Mutagenesis Advance Access originally published online on September 28, 2007
Mutagenesis 2007 22(6):409-416; doi:10.1093/mutage/gem036
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
An analysis of results from 305 compounds tested with the yeast RAD54-GFP genotoxicity assay (GreenScreen GC)—including relative predictivity of regulatory tests and rodent carcinogenesis and performance with autofluorescent and coloured compounds
1Gentronix Ltd, CTF Building, 46 Grafton Street, Manchester M13 9NT, UK 2Safety and Environmental Assurance Centre, Unilever, Colworth Park, Sharnbrook, Bedford MK44 1LQ, UK 3Department of Genetic Toxicology, GlaxoSmithKline, Park Road, Ware, Hertfordshire SG12 0DP, UK 4Genetics Department, School of Medicine, University of Wales—Swansea, Singleton Park, Swansea SA2 8PP, UK 5Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge CB2 1QW, UK 6Centre for Sustainable Water Management, Lancaster Environment Centre, Lancaster LA1 4YQ, UK
Data from 305 non-proprietary compounds tested using the yeast RAD54-GFP (Green Fluorescent Protein) assay, GreenScreen GC, are presented, together with a detailed comparison with results from in vitro and in vivo genotoxicity tests and rodent carcinogenesis. In addition, observations on reproducibility and the performance of the test with autofluorescent and coloured compounds are described. Like the Ames test, the GreenScreen assay is shown to exhibit high specificity (82%), meaning that compounds with positive results are very likely to be genotoxic carcinogens. This is in contrast to mammalian cell tests established for use in regulatory testing that provide disappointingly low specificity and the inevitable generation of confounding false positive data. The analysis confirmed the observations of earlier studies, showing that a combination of an Ames test (or surrogate) with the yeast test provides high specificity as well as high sensitivity in the identification of rodent carcinogens.
 |
Introduction |
|---|
 Top Introduction Materials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
Preliminary hazard assessment for carcinogenicity relies on genotoxicity testing. The International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) battery of tests, which includes a bacterial reversion assay (Ames), an in vitro assessment of mutation using the mouse lymphoma assay (MLA) or chromosome aberrations in mammalian cells and an in vivo assessment of chromosome damage, is effective in identifying the majority of rodent carcinogens (1
). However, this high sensitivity comes at the cost of low specificity, and thus many non-carcinogens give positive results in in vitro mammalian cell tests. Recently, Kirkland et al. (2) reviewed the ability of the in vitro genotoxicity testing battery to predict rodent carcinogenicity. The majority of 183 non-carcinogens studied had positive data in at least one of the in vitro tests. Kirkland et al. introduced a new analytical tool, ‘relative predictivity’ (RP)—the ratio of real results to false results. Highlighting the problems with placing too much emphasis on concordance which inadequately reflects the false positive and false negative attributes of the tests, it was proposed that RP would be a ‘useful tool for industry to assess the likelihood of a chemical possessing carcinogenic potential from batteries of positive or negative results’. The ICH battery relies on genetic endpoints, i.e. outcomes of genetic damage. There is an increasing interest in the use of biomarkers in risk assessment, and for genetic toxicology probably the most promising source of these is the cellular response to DNA damage. Not all genotoxic stress results in gene mutations or chromosomal aberrations. Evolution has equipped cells with effective surveillance methods for DNA damage, which in turn can trigger appropriate responses such as growth arrest, DNA repair and in extreme circumstances apoptosis. Thus, for every genetic endpoint identified, there will have been many other events which are effectively repaired to ensure genetic integrity and are undetectable. Biomarker changes associated with the process of DNA damage repair will therefore occur in a greater proportion of the population, and to a greater degree, than discernable genetic damage or other genetic endpoints.
