TY - JOUR
AU - Tcheremenskaia, Olga
AB - Abstract The study of the chemical carcinogenesis mechanisms and the design of efficient prevention strategies and measures are of crucial importance to protect human health. The long-term carcinogenesis bioassays have played a central role in protecting human health, but for ethical and practical reasons their use is dramatically diminishing, and the genotoxicity short-term tests have taken the pivotal role in the pre-screening of carcinogenicity. However, there is evidence that this strategy is not sensitive enough to detect all genotoxic carcinogens and it cannot detect nongenotoxic carcinogens. In a previous article, we have shown that an integrated strategy consisting of the in vitro Ames and Syrian Hamster Embryo cells transformation assays, combined with structure–activity relationships, is a valid alternative to the present pre-screening strategies. Here, we expand the previous investigation by (i) including results of cell transformation assays on inorganics, together with an additional assay (Bhas 42), and (ii) considering new structural alerts for nongenotoxic carcinogenicity. We also present a new analysis on global relationships between toxicological endpoints. The new results confirm that the previously proposed integrated, alternative strategy is an efficient tool to identify both genotoxic and nongenotoxic carcinogens, with an estimated 90–95% sensitivity. Introduction The study of the chemical carcinogenesis mechanisms and the design of efficient prevention strategies and measures are of crucial importance to protect human health (1,2). Cancer is the second leading cause of death in Western countries, just after circulatory diseases (3). It should be emphasised as well that the causes of cancer are predominantly environmental in nature (this includes chemical exposure, lifestyle, diet, and work): different types of evidence converge toward an estimated 80% environmental component of cancer (1,4–8). To identify potential carcinogenic effects for humans, experimental long-term chemical carcinogenesis bioassays are designed and carried out. Carcinogenesis long-term bioassay results in rodents, mainly rats and mice, have been shown to be a consistent and reliable indicator and predictor of human cancer risk (4,9,10–15). For chemicals discovered to be carcinogenic to laboratory animals, prudent public health policy suggests strongly that eliminating exposures to these agents would reduce or eliminate certain environmentally associated cancers (4,16–18). The obvious negative side of the standard bioassay in rodents is that it is time-consuming and costly, and it requires the sacrifice of large numbers of animals. This makes the bioassay not usable in practice for large screening programs [e.g. it is foreseen that in the next few years around 30 000 chemicals will need assessment under the new European legislation called REACH (i.e. Registration, Evaluation, Authorisation and Restriction of Chemicals; (19–21)]. For all these different reasons, chemical carcinogenicity has been the target of numerous attempts to create alternative predictive models, ranging from short-term biological assays (22–24) to theoretical models [namely structure–activity relationships (SAR); (2,25,26)]. Recently, this trend has been strongly stimulated by changes at regulatory level, aiming at drastically reducing the number of new cancer bioassays. Among others, REACH provides a basis for the use of ‘non-testing’ approaches [including (Q)SAR, (quantitative) structure–activity relationships] for predicting the environmental and toxicological properties of chemicals, in the interests of time-effectiveness, cost-effectiveness and animal welfare (26–30). The somatic mutation theory of cancer (31) has been by and large regarded as the predominant paradigm for the induction of cancer by chemicals, and it has inspired the majority of studies as well as the general lines of the pre-screening strategies for cancer (23,24,32). The regulatory schemes and strategies vary to a large extent, depending on the types of chemicals and intended use. However, a dominant trend can be recognised. Most often, a two-tiered integrated testing approach is used. Tier 1 includes in vitro assays. In this tier, bacterial mutation assays (such as the Ames test) are used first, followed by tests based on in vitro mammalian cells (detecting gene mutations or chromosomal aberrations). Tier 2 involves the use of short-term in vivo studies (usually a bone-marrow cytogenetics assay) to assess whether any potential for mutagenicity detected at the Tier 1 in vitro stage is actually expressed in the whole animal. Thus, negative results in vitro are usually considered sufficient to indicate lack of mutagenicity, whereas a positive result in vitro is not considered sufficient to indicate that the chemical represents a mutagenic hazard (i.e. it could be a false or misleading positive) (33–38). The evidence accumulated in more than 30 years of genotoxicity testing permits to draw conclusions on the scientific hypotheses underlying the pre-screening strategies for carcinogens (32). It appears that the Ames test is the best predictor of carcinogenicity for the DNA-reactive chemicals. For the chemicals that are negative in Salmonella, but positive in other in vitro assays (e.g. clastogenicity) no correlation with, and predictive ability for, carcinogenicity is apparent (the other in vitro assays generate an exaggerated rate of false-positive results) (24,39). The other working hypothesis is that in vitro positives should be confirmed through an in vivo genotoxicity assay; however, it is demonstrated that existing in vivo tests are insensitive and give a majority of false-negative results for many clearly genotoxic carcinogens (35,40). As a consequence, genotoxic chemicals may go undetected under the present pre-screening strategies. The other critical issue is that of the identification of epigenetic or nongenotoxic carcinogens. Nongenotoxic carcinogenicity is more elusive and less studied, and sometimes the risk to humans is not adequately appreciated (41–43). Traditionally, nongenotoxic carcinogens have been detected in rodent 2-year cancer bioassays (12,13), but new regulatory policies tend to drastically reduce the number of new cancer bioassays [see also e.g. (44)] and to rely on the pivotal role of genotoxic endpoints; however, nongenotoxic carcinogens are negative in all genotoxicity tests and thus go undetected. Justification for the development of alternative methods for the detection of nongenotoxic carcinogens includes their remarkable presence among the known human carcinogens [up to 25% in Class 1 human carcinogens as classified by the International Agency for Research on Cancer (IARC)] and the considerable potential risk associated with them (45). In addition, it should be emphasised that the proportion of nongenotoxic to genotoxic carcinogens in the environment is likely bound to increase in the near future, since the scientific knowledge on DNA reactivity [including structural alerts (SA)] allows the industrial chemists to design new compounds without overtly reactive moieties. The evident deficiencies of the pre-screening strategies adopted to avoid a massive use of the rodent bioassay have taken to the limelight again the cell transformation assays (CTAs), which have been proposed for assessing nongenotoxic carcinogens for many years but have undergone several cycles of favour and disfavour within the scientific community. CTAs 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 tumourigenic cells (46). 