83 100 67 72 81 67 76 77 54 78 31 80 50

The table displays the positive predictivity, with respect to rodent carcinogenicity, of the individual SAs implemented in the Toxtree rulebase. The SAs are coded as in Toxtree [35] and are collected into broad mechanistic categories. The positive predictivity is calculated over the ISSCAN v3a database. SA: Structural alert.

calculation -- through one of the many specialized softwares -- of a series of parameters that describe the structure and/or the physical chemical properties of the chemicals. Then mathematics and statistics are applied in order to discover which (if any) out of the calculated chemical parameters correlates with the biological activity, and a mathematical model is established. The last step is the validation of the QSAR model: very often, the validation is performed through

different statistical analyses (e.g., cross-validation, scrambling [40]), but the ultimate proof of the general validity of a QSAR model is the proof that, when applied to a new set of chemicals not used for the modeling, it predicts correctly their biological activity based on the knowledge of their chemical structure [41,42]. A great aspect of the QSARs for the individual chemical classes is that they point to the physical chemical determinants of the biological activity of the compounds; thus, they can considerably contribute to the understanding of the mechanisms of chemical mutagenicity and carcinogenicity. It should be added as well that this approach is mostly powerful when applied to congeneric groups of chemicals, that is, molecules belonging to the same chemical class and acting through the same mechanism of action (Hansch or extra-thermodynamic approach to QSAR) [34,39]. QSARs have been generated for a number of individual classes of chemical carcinogens and mutagens (including aromatic amines, nitroarenes, quinolines, triazenes, polycyclic aromatic hydrocarbons, lactones) [43-52]. The majority are relative to in vitro mutagenicity; however, a number of QSAR models for the animal carcinogenicity are available as well. Recently, the expanding role of QSAR for regulatory purposes has brought out more strongly the need for elucidation of the issue of predictivity: within this context, a survey on the (Q)SARs for mutagens and carcinogens in the public domain has been performed in a collaborative effort between the Istituto Superiore di Sanita and the European Chemicals Bureau. The details of the study are published in [53] and are put into a wider perspective in [41]. The analysis considered both QSARs that described the gradation of potency of toxic chemicals (mutagens or carcinogens) and QSARs aimed at discriminating between active and inactive chemicals. The biological activities included Salmonella mutagenicity and rodent carcinogenicity, for the classes of aromatic amines, nitroarenes and aliphatic aldehydes. The QSARs for potency (applicable only to toxic chemicals) generated predictions 30 -- 70% correctly, whereas the QSARs for discriminating between active and inactive chemicals were 70 -- 100% correct in their external predictions. It should be emphasized that this study challenged the QSAR models to predict rigorously external test sets, that is, chemicals belonging to the classes for which the QSARs were derived, but not included in the training sets used by the authors of the models. The study also showed that statistical validation methods (e.g., cross-validation, leave-oneout, leave-many-out, re-sampling), which are often assumed to be good diagnostics for predictivity, did not correlate well with the predictivity of the QSARs when challenged in external prediction tests. It should be emphasized that the external predictivity of activity models is quite high (70 -- 100% accuracy). This is even more remarkable when compared with the range of the variability of the experimental tests (e.g., the 80 -- 85% inter-laboratory repeatability of the Ames test [38]). Thus,

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

815

Alternatives to the carcinogenicity bioassay

1.0 SAs + Ames

True positive rate

0.8

SAs

0.6 Ames 0.4

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

0.2 0.0 0.0

0.2

0.4

0.6

0.8

1.0

False positive rate

Figure 2. Concordance of the Ames test, the structural alerts and their combination, with the rodent carcinogenicity.

the level of uncertainty of ‘good’ QSARs is comparable to that of ‘good’ experimental assays. 5.