The response to DNA damage in Saccharomyces cerevisiae (yeast) is well characterized. Several laboratories have generated data using reporters of gene transcription or microarray experiments, demonstrating that agents capable of causing mutation in yeast will lead to induction of the RAD54 promoter (3–6). RAD54 is a member of the RAD52 epistasis group of DNA repair genes and is induced above a constitutive level by a variety of different DNA lesions, yet the promoter does not respond to non-genotoxic oxidative or reductive stresses, heat or osmotic shocks or amino acid starvation. RAD54 encodes a structural element of the homologous recombinational repair pathway and is transcriptionally up-regulated in response to exposure of the yeast to a broad spectrum of genotoxins and thus is a good surrogate for monitoring genetic endpoints. The stimulation of RAD54 transcription by DNA damage can be effectively monitored by an operatively linked promoter–GFP reporter fusion and this is exploited in the GreenScreen assay (GSA). Several studies have revealed the utility of this assay (7–9).
In this paper, we present all the data we have generated in-house to date from the GSA. In order to avoid authors' bias in the compilation of this compound collection, a number of industrial companies (pharmaceutical and consumer goods) were asked to suggest compounds that would be of interest to them in assessing this genotoxicity screening test. Thus, the collection is diverse in nature and contains many compounds for which there are conflicting data from other genotoxicity tests. An analysis is presented for the sensitivity and specificity of the test in comparison with a number of ICH battery tests and with rodent carcinogenicity. RP is included in this analysis. This collection includes many compounds that were assessed because their particular physical properties such as colour, autofluorescence, etc. and were of interest in defining performance limits for the test, since GSA uses an optical end point.
|
Materials and methods |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
The GreenScreen microplate assay protocol has been described fully elsewhere (7
). In brief, nine serial dilutions of each test chemical were made in a 96-well, black, clear-bottomed microplate. Four compounds were tested per microplate. GSA requires 375 µl of solubilized test compound in a typical diluent of 4% v/v dimethyl sulphoxide (DMSO) in sterile water. For a top test concentration of 10 mM, assuming an average molecular mass of 300 g/mol and taking account of the 50/50 dilution on the microplate when yeast is added, the compound requirement is <3 mg per test, making it suitable for relatively high-throughput screening. The diluted test chemicals were mixed separately with both a test yeast strain, GenT01 (expressing GFP from the RAD54 promoter), and a non-fluorescent control yeast strain, GenC01 (with an out-of-frame GFP gene), in a defined growth medium.
Following overnight incubation (16 - 20 hrs), cell density and fluorescence were measured. These raw data were transferred to the GreenScreen data handling software, which provides clear decision making for positive and negative results based on statistically relevant thresholds. For each microplate well analysed, the fluorescence data were divided by the absorbance data to give a normalized ‘brightness’ value. Brightness results for the control strain were subtracted from those of the test strain results to allow correction for moderate compound autofluorescence. A positive genotoxicity result was concluded if a sample produced an induction in brightness 
30% compared to an untreated control. The 30% significance threshold is conservative and based on three times the standard deviation in the brightness data from constitutive GFP expression when testing dilution series of both non-toxic and cytotoxic but non-genotoxic chemicals. A strong positive genotoxicity result was concluded if three or more serial dilutions of the compound produced an induction in brightness >30% or at least one dilution produced an induction >60%.
Cytotoxicity was assessed by measuring the relative total growth (RTG) in cell density compared to an untreated control (100%). A positive cytotoxicity result was concluded if RTG was reduced to <80% at one or more test concentrations. A strong positive cytotoxicity result was concluded if RTG was reduced to <80% over three or more serial dilutions of the compound or at least one dilution reduced RTG to <50%. It should be noted that this method of assessing cytotoxicity is not a measure of cell death or cell viability; it quantifies the extent of cell proliferation throughout the incubation period and assesses the reduction in RTG caused by cytotoxic chemicals.
Standard compounds were also run, each at two concentrations to provide dose-dependent cytotoxicity and genotoxicity controls. These chemicals were methanol and methyl methanesulphonate (MMS), respectively. These chemicals have the added advantage of being liquids, thus simplifying chemical handling in the assay protocol; however, MMS should be freshly prepared and used as it can degrade during prolonged storage.
In the initial screening validation protocol (7), 1% v/v DMSO in sterile distilled water was used as a solvent throughout. Some of the additional compounds listed in this study were tested in 2 or 0% DMSO. Where no DMSO was used, compounds were prepared in water alone, and this is noted in Appendix 1 (Supplementary data are available at Mutagenesis Online).