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 tumours in susceptible animals (47,48). Recently, the contribution of the CTAs to the pre-screening of carcinogenicity has been re-explored. The Organization for the Economic Cooperation and Development (OECD) has reconsidered the CTAs and published a detailed report (49,50), while the European Centre for the Validation of Alternative Methods (ECVAM) has performed a pre-validation of CTAs (51,52). In a previous article, we analysed the CTA data from the OECD compilation (53). The Syrian Hamster Embryo cells transformation assay (with pH = 7 protocol) performed best in predicting rodent carcinogenicity, and it was shown to be complementary to the Ames test in a tiered approach. This article expands considerably our previous analyses, by including (i) CTA results for inorganics (data from (49) and (ii) results from a further CTA [the Bhas 42 system (54)] not considered in the OECD report. In addition, within the framework of an integrated approach to identification of carcinogens, we have considered a new set of SA for nongenotoxic carcinogenicity. These results are discussed within the perspective of the risk assessment of potential carcinogens. Materials and methods Experimental systems and data This article focuses on cell models used for CTAs that have either primary or secondary normal cells [the Syrian Hamster Embryo (SHE) assay] or established cell lines (the BALB/c 3T3, the C3H10T1/2 and the Bhas 42 assays). The SHE model uses karyotypically normal cells and is intended to detect early stages of carcinogenesis. Its cell transformation results in phenotypic abnormalities in the colonies after a 7-day assay. For the SHE assay, two protocols are considered as separate systems, with pH = 7.0 or higher and with pH = 6.7. The other three assays use immortalised aneuploid murine cell lines and measure later stages of carcinogenesis. The endpoint is induction of morphologically aberrant foci several weeks after exposure to the test sample. The Bhas 42 cells were established from the BALB/c 3T3 cells through the transfection with a plasmid pBR322 containing Ha-MuSV-DNA, clone H1 (v-Ha-ras). Detailed description and data compilation for the CTAs are available in the OECD report (49), except for Bhas 42 that is presented in (54). Regarding the OECD data compilation, in this article we used the information in Tables 11–13 (which are the basis of the performance analyses presented in the OECD paper). Intra- or inter-laboratory contradictory outcomes were not considered. In our laboratory, the data were collected into a chemical relational data file (called ISSCTA) including (i) biological data and (ii) chemical structures in a computer-readable format. ISSCTA is freely available at the web-based ISSTOX cluster of toxicological databases: http://www.iss.it/ampp/dati/cont.php?id=233&lang=1&tipo=7. Another ISSTOX database (ISSCAN) was used as the source of carcinogenicity and Ames mutagenicity data. The chemical relational databases are described in (55). All ISSTOX databases are also implemented into the OECD (Q)SAR Toolbox, where they can be used as the basis for chemical risk assessment (29) (free download): http://www.oecd.org/document/54/0,3746, en_2649_34379_42923638_1_1_1_1,00.html. Structural alerts The SA are one of the most popular approaches to SAR (2). A compilation of SA was curated by John Ashby, based on the evidence on chemical carcinogenicity (56). Since most of the carcinogens known at the time were DNA-reactive (57,58), the Ashby’s compilation of SA applies to Salmonella mutagenicity as well (59). An updated version of the carcinogenicity/mutagenicity SA has been prepared in our laboratory and implemented as a module of the expert system Toxtree (free download: http://ecb.jrc.ec.europa.eu/qsar/qsar-tools/index.php?c=TOXTREE). This implementation was composed by and large by SAs for genotoxic carcinogenicity, with very few SAs for nongenotoxic carcinogenicity (2,60). This carcinogenicity/mutagenicity rulebase has been implemented into the OECD (Q)SAR Toolbox as well. The Toolbox is a freely available software aimed at the regulatory use of QSAR approaches, by using structure–activity methodologies for the formation of toxicologically meaningful categories allowing to fill data gaps by read-across or trend analysis: http://www.qsartoolbox.org/ (29). Recently, we have expanded the SAs rulebase with a large number of new SAs for nongenotoxic carcinogenicity. The expanded rulebase has been implemented into the OECD (Q)SAR Toolbox version 2.5 and is going to be implemented into the next version of Toxtree in the near future. The new SA and a detailed explanation of their mechanistic basis are available in the Toolbox. In this article, we used the expanded SA rulebase that includes a large representation of both genotoxic and nongenotoxic alerts for carcinogenicity. Data analysis methods This article presents two main types of data analyses. One is simply the calculation and visualisation of the agreement between two dichotomic classification systems, e.g. CTA results versus carcinogenicity results for the same set of chemicals. This can be conveniently described through a receiver operating characteristics (ROC) graph analysis. A ROC graph reports (i) true-positive rate (sensitivity) on the Y-axis and (ii) false-positive rate ( = 1 −specificity) on the X-axis. In a ROC graph, perfect performance is located at the left upper corner; the diagonal line represents random results (61,62). Another procedure is aimed at comparing globally various profiles of responses to the chemicals, e.g. sets of results obtained when the same group of chemicals is tested with different biological assays (rodent carcinogenicity, Ames test, CTAs, etc.). The biological responses are coded as 0 (negative) or 1 (positive). The distance or dissimilarity between each pair of tests is calculated. A convenient measure is the percentage of chemicals with different results, out of all chemicals in common. This generates a N × N data matrix that includes all test relationships. Then all dissimilarities between pairs of biological systems are reorganised through a principal component analysis of the N × N matrix; this provides a global view of systems inter-relationships (63). For some assays (e.g. Ames test, carcinogenicity), more than one compilation of results is available; in this article, the different compilations were used as independent assessments of the test performance (different databases may diverge in the chemicals included and in the criteria for defining the calls). It should be emphasised that the calculation of a distance matrix is considered to be a very robust approach to solving the problem of missing data, since it exploits all the available information regarding the relationship between two systems (64). Finally, cluster analysis—a method that identifies groups of similar objects—is applied to the principal components of the N × N distance matrix (63). Results The core of this article is an analysis on CTA performance that expands our previous one (53), by including (i) CTA results for inorganics [67 further chemicals from (49)] and (ii) results (n = 98) for the Bhas 42 CTA [not considered by the OECD report; data from (54)]. In a first analysis, the concordance between the results of the CTAs and of the rodent carcinogenicity bioassay for organic and inorganic chemicals of very different classes was studied. The CTAs are Syrian Hamster Embryo cells assay pH 7 (SHE_7), Syrian Hamster Embryo cells assay pH 6.7 (SHE_6.7), BALB/c 3T3 (Balb), C3H10T1/2 (C3H) and Bhas 42. The predictive abilities of the CTAs are displayed as a ROC graph in Figure 1. The comparison shows that SHE_7 performs best in predicting rodent carcinogenicity, with the highest sensitivity to carcinogens (sensitivity = true-positive rate = 0.92) and a fair specificity (false-positive rate = 1 – specificity = 0.33). SHE_6.7 has the lowest rate of false positive results (0.14), but the sensitivity (0.69) is significantly lower than that of SHE_7. For a comparison, the performance of the Ames test (STY) is displayed in Figure 1 as well. Since STY identifies only DNA-reactive/genotoxic carcinogens, it appears that its sensitivity is low in this database skewed toward nongenotoxic carcinogens (the rate of false negatives is low as well). Fig. 1. Cell transformation assays and rodent carcinogenicity. View largeDownload slide The figure displays (in the form of an ROC graph) the agreement of various CTAs with rodent carcinogenicity. For comparison, the graph