Conclusions

The need for creating tools able to predict chemical carcinogens in shorter times and at a lower cost in terms of animal lives and money is still, after several decades, a research priority. New regulatory requirements (e.g., REACH) have even increased expectations in this field. Historically, the convergence between the basic genetic research on chemicallyinduced mutagenesis and the work of the Millers on the electrophilic, DNA-reactive chemical carcinogens has stimulated the scientific community to privilege the search for mutation-based STTs over other possible approaches. The above research, lasting since the 1970s, has both produced great successes and pointed to evident limitations of the available scientific concepts. In particular, the continued work to develop mutagenicitybased STTs to predict carcinogenicity has generated useful results only for a limited area of the chemical space, that is, those chemicals that are able to interact with the DNA. The DNA-reactive chemicals are able to induce both cancer and a wide spectrum of mutations, that is, gene mutations (involving one or two bases, and small deletions and insertions) and chromosome mutations (which can be structural or numerical changes). The assay that has, by and large, the best correlation with carcinogenicity is the Ames test, which was specifically engineered in such a way as to be sensitive to the chemical classes for which the hypothesis DNA-reactivity -mutation -- cancer seems valid. Thus, the Ames test is a kind of mirror of the chemistry of DNA interaction. It should be mentioned as well that the Ames-test results can be predicted with around 80% accuracy using the SAs for genotoxic carcinogenicity [10], that is, with an uncertainty 816

comparable to its inter-laboratory reproducibility. To close the circle, it should be recalled that the SAs predict the carcinogenicity potential at a level comparable to that of Salmonella in the available database of bioassay results (Figure 2). The high positive predictivity of many SAs (as shown in Table 5) supports and details the above evidence. This implies that -- for the prediction of carcinogenicity -SAR concepts are both a valid substitute of the Salmonella test, and the foundation of the Salmonella assay itself. Thus, the data confirm the reliability of the translation of the Millers hypothesis on the DNA-interaction mediated carcinogenesis into a biological test. However, at odds with the hopes of the developers of many STTs, the link between mutation and carcinogenesis is not a necessary mechanistic link, independent of the type of chemicals and suited as general screening criterion of the carcinogenic potential for regulatory purposes, but is a specific feature of a limited number of carcinogens. For the non-DNA-reactive chemicals that are negative in Salmonella, and are mutagenic only in in vitro assays such as, for example, chromosomal aberrations or mouse lymphoma mutation assays, no correlation with carcinogenicity is apparent (Table 3). This discrepancy between the intended role of the additional (to Salmonella) genotoxicity tests and their actual response to carcinogens and noncarcinogens is still to be solved, and often creates problems in the interpretation of contradictory in vitro genotoxicity results. The main controversy regards the role of the in vitro assays in cultured mammalian cells (e.g., mouse lymphoma mutation, chromosomal aberrations in CHO cells). Because many positive results are not correspondent with positive carcinogenicity results, some authors propose to discount the importance of the additional in vitro assays. On the contrary, other authors believe that agents negative in rodent bioassays and positive in in vitro tests are not necessarily false positives and support the retention of the additional in vitro genetic toxicity tests as sentinels of genotoxic effects that merit further investigation [54]. Another open issue regards the role of in vivo assays as confirmation of in vitro positive results. As shown in Table 4, the in vivo micronucleus assay -- as it is now -- is not an adequate tool, because it misses too many rodent carcinogens [25]. For applicative purposes, a chemical mutagenic in Salmonella and/or with SAs should be seriously considered as a potential carcinogen. Unfortunately, at present, no reliable, widely-accepted alternative tools exist to assess the risk of chemicals outside the domain of the above characteristics, that is, the risk posed by nongenotoxic carcinogens. Thus, the in vivo long-term experimentation is still the ultimate way to assess whether the non-DNA-reactive, Salmonella-negative chemicals are really negatives or are nongenotoxic carcinogens. It should be emphasized as well that overall these results point to the importance of exploiting structure--activity concepts for giving some general guidance to the generation