Since GreenScreen is based on a microplate protocol with sparing use of compound, the assay is easily adapted for high-throughput screening using conventional, laboratory and liquid handling automation. Approximately one-third of the compounds in this study were tested using microplates set-up by a liquid handling robot (Microlab S, Hamilton AG, Bonaduz, Switzerland) and the rest set up manually using a multichannel pipette. The results were wholly comparable irrespective of the instrumentation used.
Calculations
Commonly used terms and their definitions are taken from Cooper et al. (10) and are paraphrased here. A compound test, with for example GSA, can have a positive outcome for which there may be either a positive result (a) or a negative result (b), from a second, comparative test (for example MLA or rodent carcinogenesis). The total number of positives for GSA is thus (a + b). Similarly, the test with GSA can have a negative outcome for which there might be either a positive (c) or a negative (d), result from the second test. The total number of negatives for GSA is thus (c + d). It follows that the total number of positive results from the second test is (a + c) and the total number of negative results from the second test is (b + d). Similarly, the total number of compounds for which there are data for both tests, represented by N, is (a + b + c + d). The following terms were calculated from these basic figures.
- Sensitivity, the percentage of correctly identified positives = [a/(a + c)] x 100
- Specificity, the percentage of correctly identified negatives = [d/(b + d)] x 100
- Concordance (or accuracy), the percentage of correctly identified results, both positive and negative = [(a + d)/N] x 100
- Prevalence, percentage of positive cells in the second, comparative test results = [(a + c)/N] x 100
- RP of a positive result for carcinogenicity is the percentage of carcinogens giving a positive genotoxicity result divided by the percentage of non-carcinogens giving a positive genotoxicity result = [(a/(a + c))/(b/(b + d))]
- RP of a negative result for non-carcinogenicity is the percentage of non-carcinogens giving a negative genotoxicity result divided by the percentage of carcinogens giving a negative genotoxicity result = [(d/(b + d))/(c/(a + c))]
- Specificity, the percentage of correctly identified negatives = [d/(b + d)] x 100
The higher the RP value, the more predictive the testing strategy.
Note that where comparisons with carcinogenicity data are made in this paper these refer only to rodent carcinogenicity.
|
Comparative results |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
Appendix 1 (Supplementary data are available at Mutagenesis Online) contains a complete list of all 305 compounds tested in-house with GSA including 102 which have been previously published (7
). Appendix 1 also includes data, where available, from the battery of regulatory genotoxicity tests and rodent cancer studies. The data have been collected both from the peer-reviewed literature and from freely available Internet resources. These include Chemical Carcinogenesis Research Information System (http://toxnet.nlm.nih.gov), National Toxicology Program Reports (http://ntp-server.niehs.nih.gov), International Agency for Research on Cancer (IARC, http://www.iarc.fr), International Programme on Chemical Safety INCHEM (http://www.inchem.org), National Institute for Occupational Safety and Health (http://www.cdc.gov/niosh/homepage.html), Physicians' Desk Reference (2007) edition 61, published by Thompson PDR, Montvale NJ, USA and the US Environmental Protection Agency (http://www.epa.gov/iris/search.htm). There are several compounds where there are contradictory in vitro and in vivo genotoxicity data from different sources, and these are recorded as +/–. However, where comparisons have been drawn these are recorded as positive to reflect the potential for a positive result in a regulatory test. For the purposes of the present study, a compound was considered to be a rodent carcinogen if at least one gender of one species (mouse or rat) had significant increased tumour incidence at least at one site. This is the same approach adopted in recent studies to evaluate the effectiveness of genotoxicity tests to discriminate rodent carcinogens from non-carcinogens (2) or characterize the genotoxicity of compound collections (11). It is not intended in this paper to reclassify chemicals on the basis of other test results and the authors apologize for any inadvertent gaps or confounding decisions. |
Technical issues of compound testing |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
Requirement for a control strain and autofluorescence correction
The GSA makes use of two yeast strains, primarily to combat the effects of any autofluorescence of the test compound that might otherwise be misinterpreted as an induction in GFP expression, and hence lead to a false positive genotoxicity result. The use of the non-fluorescent control strain (GenC01) alongside the test strain (GenT01) allows effective comparative correction for autofluorescence, since the test compound is exposed to the same environment and any degradation due to cellular metabolism when combined with either strain. In the assessment of genotoxicity, the relative fluorescence of the control strain was subtracted from that of the test strain. This approach worked effectively for the vast majority of compounds tested.