also includes the agreement of the Ames test (STY) with rodent carcinogenicity (relatively to the set of chemicals for which CTA results are available). An ROC graph has 1 − specificity (false-positive rate) on the X-axis and sensitivity (true-positive rate) on the Y-axis (61,62). Fig. 1. Cell transformation assays and rodent carcinogenicity. View largeDownload slide The figure displays (in the form of an ROC graph) the agreement of various CTAs with rodent carcinogenicity. For comparison, the graph also includes the agreement of the Ames test (STY) with rodent carcinogenicity (relatively to the set of chemicals for which CTA results are available). An ROC graph has 1 − specificity (false-positive rate) on the X-axis and sensitivity (true-positive rate) on the Y-axis (61,62). Figure 2 shows the differential sensitivity of the CTAS with respect to the two categories of putative: (i) DNA-reactive and (ii) non-DNA-reactive/nongenotoxic carcinogens. The stratification of carcinogens was based on structural considerations, using the SA rulebase described in ‘Materials and methods’. It appears that SHE_7 is more sensitive than the other CTAs with both categories of carcinogens (Figure 2). It also appears that some CTAs (e.g. Bhas 42, Balb) perform better with genotoxic than with nongenotoxic carcinogens. As expected, the Ames test is sensitive almost exclusively to DNA-reactive carcinogens. Fig. 2. View largeDownload slide Differential sensitivity of cell transformation assays to genotoxic and nongenotoxic carcinogens. The carcinogens were stratified into putative DNA-reactive and non-DNA-reactive based on structural considerations (with the use of SA), and the sensitivity to the two subclasses was calculated. For comparison, the Ames test (STY) was included in the analysis. Fig. 2. View largeDownload slide Differential sensitivity of cell transformation assays to genotoxic and nongenotoxic carcinogens. The carcinogens were stratified into putative DNA-reactive and non-DNA-reactive based on structural considerations (with the use of SA), and the sensitivity to the two subclasses was calculated. For comparison, the Ames test (STY) was included in the analysis. Table I details the concordance between rodent carcinogenicity and SHE_7 for the various categories of bioassay results. It appears that the majority of noncarcinogens are negative in SHE_7, whereas an overwhelming majority of carcinogens are positive in SHE_7, both in the categories of DNA-reactive and nongenotoxic/non-DNA-reactive carcinogens. Table I. The Syrian Hamster Embryo cells transformation assay vs. rodent carcinogenicity Carcinogenicity  SHE pH 7  Negative  Positive  Noncarcinogens  36  18  Non-DNA-reactive carcinogens  6  70  DNA-reactive carcinogens  5  59  Carcinogenicity  SHE pH 7  Negative  Positive  Noncarcinogens  36  18  Non-DNA-reactive carcinogens  6  70  DNA-reactive carcinogens  5  59  The table compares the experimental responses of the rodent bioassay and of the Syrian Hamster Embryo cells CTA (pH = 7.0 or higher) to a set of 194 chemicals, including both carcinogens and noncarcinogens. The data were retrieved from an OECD compilation (49). The carcinogens are separated into putative DNA-reactive and non-DNA-reactive classes, based on structural considerations (see details in the text). View Large A further, interesting view on the relationships among the various CTAs and of the CTAs with the rodent bioassay is provided in Figure 3. Here the systems are globally compared with each other based on the entire profiles of responses to the chemicals: in the map, assay systems are close to each other if they have similar profiles of responses to the chemicals. Multivariate data analysis methods, which are particularly suited for the analysis of large and complex datasets, were applied to generate Figure 3. In practice, a matrix of distances between systems was built based on the profiles of response to the chemicals, and then principal component analysis was applied to the distance matrix (63). For rodent carcinogenicity, more than one compilation of results is available [i.e. OECD (49), ISSCAN database (55), and Japanese database used for the assessment of Bhas42 (54)]. Since different databases may diverge in the chemicals included and in the criteria for defining the calls, in the analysis these compilations were used as independent assessments of the rodent bioassay results. Fig. 3. View largeDownload slide Cell transformation assays and the rodent bioassay: a global comparison. The graph displays the global similarities of biological endpoints based on the response profiles to chemicals (see analytical procedure in the text). The final plot was obtained with principal component analysis, and it explains 80% of the entire variance of the data. For the rodent bioassay endpoint, three separate databases were used as independent assessments: Canc_OECD: bioassay results reported in the OECD compilation (49); Canc_iss4: bioassay results reported in the ISSCAN database (free download from the ISSTOX website http://www.iss.it/ampp/dati/cont.php?id=233&lang=1&tipo=7); Canc_jap: bioassay results reported in (54). Fig. 3. View largeDownload slide Cell transformation assays and the rodent bioassay: a global comparison. The graph displays the global similarities of biological endpoints based on the response profiles to chemicals (see analytical procedure in the text). The final plot was obtained with principal component analysis, and it explains 80% of the entire variance of the data. For the rodent bioassay endpoint, three separate databases were used as independent assessments: Canc_OECD: bioassay results reported in the OECD compilation (49); Canc_iss4: bioassay results reported in the ISSCAN database (free download from the ISSTOX website http://www.iss.it/ampp/dati/cont.php?id=233&lang=1&tipo=7); Canc_jap: bioassay results reported in (54). The most remarkable observation from Figure 3 is that the different carcinogenicity databases cluster together, with SHE_7 clustering with them as well. This is in agreement with the high predictability for carcinogenicity shown by this CTA. The other CTAs are more distant from the carcinogenicity results. Balb and C3H are close to each other, whereas SHE_67 and Bhas 42 seem to have more peculiar profiles of responses to the chemicals. A tiered approach to the identification of carcinogens In a previous article including only results for the organics present in the OECD compilation, a tiered testing strategy to the prediction of carcinogenicity has been presented (53). First the Ames test (or, alternatively, the SA) was used in Tier 1. Then, the chemicals negative in Tier 1 were ‘tested’ with SHE_7. For this article, the exercise is repeated with the inclusion of inorganics (Table II). The carcinogenicity and SHE_7 data were retrieved from (49), and the Ames test data were retrieved from the ISSCAN database. In one option (Tiered Approach A), the Ames test is used as the first screening tool. Out of the initial 122 chemicals (36 noncarcinogens and 86 carcinogens), 68 are negative in the Ames test and go to the next tier, where they are tested with SHE_7. The chemicals negative in SHE_7 are 25; this includes 17 noncarcinogens and 8 carcinogens. Thus Tiered Approach A reduces the unidentified carcinogens to 9% of the initial number (8/86). Table II. Alternative tiered approach to the identification of carcinogens Tiered approach A  Tiered approach B    Noncarcinogens  Carcinogens    Noncarcinogens  Carcinogens  Initial sample  36  86  Initial sample  52  130    After Tier 1    After Tier 1a  STY negative  27  41  SA genonegative  32  66          After Tier 1b        SA nongenonegative  27  43    After Tier 2    After Tier 2  SHE negative  17  8  SHE negative  17  5  % Initial sample  47%  9%    33%  4%  Tiered approach A  Tiered approach B    Noncarcinogens  Carcinogens    Noncarcinogens  Carcinogens  Initial sample  36  86  Initial sample  52  130    After Tier 1    After Tier 1a  STY negative  27  41  SA genonegative  32  66          After Tier 1b        SA nongenonegative  27  43    