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

Benigni, Bossa, Tcheremenskaia & Giuliani

of a ‘safer chemistry’; in fact, the DNA-reactivity SAs provide a tool to decide whether a chemical can be assessed through SAs and or/Salmonella or needs to be studied with different (longer-terms) assays. One last point to be emphasized is that the present integrated view on the alternatives to the rodent bioassay has been made possible by the exploitation of a wide range of in silico tools. Most notably, these tools include expert systems (e.g., Toxtree) that code for mechanistic knowledge on the chemistry of carcinogenicity (in the form of SAs) and chemical relational databases, such as ISSCAN. A chemical relational database is a special type of relational database whose main informational unit is a chemical structure and whose fields are attributes or data (e.g., toxicity) associated with that chemical structure. Whereas the indexing (identifier) elements in traditional databases, such as names and CAS numbers, are non-unique, prone to errors and devoid of intrinsic information, chemical structure as a chemical identifier has a universally understood meaning and scientific relevance. The combination of information technology with the chemical structure as identifier of the molecules permits an enormous range of operations (e.g., retrieving chemicals or chemical classes, describing the content of databases, finding similar chemicals, crossing biological and chemical interrogations) that other more classical databases cannot allow. In this paper, the use of this modern in silico technology has permitted the rapid assessment of, for example, the predictivity of the SAs both individually and as a whole (Figure 1 and Table 5), as well as of the complementarity between the SAs and the Ames test (Figure 2). 6.

Expert opinion

The biggest success of the research on the mutagenicity-based alternative assays has been the generation of the Salmonella typhimurium (Ames) test, which is able to identify with great efficiency the DNA-reactive carcinogens. On the other hand, no mutagenicity-based assays complementary to the Ames test for identifying the non-DNA-reactive carcinogens have been found. Chemicals negative in Salmonella and positive only for other genetic end points (e.g., clastogenicity, aneuploidy) have not a significantly high probability of being carcinogenic. The above evidence, overall, seems to question the wellestablished general equation mutation = cancer. Within this perspective, the link between Ames test and carcinogenesis is probably not rooted in the biological process of carcinogenesis that necessarily would follow the specific mutational events recognized by the test, but in the ability to respond to general chemical physical properties of organic compound classes able to react with DNA. The above concept finds support in recent research in tumor development that stresses the two complementary

aspects of microenvironment and attractor-like dynamics of tumor progression [55,56], and suggests that the initial stimulus to cancer development is largely aspecific and noisy and, more importantly, not necessarily restricted to the single cell-, single event level. If this is the case, the search for ‘all the possible effective perturbations’ could be a never ending task and could give rise mainly to ‘chance correlations’. In our opinion, two avenues may be fruitful for future research. The first approach is more operational and can give results in a short time. It involves the recognition of individual classes of nongenotoxic carcinogens, and the coding of this chemical knowledge into SAs and QSARs, which can be implemented into expert systems (such as Toxtree). Another avenue is that of looking for assays of more general applicability. Possible candidates are the cell transformation assays. These STTs are not directly based on the concept of genetic mutation and mimic some stages of in vivo multistep carcinogenesis. Cell transformation has been defined as the induction of certain phenotypic alterations in cultured cells that are characteristic of tumorigenic cells [57]. These phenotypic alterations can be induced by exposing mammalian cells to carcinogens. Transformed cells that have acquired the characteristics of malignant cells have the ability to induce tumors in susceptible animals [58,59]. The CTAs have been proposed for predicting carcinogenic potential of chemicals for many years; however, they have undergone different cycles of favor and disfavor among the scientific community and have never been consistently included in the regulatory testing schemes. More experimentation, with larger sets of chemicals, could provide useful information on the reliability of the cell transformation assays. Another good ‘middle term’ candidate could be based on in vivo metabolomics. These studies are already producing very promising results in both cancer staging and identification [60]. In the toxicological realm this approach is still in its infancy [61]; however, it holds promises for drastic animal reduction. As a matter of fact, it has been known for decades that the detection of general metabolic effects of cancer, such as the Warburg oxidative glycolysis typical of tumors [62], can be used to discriminate tumor and normal tissues, and thus could be at the basis of early predictions of cancer development. In our opinion these ‘hybrid’ (biological fluids ex vivo), non-invasive approaches could open promising avenues of research in the next few years.

Declaration of interest C Bossa received funding from the EU OSIRIS FP6 project. O Tcheremenskaia received funding from the EU OpenTox FP7 project. The other authors state no conflict of interest and have received no payment in preparation of this manuscript.

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

817

Alternatives to the carcinogenicity bioassay

Bibliography 1.

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

818

genotoxic carcinogenesis among 222 chemicals tested by the U.S.NCI/ NTP. Mutat Res 1988;204:17-115

Huff J. Value, validity, and historical development of carcinogenesis studies for predicting and confirming carcinogenic risks to humans. In: Kitchin KT, editor, Carcinogenicity. Testing, predicting, and interpreting chemical effects. Marcel Dekker, Inc., New York; 1999. p. 21-123

12.