An exercise was carried out to test the requirement for the control yeast strain. Data from a subset of 102 compounds, as listed in the study by Cahill et al. (7), were reprocessed without correcting for autofluorescence in the control strain. If the control strain could be omitted, this would effectively double the capacity for compound testing on each microplate and may lead to an increased sensitivity in the measurement of GFP induction.
When the control strain data were not used, 21% of the ‘positive’ results were reclassified as ‘strong positive’, however, 10% of non-genotoxic compounds were sufficiently fluorescent to give a false positive result, although this result is highly dependent on the chemical space analysed. Overall there was an 8% change in qualitative results from the whole dataset. Of those compounds that gave a positive result with or without the control strain (64 compounds), 73% gave the same lowest effective concentration (LEC) results (i.e. the same sensitivity), 19% were apparently one serial dilution more sensitive and 8% more than one serial dilution more sensitive. Hence, correction for autofluorescence is necessary to give credible results with high specificity.
In a few cases (eight in this study, 2.6%), the autofluorescence of the test compound was sufficiently high at the same wavelength as GFP as to significantly obscure the GFP fluorescence. Hence, across a dilution series of the compound, the control strain appears almost as bright as the test strain, preventing accurate correction. These compounds were principally conjugated polyheterocyclic compounds or fluorescein analogues. For these compounds, a modified protocol employing fluorescence polarization (FP) was used, which has been described elsewhere (12). The protocol exploited the high fluorescence anisotropy of GFP compared to the generally smaller, more mobile molecules of the test compound. When illuminated with polarized light, the emitted fluorescence from GFP is still highly polarized with respect to the excitation light. Hence, the difference between the intensity of the fluorescence polarized parallel and perpendicular to the excitation light is large for GFP, yet much smaller for other autofluorescent species. Taking this calculated difference as the analytical signal allows the fluorescence from GFP to be distinguished from that of the test compound.
This approach worked well for these highly autofluorescent compounds and in addition even removed the background, low level, autofluorescence of the control strain and growth medium. However, if the test compound also shows high fluorescence anisotropy, the effectiveness of this approach will be reduced. This may be the case if the test compound is very large or is a small molecule which binds strongly to larger macromolecules within the yeast cell or the microplate surface itself, such that molecular rotation is restricted. Although this rarely occurs, in this collection, the effect was observed with the highly fluorescent dye and biological stain erythrosin B, an iodine-substituted derivative of fluorescein and the only example in this study where the FP method did not successfully correct for compound autofluorescence. Unless a compound is also highly coloured, the autofluorescence does not affect the measurement of cytotoxicity.
Coloured compounds
Compounds which are highly coloured present potential interference in any optical assay. In GSA, coloured substances increase the measured absorbance, artificially increasing the estimation of cell proliferation and hence diminishing both the cytotoxicity and genotoxicity response. Coloured substances also potentially absorb both the excitation and emission light of the GFP fluorescence, which may further diminish the GFP induction signal. However, for the majority of compounds in this study, including many coloured dyes, the intensity of colour was not significant at diluted concentrations where the compound was still toxicologically active.
Four coloured compounds (two food dyes and two iron chlorides) significantly interfered with the absorbance measurements in this study such that a modified protocol was required. In the revised protocol, a separate yet identical microplate was created with the diluted test compound combined with growth medium alone (no yeast cells). Absorbance measurements from this plate could then be subtracted from those of the test plate. The method was effective, although there is a risk that the light-absorbing properties of the compound may change in the environment created by the growing yeast culture, which is not reflected on the control plate.
It is possible to estimate the limits of this approach. The method should work up to the limit of Beer's Law, where the absorbance no longer increases linearly with compound concentration. Conservatively, approximating this limit to be 0.5 OD600 for a 1-cm path length, since the path length of the liquid in the microplate well is much shorter (
4 mm), and taking account of the fact that the test compound is diluted by 50% when tested due to the addition of the yeast cell culture, the limit of the comparative correction method equates to a test compound exhibiting an absorbance OD600 = 2.5, at the standard 1-cm path length.