After Tier 2    After Tier 2  SHE negative  17  8  SHE negative  17  5  % Initial sample  47%  9%    33%  4%  Tiered strategies for assessing chemical carcinogenicity: the initial sample of chemicals is subjected to two strategies, one including the Ames test first and then SHE_7 (Approach A), and the second one including the SA and then SHE_7 (Approach B). The SA include both genotoxic and nongenotoxic alerts. After each step of the tier, the boxes report the chemicals that are negative in the step (see details in the text). Chemicals detected as positive are excluded from the continuation of the analysis. View Large In the second option (Tiered approach B), the Ames test is replaced by the fast and inexpensive SA. We have further distinguished between SA for DNA-reactive chemicals and SA for nongenotoxic carcinogens. The SA used here are those implemented in a new rulebase, which includes an expanded list of nongenotoxic SA (see ‘Materials and methods’). The initial sample of chemicals for this exercise is 52 noncarcinogens and 130 carcinogens (total number 182). It appears that 70 chemicals (out of the 182 in the initial sample) do not possess SAs, and so they go to SHE_7 testing. The application of the SAs reduces the number of carcinogens in the sample from 130 to 43. Out of the 70 chemicals without SAs, 22 are SHE_7 negative, with only 5 unidentified carcinogens: thus, a considerable reduction of the carcinogens (to a final 4% of the initial number, 5/130) is obtained. A closer inspection of Table II shows that Tier A is less sensitive to carcinogens and more specific (lower number of false-positive predictions), whereas Tier B is more sensitive but less specific. This has to do with the original difference between the Ames test and the SA: the Ames test is less sensitive but more specific than the SA [see Figure 2 in (39)]. The number of false positives is an issue that needs further work and refinement; however, it should be emphasised that the use of the traditional in vitro genotoxicity assays (e.g. mouse lymphoma mutation) instead of SHE in a tiered approach generates a remarkably higher number of false-positive results (53). Mechanisms of cell transformation: learning more from data The overall evidence points to the CTAs as reliable tools for the identification of both genotoxic and nongenotoxic carcinogens, so it is of interest to put the CTAs in a wider perspective in relation to other mechanisms of toxicity, namely carcinogenicity and genotoxicity. A very large number of experimental results—relative to many chemical classes—have been generated in the last 30–35 years of testing and compiled into databases. Since explicit knowledge on mechanisms/modes of action has serious limitations, one approach is to consider these panels of chemicals as probes for the different mechanisms/modes of action and then let relationships arise spontaneously from the data themselves. Technically, the same analytical procedure used for generating Figure 3 was applied: principal component analysis of systems distances provided an overall view of systems inter-relationships. For various assays (e.g. Ames test, carcinogenicity), more than one compilation of results is available; they were used as independent assessments of the test performance (different databases may diverge in the chemicals included and in the criteria for defining the calls). A difference with Figure 3 is that the number of systems and chemicals (n = 1150) is much larger, so also the resulting view is much more complex. A complementary cluster analysis was performed as well. Figure 4 displays a projection of the test relationships on the first two principal components (variance explained 57%). A third dimension explains another 17% of variance and is relative to the in vivo micronucleus test, which should be imagined up from the plane where the other systems are. Thus, the view from the first three principal components is a virtually exhaustive description of the relationships among the toxicological endpoint systems and how they respond to large panels of chemicals. Fig. 4. View largeDownload slide A data-based global view of toxicological endpoints. The response profiles of carcinogenicity, cell transformation and genotoxicity endpoints were analysed (see details in the text). The total number of chemicals considered is around 1150. The first three principal components explain 74% of total variance. The codes of the cell transformation assays are self-explanatory. The codes of the other toxicological systems are as follows: STY: Ames test; CANC: rodent bioassay; vitroMIC: in vitro micronucleus test; vivMIC: in vivo micronucleus test; cha: in vitro chromosomal aberrations; MLY: in vitro mouse lymphoma cells mutation test. The results of a number of endpoints are compiled in different databases. These are as follows: Lha: Kirkland–Lhasa compilation of genotoxicity results (38); Ls: Leadscope toxicity database (http://www.leadscope.com/toxicity_database/); _oecd: bioassay results reported in the OECD compilation (49); _jap: bioassay results reported in (54); _iss: results from the ISSCAN and ISSMIC databases in ISSTOX. Fig. 4. View largeDownload slide A data-based global view of toxicological endpoints. The response profiles of carcinogenicity, cell transformation and genotoxicity endpoints were analysed (see details in the text). The total number of chemicals considered is around 1150. The first three principal components explain 74% of total variance. The codes of the cell transformation assays are self-explanatory. The codes of the other toxicological systems are as follows: STY: Ames test; CANC: rodent bioassay; vitroMIC: in vitro micronucleus test; vivMIC: in vivo micronucleus test; cha: in vitro chromosomal aberrations; MLY: in vitro mouse lymphoma cells mutation test. The results of a number of endpoints are compiled in different databases. These are as follows: Lha: Kirkland–Lhasa compilation of genotoxicity results (38); Ls: Leadscope toxicity database (http://www.leadscope.com/toxicity_database/); _oecd: bioassay results reported in the OECD compilation (49); _jap: bioassay results reported in (54); _iss: results from the ISSCAN and ISSMIC databases in ISSTOX. Figure 4 points to four main attractors: (i) the carcinogenicity response profiles (four databases), with SHE_7 being very close to the responses of the rodent bioassay. SHE_6.7 and Bhas 42 give more dissimilar responses to the chemicals; (ii) the Ames test (three databases), which is known to be specifically sensitive to DNA-reactive chemicals; (iii) the in vitro mammalian mutagenicity assays (gene mutation and clastogenicity). These endpoints depend on interactions with both DNA and proteins. Unexpectedly, two CTAs (BALB/c 3T3, C3H10T1/2) have profiles of responses to the chemicals similar to those of the assays in this group and (iv) the in vivo mutagenicity (micronucleus) assay (two databases). As expected, the in vivo micronucleus is in the same region of the other chromosomal effects, but its response to the chemicals is further mediated by ADME properties that are not present in the in vitro systems. Figure 5 is a cluster analysis of the responses of the toxicological endpoints to the chemicals and contributes to better grasp the significance of the complex, three-dimensional situation displayed in Figure 4. Whereas principal component analysis is sensitive to the large-scale trends in the data, cluster analysis describes efficiently the data at the local scale (63,65). Figure 5 confirms very clearly the existence of the four poles described above. Fig. 5. View largeDownload slide A data-based global view of toxicological endpoints: cluster analysis. The clusters of similar response profiles of toxicological endpoints are displayed in the figure. The analysis complements the results in Figure 4 (see details in the text). The codes are as in Figure 4. Fig. 5. View largeDownload slide A data-based global view of toxicological endpoints: cluster analysis. The clusters of similar response profiles of toxicological endpoints are displayed in the figure. The