Huff J. Long-term chemical carcinogenesis bioassays predict human cancer hazards. Issues, controversies, and uncertainties. Ann NY Acad Sci 1999;895:56-79

13.

Miller JA, Miller EC. Ultimate chemical carcinogens as reactive mutagenic electrophiles. In: Hiatt HH, Watson JD, Winsten JA, editors, Origins of human cancer. Cold Spring Harbor: Cold Spring Harbor Laboratory; 1977. p. 605-28

14.

Miller EC, Miller JA. Searches for ultimate chemical carcinogens and their reactions with cellular macromolecules. Cancer 1981;47:2327-45 Arcos JC, Argus MF. Multifactor interaction network of carcinogenesis - a “Tour Guide”. In: Arcos JC, Argus MF, Woo YT, editors, Chemical induction of cancer. Modulation and combination effects. Birkhauser, Boston; 1995. p. 1-20 Woo YT. Mechanisms of action of chemical carcinogens, and their role in Structure-Activity Relationships (SAR) analysis and risk assessment. In: Benigni R, editor, Quantitative Structure-Activity Relationship (QSAR) models of mutagens and carcinogens. CRC Press, Boca Raton; 2003. p. 41-80 Ames BN. The detection of environmental mutagens and potential carcinogens. Cancer 1984;53:2030-40

15.

16.

17.

18.

Maron DM, Ames BN. Revised methods for the Salmonella mutagenicity test. Mutat Res 1983;113:173-215 Zeiger E. Carcinogenicity of mutagens: predictive capability of the Salmonella mutagenesis assay for rodent carcinogenicity. Cancer Res 1987;47:1287-96 Benigni R, Bossa C. Structure Alerts for carcinogenicity, and the Salmonella assay system: a novel insight through the chemical relational databases technology. Mutat Res Revs 2008;659:248-61 Ashby J, Tennant RW. Chemical structure, Salmonella mutagenicity and extent of carcinogenicity as indicators of

19.

20.

21.

and performance. Regul Pharmacol Toxicol 1998;28:85-95 22.

Zeiger E, Haseman JK, Shelby MD, et al. Evaluation of four in vitro genetic toxicity tests for predicting rodent carcinogenicity: confirmation of earlier results with 41 additional chemicals. Environ Mol Mutagen 1990;16:1-14

Zeiger E. Mutagens that are not carcinogens: faulty theory or faulty tests? Mutat Res 2001;492:29-38

23.

Fetterman BA, Soo Kim B, Margolin BH, et al. Predicting rodent carcinogenicity from mutagenic potency measured in the Ames salmonella assay. Environ Mol Mutagen 1997;29:312-22

Benigni R, Bossa C, Richard AM, Yang C. A novel approach: chemical relational databases, and the role of the ISSCAN database on assessing chemical carcinogenicity. Ann Ist Super Sanita` 2008;44:48-56

24.

Tennant RW, Margolin BH, Shelby MD, et al. Prediction of chemical carcinogenicity in rodents from in vitro genetic toxicity assays. Science 1987;236:933-41

25.

Benigni R, Bossa C, Worth AP. Structural analysis and predictive value of the rodent in vivo micronucleus assay. Mutagenesis 2010. [Epub ahead of print]

26.

EFSA. Existing approaches incorporating replacement, reduction and refinement of animal testing: applicability in food and feed risk assessment. EFSA J 2009;1052:1-57

27.

Dix DJ, Houck KA, Martin MT, et al. The ToxCast program for prioritizing toxicity testing of environmental chemicals. Toxicol Sci 2007;95:5-12

28.

Schmidt C. Testing for carcinogens: shift from animals to automation gathers steam--slowly. J Natl Cancer Inst 2009;101:910-2

29.

Kavlock R, Austin CP, Tice RR. Toxicity testing in the 21st century: implications for human health risk assessment. Risk Anal 2009;29:485-7

30.

Young D. Drug innovation on the decline. Am J Health Syst Pharm 2009;64:228-31

31.