Blue and green compounds strongly absorb the red light (600–620 nm) commonly used to quantify cell density. However, since the yeast cells are of a sufficiently large size to absorb light to approximately the same degree irrespective of wavelength across the visible spectrum, the GSA works just as well using blue light for absorbance measurements. Hence, when analysing blue/green compounds such as mitomycin C, more accurate cell density measurements were made at 492 nm (aqua green) where potential interference was markedly reduced. Red compounds absorb blue/green light and hence have the potential to interfere primarily with the fluorescence measurement and the genotoxicity evaluation, even at low test concentrations. For fluorescence measurements switching wavelengths is not an option and red-coloured compounds are likely to be the most problematic for this assay. This, in part, is one of the complications of implementing a protocol with the S9 liver homogenate.
|
Protocol development with S9 for promutagens |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
Significant effort was spent attempting to develop a protocol to enhance the metabolic competency of the yeast by the addition of rat liver S9 (Moltox, Boone, NC, USA), with the aim of detecting genotoxicity from a greater number of promutagenic compounds. S9 fraction as a material presents several challenges to any optical assay and these were significant in the GreenScreen protocol. First, the red colour and particulate nature of the material interfered with absorbance measurements. S9 fraction has a broad absorbance spectrum with a peak at 415 nm. Absorbance at 600 nm was significant in the GSA down to concentrations of 0.38% v/v for S9 fraction and 0.75% v/v for the supernatant obtained after further centrifugation of S9, since the red colour was retained. Positive interference reduced the apparent sensitivity of the assay in the measurement of cell density and hence cytotoxicity. Second, S9 interfered on two levels with the measurement of fluorescence. S9 itself was autofluorescent with a broad spectrum from 500 to 600 nm (excitation 485 nm), and while the use of FP as previously described did not significantly diminish its autofluorescence, the use of the control strain for autofluorescence correction was effective. Of greater significance, however, was that its red colour absorbed strongly at both the GFP fluorescence excitation and emission wavelengths diminishing the sensitivity for the measurement of GFP induction and hence genotoxicity. This fluorescence interference was significant down to 0.20% v/v for S9 fraction and 0.38% v/v for the centrifuged supernatant.
Various protocols were examined based on approaches used in Vitotox and Ames assays as well as adapted methods devised in-house. Variables examined included the use of mid-log versus stationary phase cells, incubation time, temperature, concentration, pH and buffer conditions for the assay and S9 treatment stages and the use of dialysis membranes (data not presented). Interactions between these variables in addition to factors, such as the variation of S9 activity between batches, instability of S9 solutions and low solubility of test compounds, compounded by the issues of interference with optical measurements led to inconsistent results and meant that development of a standardized and reproducible S9 protocol for GreenScreen was not possible.
However, yeast has a spectrum of metabolizing enzymes, albeit less comprehensive than mammalian cells, that includes a number of CYP (cytochrome P450) type enzymes and glutathione S-transferases (13–17). This proficiency of the yeast is reflected in the fact that of 28 compounds in Appendix 1 (Supplementary data are available at Mutagenesis Online) which require external metabolic activation, usually by S9, to give a positive result in the Ames test, eight (28.6%) gave positive results in the GSA.
Reducing the number of dilution steps
GSA uses nine serial dilutions of each compound. An analysis was made to determine over how many serial dilutions from the starting concentration, test compounds were still toxicologically active, i.e. gave an induction of GFP expression, or reduction in relative cell density, beyond the significance thresholds. In all, 166 compounds gave positive cytotoxicity data while 94 compounds gave positive genotoxicity data. After three serial dilutions, 41.6 and 35.1% of compounds were still cytotoxic and genotoxic, respectively. In all, 19.9 and 14.9% of compounds were still cytotoxic and genotoxic, respectively, after five serial dilutions.
Reducing the number of serial dilution steps would enables a higher density of compounds to be tested per microplate, increasing throughput. While these results indicated that the majority of compounds are toxicologically active in yeast over less than five serial dilutions of the top standard, testing over nine serial dilutions gives a greater chance of capturing this toxic range in the first test of a compound, often removing the need for an additional range-finding test. However, for higher throughput screening from a fixed concentration, a shorter dilution series would be appropriate.