analysis complements the results in Figure 4 (see details in the text). The codes are as in Figure 4. Overall, Figures 4 and 5 confirm that some CTAs (notably SHE_7) are reliable in vitro models for carcinogenesis, since they respond to chemicals in a way very similar to that of the rodent bioassay. The genotoxicity assays are more distant (more dissimilar) from carcinogenicity; this is in agreement with the fact that they ‘explain’ only one component in carcinogenesis. It should be noted that Salmonella and the rodent bioassay have the same values of Principal Component 2 in Figure 4: this indicates that Salmonella is a faithful model of the second (in order of importance) component of carcinogenesis, which has to be interpreted as the DNA-reactivity component. Discussion The tendency to reduce/replace the use of the rodent bioassay for identifying chemical carcinogens increases the reliance on alternative strategies that are largely based on the somatic mutation theory of cancer. However, evidence accumulated in more than 30 years of testing indicates that this strategy is quite inefficient and that even genotoxic carcinogens may go undetected. DNA-reactive carcinogens can be identified efficiently by the Salmonella typhimurium (Ames) test or the SA, but other in vitro and in vivo mutagenicity tests present in the usual testing strategies do not have added value and rather impair the prediction ability of Salmonella alone. In addition, this strategy does not address the issue of nongenotoxic carcinogens. However, there is evidence that the combination of Salmonella and in vitro cell transformation assays permits the identification of a very large proportion of carcinogens. The integration of the above in vitro assays with the mechanistic knowledge encoded into the SA makes even faster and more economical the entire process of carcinogenicity pre-screening with alternative methods (Table II). The following sections address more in detail some of the issues considered in this article. The high predictivity of CTAs: mechanistic implication/relevance Figures 1–5 indicate that the response of CTAs to potential carcinogens is very similar to that of the rodent bioassay, with a high predictivity for the carcinogenicity endpoint. This applies especially to SHE_7. In a pre-validation exercise by ECVAM, overall the CTAs have shown to be valuable to detect rodent carcinogens. It was also concluded that standardised protocols are now available that should be the basis for future use. In particular, the SHE pH 6.7 and the SHE pH 7.0 protocols and the assay systems themselves are transferable between laboratories and are reproducible within and between laboratories (51,66). Several suggestions have been made as to how CTAs may detect chemical carcinogens that operate by a range of mechanisms of action. It has been proposed that the assays may be detecting a basic carcinogenic change common to different modes of carcinogenesis. Two predominant forms of CTAs have emerged over the years: (i) assays using primary normal diploid cells of which the SHE assay is the most established example and (ii) assays employing immortalised aneuploid mouse cell lines (e.g., BALB/c 3T3, C3H10T1/2). The two forms of CTAs have been correlated to different stages of cell transformation (49). At least four stages seem to be involved in cell transformation. The stages are (i) a block in cellular differentiation (detected as morphological transformation in the SHE assay); (ii) acquisition of immortality expressed by unlimited lifespan and aneuploid karyotype and genetic instability; (iii) acquisition of tumourigenicity associated with foci formation and anchorage-independent growth obtained in the BALB/c 3T3, C3H10T1/2 and Bhas 42 assay systems and (iv) full malignancy when cells are injected in a suitable host animal (67). Within the theory on the four stages of cell transformation, SHE may be sensitive to a larger range of carcinogen types than other CTAs because it detects more basic and aspecific mechanisms. An additional proposal is that the success of the SHE assay may be due to the use of primary cells containing a wide variety of cell types susceptible to a range of different transformation pathways (68,69). In addition, SHE cells are metabolically competent. The mechanistic basis for the ability of SHE to detect epigenetic/nongenotoxic carcinogens is investigated in (70). The Ames and SHE tests in a larger perspective A subject of interest is why the Ames and SHE tests, which represent two success stories quite unique in the landscape of reductionist in vitro systems (71), are so useful in the identification of carcinogens. The Ames test is sensitive to a very large family of carcinogens that are able to interact with DNA according to various molecular mechanisms (e.g. direct or indirect alkylation, acylation, intercalation, formation of aminoaryl DNA adducts) (59) and provoke all types of mutations and genotoxic damage (including chromosomal aberrations); it is the typical example of the somatic mutation theory of cancer. In addition, chemicals able to react with DNA are usually also able to interact with proteins; thus they interfere with the cellular processes in many different ways. As shown in this article, DNA-reactive chemicals induce cell transformation as well. As a consequence, an Ames test positivity is not only a signature for the chemicals’ ability to induce the narrow category of gene mutations in bacteria, but it is a probe for the chemicals’ ability to induce a very wide range of disturbances to the physiology of cells. On its side, the cell transformation assay SHE_7 detects phenotypic alterations that are characteristic of tumourigenic cells. In vitro cell transformation can be produced by a plethora of different molecular mechanisms that do not include the induction of mutations (72); among others, these transformation assays are models of cellcell and cellstroma communication phenomena typical of cancer, which cannot be seen as a ‘single cell’ condition but is linked to modifications of the relations among cells in tissues (73–75). In other terms, cell transformation provides a model for the systemic effects advocated as one the major factors in carcinogenesis [see e.g. the tissue organisation field theory (31)]. It is worth comparing the above characteristics of Ames and SHE assays with other approaches aimed at improving the risk assessment procedures, which are being explored in present times. These approaches explore completely new tools, such as tracing molecular perturbations related to specific biochemical pathways with the use of various -omics technologies (in vitro high-throughput assay systems) (76). A related concept is the adverse outcome pathway. An adverse outcome pathway is a conceptual construct that portrays existing knowledge concerning the linkage between a direct molecular initiating event and an adverse outcome at a biological level of organisation relevant to risk assessment (77). This strategy assumes that macroscopic events, such as toxicity in tissues or whole animals, can be traced back to individual molecular events and that these molecular events are identical in in vitro high-throughput screening systems and in the whole organism. However, there are both theoretical reasons and practical evidence that this is hardly so. One reason is that a numberless range of pathways should be identified, without a clear clue on when to stop the search. At a more fundamental level, past lessons in modelling of complex systems, going from mechanical statistics to hydrodynamics and ecology, tell us that the emerging of an order or regularity—upon which we can base our predictions—happens at a statistical level, when huge numbers of elementary units are taken into consideration (78–80). Translated into a biological context, this suggests that the relevant level of investigation for deriving useful prediction on the ‘whole organism’ resides at the population level (e.g. in the cell–cell and cell–stroma interactions) and not inside the single cell. A direct experimental evidence