Benigni R. Relationships between in vitro mutagenicity assays. Mutag 1992;7:335-41

MacDonald JS, Robertson RT. Toxicity testing in the 21st century: a view from the pharmaceutical industry. Toxicol Sci 2009;110:40-6

32.

Benigni R, Giuliani A. Quantitative modeling and biology: the multivariate approach. Am J Physiol 1994;266:R1697-704

Ashby J. Fundamental structural alerts to potential carcinogenicity or noncarcinogenicity. Environ Mutagen 1985;7:919-21

33.

Woo YT, Lai DY, McLain JL, et al. Use of mechanism-based structure-activity relationships analysis in carcinogenic potential ranking for drinking water disinfection by-products. Environ Health Perspect 2002;110:75-87

Zeiger E. History and rationale of genetic toxicity testing: an impersonal, and sometimes personal, view. Environ Health Perspect 2004;44:363-71 Ashby J, Waters MD, Preston J, et al. IPCS harmonization of methods for the prediction and quantification of human carcinogenic/mutagenic hazard, and for indicating the probable mechanism of action of carcinogens. Mutat Res Fundam Mol Mech Mutagen 1996;352:153-7 Combes R, Grindon C, Cronin MTD, et al. Proposed integrated decision-tree testing strategies for mutagenicity and carcinogenicity in relation to the EU REACH legislation. ATLA 2007;35:267-87 Kirkland D, Speit G. Evaluation of the ability of a battery of three in vitro genotoxicity tests to discriminate rodent carcinogens and non-carcinogens III. Appropriate follow-up testing in vivo. Mutat Res 2008;654:114-32 Lilienblum W, Dekant W, Foth H, et al. Alternative methods to safety studies in experimental animals: role in the risk assessment of chemicals under the new European Chemicals Legislation (REACH). Arch Toxicol 2008;82:211-36

Zeiger E. Identification of rodent carcinogens and noncarcinogens using genetic toxicity tests: premises, promises,

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

Benigni, Bossa, Tcheremenskaia & Giuliani

34.

Expert Opin. Drug Metab. Toxicol. Downloaded from informahealthcare.com by Universitaet Zuerich on 06/23/14 For personal use only.

35.

36.

37.

38.

39.

40.

41.

42.

Hansch C, Leo A. Exploring QSAR. 1. Fundamentals and applications in chemistry and biology. American Chemical Society, Washington, D.C.; 1995 Benigni R, Bossa C, Jeliazkova NG, et al. The Benigni/Bossa rulebase for mutagenicity and carcinogenicity--a module of Toxtree. EUR 23241 EN. Luxenbourg, Office for the Official Publications of the European Communities. EUR--Scientific and Technical Report Series. 2008 Woo YT, Lai DY, Argus MF, Arcos JC. Development of structure-activity relationship rules for predicting carcinogenic potential of chemicals. Toxicol Lett 1995;79:219-28 Worth AP, Bassan A, de Brujin J, et al. The role of the European chemicals bureau in promoting the regulatory use of (Q)SAR methods. SAR QSAR Environ Res 2007;18:111-25 Piegorsch WW, Zeiger E. Measuring intra-assay agreement for the Ames Salmonella assay. In: Rienhoff O, Lindberg DAB, editors, Statistical methods in toxicology. Springer-Verlag, Heidelberg; 1991. p. 35-41 Franke R, Gruska A. General introduction to QSAR. In: Benigni R, editor, Quantitative Structure-Activity Relationhsip (QSAR) models of mutagens and carcinogens. CRC Press, Boca Raton; 2003. p. 1-40

43.

Hansch C. Structure-activity relationships of chemical mutagens and carcinogens. Sci Total Environ 1991;109/110:17-29

44.

Cronin MTD, Dearden JC. QSAR in toxicology. 3. Prediction of chronic toxicities. Quant Struct Act Relat 1995;14:329-34

45.

Debnath AK, Debnath G, Shusterman AJ, Hansch C. A QSAR investigation of the role of hydrophobicity in regulating mutagenicity in the Ames test: 1. Mutagenicity of aromatic and heteroaromatic amines in Salmonella typhimurium TA98 and TA100. Environ Mol Mutagen 1992;19:37-52

Benigni R, Bossa C, Netzeva TI, Worth AP. Collection and evaluation of (Q)SAR models for mutagenicity and carcinogenicity. EUR 22772 EN. 2007. Luxenbourg, Office for the Official Publications of the European Communities. EUR--Scientific and Technical Research Series. 12-7-2007

54.