A reproducibility study
In a separate study, a subset of 66 compounds representing a diverse range of substances, both non-toxic and genotoxic by a variety of mechanisms, were tested a minimum of four times by an independent laboratory. Permethrin was an exception and was tested only three times. These compounds are marked with an ‘R’ in Appendix 1 (Supplementary data are available at Mutagenesis Online). Compounds were tested in 2% v/v DMSO to 10 mM or 5 mg/ml, whichever is the lower, unless restricted by solubility. Sixty-three from 66 compounds (95.5%) gave the same result for genotoxicity (positive or negative) every time they were tested. Methotrexate was tested five times and was positive on four of five occasions. Aphidicolin and tritolyl phosphate were positive for genotoxicity on two of five and one of four occasions, respectively, and were concluded as ‘equivocal’ positives for genotoxicity. In the case of aphidicolin, results that were reported by GSA software as negative did show a dose-dependent induction of GFP, but below the significance threshold of 1.3. The results indicated a high degree of reproducibility in the determination of genotoxicity using GSA, providing the incubation time, microplate reader and solvent conditions are standardized.
|
Discussion of results |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
Seventy-three of the 305 compounds in the set (23.9%) were not fully soluble at the concentration tested in up to 4% DMSO v/v. While many of these compounds gave positive cytotoxicity and genotoxicity results, the LEC quoted may be an overestimate of the solubilized fraction of the compound that is bioavailable to the yeast. Compounds were generally tested to the limit of solubility or cytotoxicity, whichever was the lower. Hence, 57 of the compounds (18.7%) were tested at concentrations >10 mM, the ICH guideline limit for testing compounds. However, half of these compounds were liquids, including eight common solvents. Only 13 of these compounds (4.3%) were positive for genotoxicity with an LEC >10 mM and seven compounds (2.3%) with an LEC >20 mM. Excluding these 13 compounds from the following analysis made only small differences to the overall statistical results and hence they were included for completeness.
Table I (A1 and B) lists the comparative data for each single test and in paired combinations for the prediction of rodent carcinogenesis. In contrast, Table II characterizes the prediction of both rodent carcinogenesis and other regulatory genotoxicity data by GSA. It is perhaps unsurprising that Ames data is the most readily available, being a common first screen for genotoxicity (data available for 271 compounds). Compounds with positive Ames data will often not be tested further leading to fewer test results from other genotoxicity assays. For example, data from the in vitro micronucleus test (MNT) were the least represented and only found for 109 compounds (35.7% of the dataset). Positive prevalence was the highest for in vitro MNT (71.9%) but similar for the other tests at between 58.0 and 64.6%. A complete dataset for all four regulatory assays, GSA and rodent carcinogenicity was only available for 48 compounds (15.7%). In each statistical measure made, the number of compounds from which data are available is stated in Tables I and II. Note that some comparisons are approximate because they will be based on overlapping but non-identical sets of compounds.
|
|
Figure 1 graphically ranks the sensitivity and specificity of the individual tests in predicting rodent carcinogenesis from figures included in Table I (A1 and B). In the regulatory battery of tests, the cost of high sensitivity is low specificity and this is reflected in the data presented here. Figure 1 shows that an increased ability to detect rodent carcinogens (sensitivity) is accompanied by a poorer ability to correctly identify non-carcinogens (specificity). In this dataset, Ames and GSA had the highest, almost equivalent specificity (84.4 and 82.3%, respectively), which was in turn higher than all other tests or paired combinations of tests. While GSA alone exhibited the lowest sensitivity (38.8%), when examined in combination with Ames, the sensitivity was increased significantly to 66.4%, while the specificity (70.1%) was the highest of any paired combination of tests. MNT in vitro testing shows the highest sensitivity both alone (70.3%) and in combination with other tests; however, this is due in part to the higher positive prevalence in this dataset, i.e. a greater percentage of carcinogens versus non-carcinogens in the compounds for which MNT in vitro test results are available.