comes from the failure of omics in pharmacology to predict correctly the efficacy and toxicity of new drugs (81). Another cogent, even though indirect, evidence comes from the results presented in this article, showing that the best in vitro predictors of carcinogenicity are the Salmonella assay and the SHE CTA, which are both quite aspecific and large-range tools and which do not model narrow, isolated pathways. In other words, they can be considered as higher-level ‘pathways’, each including a large number of pathways that—most importantly—interact with each other and so provide the essential systemic-component information. Do ‘nongenotoxic carcinogens’ lack genotoxic activity (and vice versa)? An issue for risk assessment The successful combination of two assays (Ames and SHE) that exemplify theories often presented as antagonistic may indicate that the distinction between genotoxic and nongenotoxic carcinogens is not so sharp and that the theories on the early stages of carcinogenesis are not mutually exclusive; on the contrary, different pathways in the carcinogenesis process may co-exist and should be taken into account in testing strategies. This may have implications for the risk assessment of carcinogens. Classification systems based on labelling chemicals as genotoxic or nongenotoxic and on presumed mechanisms of action for each class may lead to ambiguous reconstructions of the carcinogenic process. One motivation for such classification is that nongenotoxic carcinogens are thought to be less hazardous to human health than are genotoxic carcinogens. This view is based on the assumption that nongenotoxic carcinogens act as tumour promoters and exhibit threshold tumour dose-responses, whereas genotoxic carcinogens act as tumour initiators and exhibit proportional responses at low doses (43). However, this does not necessarily mean that a chemical affects the carcinogenic process solely through such activities. Many examples can be quoted. Cyproterone acetate and tamoxifen are liver carcinogens that had been thought to act by nongenotoxic mechanisms. Later studies indicated that they are genotoxic and have tumour-initiating activity (43). Steroidal estrogens are predominantly nongenotoxic, but they can also generate DNA adducts upon metabolism. The same is valid for inducers of oxidative stress, which produce DNA damage as well (42). On the other hand, DNA-reactive chemicals can also interact with proteins. In addition, animal studies demonstrate that tumour promoters can cause cancer in the absence of an initiating agent, and the existence or absence of threshold dose-responses cannot be determined from current knowledge of carcinogenic mechanisms (82). More important is the fact that several nonmutagenic carcinogens have been found to be carcinogenic in experimental animals as well as in humans (e.g. benzene, 2,3,7,8-tetrachlorodibenzo-p-dioxin, diethylstilbestrol, asbestos, arsenic) (43,45). Thus classification of carcinogens into genotoxic and nongenotoxic or into initiating or promoting agents may not only be unhelpful but even an impediment to risk assessment since it may lead to the use of unjustified default assumptions. A better option is to consider them as ‘predominantly’ genotoxic or nongenotoxic. Once we can accumulate enough relevant mechanistic information about individual chemicals and their contribution to the different stages of carcinogenesis, it will be more reasonable to use this information directly for risk estimation, without expending efforts in classifying them (43,83). In addition to hazard identification, the CTAs may contribute, in combination with other tools, to decide about the prevailing mode of action and then to risk assessment. Conclusions The tiered approach presented in this report permits the identification of chemicals of high concern: SHE_7 complements the Ames test (and the SA) in the identification of a very large number of carcinogens. Another strong point is the recent progress in the mechanistically based identification of nongenotoxic carcinogens through an expanded set of SA. This is a reliable ground for refinements toward full replacement of the in vivo bioassay. Among others, a weak point of the process is a limited specificity (incidence of false positives). Inspection of Table II can help in delineating the open issues. In both tiered approaches, the identification of DNA-reactive chemicals (through the Ames test or the genotoxic SA) leads to halving the number of unidentified carcinogens: half of the problem appears to be solved at this step. Another remarkable proportion of carcinogens is identified with the nongenotoxic SA (Tier B). Thus the chemical biological knowledge permits to break down and model a considerable part of the carcinogenicity mechanisms. The last steps in Table II indicate that SHE_7 can model—experimentally in vitro—most of what is left unexplained. At this point, an issue is whether it is possible to model theoretically also this part of the problem, and how. The question can be formulated as follows: is SHE_7 measuring a systemic component in carcinogenesis that cannot be broken down into narrow mechanisms and pathways? Or, is a more global approach and a different perspective necessary to model SHE_7? For example a systemic effect like anaesthesia induced by a wide range of chemicals has been modelled by Corwin Hansch through very few global physical chemical parameters and not through the description of many individual pathways (84); this is because the rate-limiting step of the entire process depends on the limited amount of physical chemical descriptors. In any case, the translation of SHE_7 into a theoretical model (like the translation of the Ames test into SA) would help much in reducing time and cost of assessments. In order to progress further along this path, what is needed is surely the availability of more experimental results for SHE_7 on which to base the modelling work. These results will be useful for a better assessment of its predictivity toward carcinogenicity as well. Acknowledgements This work was partially supported by Italy–USA cooperation project ‘Systems effects of chemicals: evaluation of HighThroughPut profiling’. References 1. Belpomme D. Irigaray P. Hardell L. Clapp R. Montagnier L. Epstein S. Sasco A. J . 2007 The multitude and diversity of environmental carcinogens. Environ. Res ., 105, 414–– 429. Google Scholar CrossRef Search ADS PubMed  2. Benigni R. Bossa C . 2011 Mechanisms of chemical carcinogenicity and mutagenicity: a review with implications for predictive toxicology. Chem. Rev ., 111, 2507– 2536. Google Scholar CrossRef Search ADS PubMed  3. Frankish H . 2003 15 million new cancer cases per year by 2020, says WHO. Lancet  361, 1278 Google Scholar CrossRef Search ADS PubMed  4. Tomatis L. Huff J. Hertz-Picciotto I. et al.   1997 Avoided and avoidable risks of cancer. Carcinogenesis  18, 97– 105. Google Scholar CrossRef Search ADS PubMed  5. Sokal R. R. Oden N. L. Rosenberg M. S. Thomson B. A . 2000 Cancer incidences in Europe related to mortalities, and ethnohistoric, genetic, and geographic distances. Proc. Natl Acad. Sci. USA  97, 6067– 6072. Google Scholar CrossRef Search ADS   6. Belpomme D. Irigaray P. Sasco A. J. Newby J. A. Howard V. Clapp R. Hardell L . 2007 The growing incidence of cancer: role of lifestyle and screening detection. Int. J. Oncol ., 30, 1037– 1049. Google Scholar PubMed  7. Liechtenstein P. Holm N. V. Verkasalo P. K. Iliadou A. Kaprio J. Koskenvuo M. Pukkala E. Skytthe A. Hemminki K . 2000 Environmental and heritable factors in the causation of cancer. N. Engl. J. Med  ., 343, 78– 84. Google Scholar CrossRef Search ADS PubMed  8. Benigni R . 2007 Social sexual inequality, and sex difference in cancer incidence. Environ. Res ., 104, 128– 134. Google Scholar CrossRef Search ADS PubMed  9. Haseman J. K. Melnick R. L. Tomatis L. Huff J . 2001 Carcinogenesis bioassays: study duration and biological relevance. Food Chem. Toxicol ., 39, 739– 744. Google Scholar CrossRef Search ADS PubMed  10. Tomatis L. Melnick R. L. Haseman J. K. Barrett J. C. Huff J . 