Elespuru RK, Agarwal R, Atrakchi AH, et al. Current and future application of genetic toxicity assays: the role and value of in vitro mammalian assays. Toxicol Sci 2009;109:172-9

55.

Kalluri R, Zeisberg M. Fibroblasts in cancer. Nat Rev Cancer 2009;6:392-401

56.

46.

Debnath AK, Shusterman A, Lopez de Compadre RL, Hansch C. The importance of the hydrophobic interaction in the mutagenicity of organic compounds. Mutat Res 1994;305:63-72

Brock A, Chang H, Huang S. Non-genetic heterogeneity--a mutation-independent driving force for the somatic evolution of tumours. Nat Revs Genetics 2009;10:336-49

57.

47.

Smith CJ, Hansch C, Morton MJ. QSAR treatment of multiple toxicities: the mutagenicity and cytotoxicity of quinolines. Mutat Res 1997;379:167-75

Barrett JC, Ts’o PO. Evidence for the progressive nature of neoplastic transformation in vitro. Proc Natl Acad Sci USA 1978;75:3761-5

58.

48.

Lopez de Compadre RL, Debnath AK, Shusterman AJ, Hansch C. LUMO Energies and hidrophobicity as determinants of mutagenicity by nitroaromatic compounds in Salmonella typhimurium. Environ Mol Mutagen 1990;15:44-55

Berwald Y, Sachs L. In vitro transformation of normal cells to tumour cells by carcinogenic hydrocarbons. J Natl Cancer Inst 1965;35:641-61

59.

Berwald Y, Sachs L. In vitro cell transformation with chemical carcinogens. Nature 1963;200:1182-4

60.

Tiziani S, Lopes V, Gunther UL. Early stage diagnosis of oral cancer using 1H NMR-based metabolomics. Neoplasia 2009;11:269-76

61.

Griffin JL, Bollard ME. Metabonomics: its potential as a tool in toxicology for safety assessment and data integration. Curr Drug Metab 2004;5:389-98

62.

Kim J-W, Dang CV. Cancer’s molecular sweet tooth and the warburg effect. Cancer Res 2006;66:8927-30

49.

Eriksson L, Jaworska JS, Worth AP, et al. Methods for reliability and uncertainty assessment and for applicability evaluations of classification- and regression-based QSARs. Environ Health Perspect 2003;111:1361-75

50.

Benigni R, Bossa C. Predictivity of QSAR. J Chem Inf Model 2008;48:971-80

51.

Kubinyi H. Validation and predictivity of QSAR models. In: Aki-Sener E, Yalcin I, editors, QSAR and Molecular Modelling in rational design of bioactive molecules. The 15th European Symposium on Quantitative Structure-Activity Relationships and molecular modelling. CADDS, Ankara; 2005. p. 30-3

53.

52.

Benigni R, Bossa C, Netzeva TI, et al. Mechanistic QSAR of aromatic amines: new models for discriminating between homocyclic mutagens and nonmutagens, and validation of models for carcinogens. Environ Mol Mutagen 2007;48:754-71 Hansch C, Kurup A, Garg R, Gao H. Chem-bioinformatics and QSAR: a review of QSAR lacking positive hydrophobic terms. Chem Revs 2001;101:619-72 Zhang L, Sannes K, Shusterman AJ, Hansch C. The structure-activity relationships of skin carcinogenicity of aromatic hydrocarbons and heterocycles. Chem Biol Interact 1992;81:149-80 Benigni R. Structure-activity relationship studies of chemical mutagens and carcinogens: mechanistic investigations and prediction approaches. Chem Revs 2005;105:1767-800

Expert Opin. Drug Metab. Toxicol. (2010) 6(7)

Affiliation

Romualdo Benigni†, Cecilia Bossa, Olga Tcheremenskaia & Alessandro Giuliani † Author for correspondence Istituto Superiore di Sanita, Environment and Health Department, Viale Regina Elena 299, 00161 Rome, Italy Tel: +39 06 49902579; Fax: +39 06 49902999; E-mail: [email protected]

819