2001 Alleged ‘misconceptions’ distort perceptions of environmental cancer risks. FASEB J ., 15, 195– 203. Google Scholar CrossRef Search ADS PubMed  11. Bucher J. R . 2000 Doses in rodent cancer studies: sorting fact from fiction. Drug Metab. Rev ., 32, 153– 163. Google Scholar CrossRef Search ADS PubMed  12. Huff J . 1999 Long-term chemical carcinogenesis bioassays predict human cancer hazards. Issues, controversies, and uncertainties. Ann. N. Y. Acad. Sci ., 895, 56– 79. Google Scholar CrossRef Search ADS PubMed  13. Huff J . 1999 Value, validity, and historical development of carcinogenesis studies for predicting and confirming carcinogenic risks to humans . In Kitchin K. T. . (ed.), Carcinogenicity. Testing, Predicting, and Interpreting Chemical Effects  Marcel Dekker New Yorkpp. 21– 123. 14. Fung V. A. Huff J. Weisburger E. K. Hoel D. G . 1993 Predictive strategies for selecting 379 NCI/NTP chemicals evaluated for carcinogenic potential: scientific and public health impact. Fund. Appl. Toxicol ., 20, 413– 436. Google Scholar CrossRef Search ADS   15. Huff J. E. Haseman J. K. Rall D. P . 1991 Scientific concepts, value, and significance of chemical carcinogenesis studies. Annu. Rev. Pharmacol. Toxicol ., 31, 621– 625. Google Scholar CrossRef Search ADS PubMed  16. Tomatis L . 2006 Identification of carcinogenic agents and primary prevention of cancer. Ann. N. Y. Acad. Sci ., 1076, 1– 14. Google Scholar CrossRef Search ADS PubMed  17. Tomatis L . 2005 Primary prevention of cancer in relation to science, sociocultural trends and economic pressures. Scand. J. Work Environ. Health  31, 165– 167. Google Scholar CrossRef Search ADS PubMed  18. Huff J . 2011 Primary prevention of cancer. Science  332, 916– 917. Google Scholar CrossRef Search ADS PubMed  19. European Commission ( 2003 Proposal concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH). COM(2003)644Final Google Scholar 20. Pedersen F. de Brujin J. Munn S. J. Van Leeuwen K. . ( 2003 Assessment of Additional Testing Needs under REACH. Effects of (Q)SARs, Risk Based Testing and Voluntary Industry Initiatives. JRC report EUR 20863 EN  European Commission, Joint Research Centre Ispra Italy 21. Van der Jagt K. Munn S. J. Torslov J. de Brujin J. . ( 2004 Alternative Approaches Can Reduce the Use of Test Animals under REACH. Addendum to the Report ‘Assessment of Additional Testing Needs under REACH. Effects of (Q)SARs, Risk Based Testing and Voluntary Industry Initiatives’. JRC Report EUR 21405 EN  European Commission, Joint Research Centre, Ispra Italy 22. Zeiger E . 2010 Historical perspective on the development of the genetic toxicity test battery in the United States. Environ. Mol. Mutagen ., 51, 781– 791. Google Scholar CrossRef Search ADS PubMed  23. Zeiger E . 2004 History and rationale of genetic toxicity testing: an impersonal, and sometimes personal, view. Environ. Mol. Mutagen ., 44, 363– 371. Google Scholar CrossRef Search ADS PubMed  24. Zeiger E . 1998 Identification of rodent carcinogens and noncarcinogens using genetic toxicity tests: premises, promises, and performance. Regulat. Pharmacol. Toxicol ., 28, 85– 95. Google Scholar CrossRef Search ADS   25. Hansch C . 1991 Structure-activity relationships of chemical mutagens and carcinogens. Sci. Total Environ ., 109–110, 17– 29. Google Scholar CrossRef Search ADS PubMed  26. Benigni R. Netzeva T. I. Benfenati E. et al.   2007 The expanding role of predictive toxicology: an update on the (Q)SAR models for mutagens and carcinogens. J. Environ. Sci. Health C Environ. Carcinog. Ecotoxicol. Rev ., 25, 53– 97. Google Scholar CrossRef Search ADS PubMed  27. Hansch C. Kim D. Leo A. J. Novellino E. Silipo C. Vittoria A . 1989 Toward a quantitative comparative toxicology of organic compounds. Crit. Rev. Toxicol. , 19, 185– 226. Google Scholar CrossRef Search ADS PubMed  28. Worth A. P. Bassan A. de Brujin J. Gallegos Saliner A. Netzeva T. I. Pavan M. Patlewicz G. Tsakovska I. Eisenreich S . 2007 The Role of the European Chemicals Bureau in promoting the regulatory use of (Q)SAR methods. SAR QSAR Environ. Res ., 18, 111– 125. Google Scholar CrossRef Search ADS PubMed  29. Van Leeuwen K. Schultz T. W. Henry T. Diderich B. Veith G. D . 2009 Using chemical categories to fill data gaps in hazard assessment. SAR QSAR Environ. Res ., 20, 207– 220. Google Scholar CrossRef Search ADS PubMed  30. Worth A. P . 2010 The role of QSAR methodology in the regulatory assessment of chemicals. In Puzyn T. Leszczynski J. Cronin M. T. D. . (eds.), Recent Advances in QSAR Studies: Methods and Applications  Springer Heidelbergpp. 367–3– 82. 31. Baker S. G. Cappuccio A. Potter J. D . 2010 Research on early-stage carcinogenesis: are we approaching paradigm instability? J. Clin. Oncol ., 28, 3215– 3218. Google Scholar CrossRef Search ADS PubMed  32. Benigni R . 2012 Alternatives to the carcinogenicity bioassay for toxicity prediction: are we there yet? Exp. Opin. Drug Metab. Toxicol ., 8, 1– 11. Google Scholar CrossRef Search ADS   33. Zeiger E . 1994 Strategies and philosophies of genotoxicity testing: what is the question? Mutat. Res ., 304, 309– 314. Google Scholar CrossRef Search ADS PubMed  34. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use ( ICH) ( 2008 Guidance on Genotoxicity Testing and Data Interpretation for Pharmaceuticals Intended for Human Use . S2(R1). ICH Geneva, Switzerland Google Scholar 35. Kirkland D. Speit G . 2008 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  ., 654, 114– 132. Google Scholar CrossRef Search ADS PubMed  36. Elespuru R. K. Agarwal R. Atrakchi A. H. et al.   2009 Current and future application of genetic toxicity assays: the role and value of in vitro mammalian assays. Toxicol. Sci  ., 109, 172– 179. Google Scholar CrossRef Search ADS PubMed  37. EFSA 2011 Scientific opinion on genotoxicity testing strategies applicable to food and feed safety assessment. EFSA  J ., 9, 2379 Google Scholar 38. Kirkland D. Aardema M. J. Henderson L. Muller L . 2005 Evaluation of the ability of a battery of three in vitro genotoxicity tests to discriminate rodent carcinogens and non-carcinogens I Sensitivity, specificity and relative predictivity. Mutat. Res ., 584, 1– 256. Google Scholar CrossRef Search ADS PubMed  39. Benigni R. Bossa C. Tcheremenskaia O. Giuliani A . 2010 Alternatives to the carcinogenicity bioassay: in silico methods, and the in vitro and in vivo mutagenicity assays. Exp. Opin. Drug Metab. Toxicol ., 6, 1– 11. Google Scholar CrossRef Search ADS   40. Benigni R. Bossa C. Tcheremenskaia O. Battistelli C. L. Crettaz P . 2012 The New ISSMIC Database on in vivo micronucleus, and its role in assessing genotoxicity testing strategies. Mutagenesis  27, 87– 92. Google Scholar CrossRef Search ADS PubMed  41. Lai D. Y. Woo Y. T . 2012 Reducing carcinogenicity and mutagenicity through mechanism-based molecular design of chemicals. In Boethling R. S. Voutchkova A. M. . (eds.), Green Processes. Volume 9: Designing Safer Chemicals  Wiley Heidelberg 42. Woo Y. T . 2003 Mechanisms of action of chemical carcinogens, and their role in structure-activity relationships (SAR) analysis and risk assessment. In Benigni R. . (ed.), Quantitative Structure-Activity Relationship (QSAR)Models of Mutagens and Carcinogens  CRC Press Boca RatonFL, pp. 41– 80. 43. Melnick R. L. Kohn M. C. Portier C. J . 1996 Implications for risk assessment of suggested nongenotoxic mechanisms of chemical carcinogenesis. Environ. Health Perspect ., 104, 123– 134. 44. Huff J . 2008 More toxin tests needed. Science  319, 725– 726. Google Scholar CrossRef Search ADS PubMed  45. Hernandez L. G. van Steeg H. Luijten M. van Benthem J . 2009 Mechanisms of non-genotoxic carcinogens and importance of a weight of evidence approach. Mutat. Res ., 682, 94– 109. Google Scholar CrossRef Search ADS PubMed  46. Barrett J. C. Ts’o P. O . 1978 Evidence for the progressive nature of neoplastic transformation in vitro . Proc. Natl Acad. Sci. USA  75, 3761– 3765. Google Scholar CrossRef Search ADS   47. Berwald Y. Sachs L . 1965 In vitro transformation of normal cells to tumour cells by carcinogenic hydrocarbons. J. Natl Cancer Inst ., 35, 641– 661. Google Scholar PubMed  48. Berwald Y. Sachs L . 1963 In vitro cell transformation with chemical carcinogens. Nature  200, 1182– 1184. Google Scholar CrossRef Search ADS PubMed  49. OECD ( 2007 Detailed Review Paper on Cell Transformation Assays for Detection of Chemical Carcinogens  . OECD Series on Testing and Assessment, No. 31. OECD, Paris Google Scholar 50. Vasseur P. Lasne C . 2012 OECD Detailed Review Paper (DRP) number 31 on ‘Cell Transformation Assays for Detection of Chemical Carcinogens’: main results and conclusions. Mutat. Res ., 744, 8– 11. Google Scholar CrossRef Search ADS PubMed  51. Corvi R. Aardema M. J. Gribaldo L. Hayashi M. Hoffmann S. Schechtman L. M. Vanparys P . 2011 ECVAM prevalidation study on in vitro cell transformation assays: general outline and conclusions of the study. Mutat. Res ., 744, 12– 19. Google Scholar CrossRef Search ADS PubMed  52. Vanparys P. Corvi R. Aardema M. J. Gribaldo L. Hayashi M. Hoffmann S. Schechtman L. M . 2012 Application of in vitro cell transformation assays in regulatory toxicology for pharmaceuticals, chemicals, food products and cosmetics. Mutat. Res ., 744, 111– 118. Google Scholar CrossRef Search ADS PubMed  53. Benigni R. Bossa C . 2011 Alternative strategies for carcinogenicity assessment: an efficient and simplified approach based on in vitro mutagenicity and cell transformation assays. Mutagenesis  26, 455– 460. Google Scholar CrossRef Search ADS PubMed  54. Sakai A. Sasaki K. Muramatsu D. et al.   2010 A Bhas 42 cell transformation assay on 98 chemicals: the characteristics and performance for the prediction of chemical carcinogenicity. Mutat. Res ., 702, 100– 122. Google Scholar CrossRef Search ADS PubMed  55. Benigni R. Bossa C. Richard A. M. Yang C . 2008 A novel approach: chemical relational databases, and the role of the ISSCAN database on assessing chemical carcinogenicity. Ann. Ist. Super. Sanità  44, 48– 56. Google Scholar PubMed  56. Ashby J . 1985 Fundamental structural alerts to potential carcinogenicity or noncarcinogenicity. Environ. Mutagen ., 7, 919– 921. Google Scholar CrossRef Search ADS PubMed  57. Miller E. C. Miller J. A . 1981 Mechanisms of chemical carcinogenesis. Cancer  47, 1055– 1064. Google Scholar CrossRef Search ADS PubMed  58. Miller J. A. Miller E. C . 1977 Ultimate chemical carcinogens as reactive mutagenic electrophiles. In Hiatt H. H. Watson J. D. Winsten J. A. . (eds.), Origins of Human Cancer  Cold Spring Harbor Laboratory Press Cold Spring Harborpp. 605–6– 28. 59. Benigni R. Bossa C . 2008 Structure alerts for carcinogenicity, and the Salmonella assay system: a novel insight through the chemical relational databases technology. Mutat. Res. Rev ., 659, 248– 261. Google Scholar CrossRef Search ADS   60. Benigni R. Bossa C. Jeliazkova N. G. Netzeva T. I. Worth A. P. . ( 2008 The Benigni/Bossa Rulebase for Mutagenicity and Carcinogenicity—A Module of Toxtree. EUR 23241 EN  Office for the Official Publications of the European Communities, EUR—Scientific and Technical Report Series, Luxembourg 61. Swets J. A . 1988 Measuring the accuracy of diagnostic systems. Science  240, 1285– 1293. Google Scholar CrossRef Search ADS PubMed  62. Kraemer H. C . 1988 Assessment of 2x2 associations: generalization of signal-detection methodology. Am. Stat. , 42, 37– 49. 63. Benigni R. Giuliani A . 1994 Quantitative modeling and biology: the multivariate approach. Am. J. Physiol ., 266, R1697– R1704. Google Scholar PubMed  64. Sneath P. H. A . 1983 Distortions of taxonomic structure from incomplete data on a restricted set of reference strains. J. Gen. Microbiol ., 129, 1045– 1073. 65. Lebart L. Morineau A. Warwick K. M . 1984 Multivariate Descriptive Statistical Analysis  Wiley New York 66. Maire M. A. Pant K. Phrakonkham P. et al.   2012 Recommended protocol for the Syrian hamster embryo (SHE) cell transformation assay. Mutat. Res ., 744, 76– 81. 67. Creton S. Aardema M. J. Carmichael P. et al.   2012 Cell transformation assays for prediction of carcinogenic potential: state of the science and future research needs. Mutagenesis  62, 93– 101. Google Scholar CrossRef Search ADS   68. Mauthe R. J. Gibson D. P. Bunch R. T. Custer L . 2001 The Syrian Hamster Embryo (SHE) cell transformation assay: review of the methods and results. Toxicol. Pathol ., 29, 138– 146. Google Scholar CrossRef Search ADS PubMed  69. Haga K. Ohno S. Yugawa T. Narisawa-Saito M. Fujita M. Sakamoto M. Galloway D. A. Kiyono T . 2007 Efficient immortalization of primary human cells by p.16INK4a-specific short hairpin RNA or Bmi-1, combined with introduction of hTERT. Cancer Sci ., 98, 147– 154. Google Scholar CrossRef Search ADS PubMed  70. Nguyen-Ba G. Vasseur P . 1999 Epigenetic events during the process of cell transformation induced by carcinogens (review). Oncol. Rep ., 6, 925– 932. Google Scholar PubMed  71. Benigni R. Bossa C. Giuliani A. Tcheremenskaia O . 2010 Exploring in vitro/in vivo correlations: lessons learned from analyzing phase I results of U.S. EPA’s ToxCast project. J. Environ. Sci. Health. C Environ. Carcinog. Ecotoxicol. Rev ., 28, 272– 286. Google Scholar CrossRef Search ADS PubMed  72. Bignami M. Ficorella C. Dogliotti E. Norman R. L. Kaighn M. E. Saffiotti U . 1984 Temporal dissociation in the exposure times required for maximal induction of cytotoxicity, mutation, and transformation by N-methyl-N’-nitro-N-nitrosoguanidine in the BALB/3T3 ClA31-1-1 cell line. Cancer Res ., 44, 2452– 2457. Google Scholar PubMed  73. Kalluri R. Zeisberg M . 2006 Fibroblasts in cancer. Nat. Rev. Cancer  6, 392– 401. Google Scholar CrossRef Search ADS PubMed  74. Soto A.M. Sonnenschein C . 2011 The tissue organization field theory of cancer: a testable replacement for the somatic mutation theory. Bioassays  33, 332– 340. Google Scholar CrossRef Search ADS   75. Potter J. D . 2007 Morphostats, morphogens, microarchitecture and malignancy. Nat. Rev. Cancer  7, 464– 474. Google Scholar CrossRef Search ADS PubMed  76. Kavlock R. J. Austin C. P. Tice R. R . 2009 Toxicity testing in the 21st century: implications for human health risk assessment. Risk Anal.   29, 485– 487. Google Scholar CrossRef Search ADS PubMed  77. Ankley G. T. Bennett R. S. Erickson R. J. et al.   2011 Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ. Toxicol. Chem ., 29, 730– 741. Google Scholar CrossRef Search ADS   78. Butcher E. C . 2005 Can cell systems biology rescue drug discovery? Nat. Rev. Drug Discov ., 4, 461– 467. Google Scholar CrossRef Search ADS PubMed  79. May R. M . 2001 Stability and Complexity in Model Ecosystems  Princeton University Press Princeton 80. Badii R. Politi A . 1997 Complexity  Hierarchical Structures and Scaling in Physics   Cambridge University Press Cambridge 81. MacDonald J. S. Robertson R. T . 2009 Toxicity testing in the 21st century: a view from the pharmaceutical industry. Toxicol. Sci ., 110, 40– 46. Google Scholar CrossRef Search ADS PubMed  82. Perera F. P . 1991 Perspectives on the risk assessment for non-genotoxic carcinogens and tumor promoters. Environ. Health Perspect ., 94, 231– 235. Google Scholar CrossRef Search ADS PubMed  83. Yamasaki H . 1995 Non-genotoxic mechanisms of carcinogenesis: studies of cell transformation and gap junctional intercellular communication. Toxicol. Lett. , 77, 55– 61. Google Scholar CrossRef Search ADS PubMed  84. Hansch C. Vittoria A. Silipo C. Jow P. Y . 1975 Partition coefficients and the structure-activity relationships of the anesthetic gases. J. Med. Chem ., 18, 546– 548. Google Scholar CrossRef Search ADS PubMed  © The Author 2012. Published by Oxford University Press on behalf of the UK Environmental Mutagen Society. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
TI - In vitro cell transformation assays for an integrated, alternative assessment of carcinogenicity: a data-based analysis
JF - Mutagenesis
DO - 10.1093/mutage/ges059
DA - 2012-11-06
UR - https://www.deepdyve.com/lp/oxford-university-press/in-vitro-cell-transformation-assays-for-an-integrated-alternative-2zcvwTdwBt
SP - 107
EP - 116
VL - 28
IS - 1
DP - DeepDyve
ER -