|
In Appendix 1 (Supplementary data are available at Mutagenesis Online), 50 compounds (16.4%) were listed as +/– for rodent carcinogenicity and the comparisons described thus far are based on classifying these as positive, i.e. true carcinogens. The use and prevalence of the term +/– in Appendix 1 with respect to the rodent carcinogenicity data reflects the frequent lack of correlation of carcinogenicity results between reported assays, species and sex due to differing metabolism, excretion, etc. This invariably leads to the weight of evidence approach in interpreting data, as adopted by the IARC for example. Hence, in this study, the prediction of rodent carcinogenicity by all the genotoxicity assays considered was also separately reassessed, first treating the +/– carcinogenicity results as negative and second with the +/– results excluded from the calculations.
The effect of classifying the +/– rodent carcinogenicity results as negative is shown in Table I A2. The key comparative results varied slightly, increasing the sensitivity of GreenScreen from 38.8 to 43.9% and of GreenScreen in combination with Ames from 66.4 to 74.2%, yet this largely reflects the decrease in positive prevalence in the dataset. The specificity for GreenScreen was reduced from 82.3 to 76.7%, however, this figure was higher than for any other test or paired combination of tests. The effect of removing the 50 +/– rodent carcinogenicity results from the analysis is shown in Table I A3. Note that in this instance, the number of comparisons was significantly reduced in each case. Specificity was unchanged from the initial analysis since this figure is based on negative carcinogenicity data, yet sensitivity was increased to 43.9% as in the previous analysis.
The compound collection is diverse and contains a significant proportion of chemicals considered or suspected to be non-genotoxic carcinogens. These amount to 35% (41 from 116) of carcinogens listed and are labelled as such in Appendix 1 (Supplementary data are available at Mutagenesis Online). Reference sources used by the authors in categorizing each chemical as a genotoxic or non-genotoxic carcinogen were those previously quoted for determining results for the battery of regulatory genotoxicity tests listed earlier in the Comparative Results section. Carcinogens which predominately act via non-genotoxic mechanisms may not necessarily be expected to be detected by in vitro cellular genotoxicity assays such as GSA. However, 46.3% of these compounds (19 from 41) from this dataset have some positive data in the regulatory battery of tests examined here and 26.8% (11 from 41) are positive in GSA, demonstrating how individual compounds can act by a number of different genotoxic and carcinogenic mechanisms.
If these 41 non-genotoxic carcinogens are removed from the dataset, the sensitivity of GSA increases to 45.3% and further to 85.3% when GSA is taken in combination with Ames. RP for a positive result was >2, the significance threshold suggested by Kirkland et al. (2), for GSA alone (2.19) and in combination with Ames (2.22). RP increases to 2.56 for GSA and 2.86 for GSA in combination with Ames if non-genotoxic carcinogens are removed from the dataset. In addition, the RP for a negative result was >2 for the combination of GSA and Ames (2.09). These figures demonstrate the power of combining an Ames test with the eukaryotic GSA, both of which can be implemented for high-throughput screening if a screening version of the Ames test such as AmesII is used (18). With high specificity, a positive result in GSA should be considered a reliable indicator of a positive result in a rodent carcinogenicity study and a valuable tool in compound ranking.
Table II shows the correlation of GSA with carcinogenicity and the other genotoxicity assays considered in this exercise and its ability to predict positive and negative outcomes in these assays. GSA had both high specificity and the highest RP for positive results when predicting rodent carcinogenesis compared to GSA's prediction of the results of other tests. Thus, the GSA gives a very low rate of false positive results, essential if potentially useful drug candidates are not to be discarded as a result of a genotoxicity screen and particularly relevant to an early ‘filtering’ screen. These results were in contrast to the lower RP for positive results when GSA is compared to Ames, reflecting the eukaryotic nature of GSA. Hence, GSA gives complementary data to Ames, indicating that these two assays will be effective when used together in a screening battery of tests. This was recently demonstrated in a study by Van Gompel et al. (8), whereby the analysis of 2351 proprietary drug candidates showed that 7% were positive in AmesII and 7.5% were positive in GSA. However, only 7% of the GSA positives were also positive in AmesII, showing that each assay detects a different but overlapping spectrum of genotoxins.
The high specificity of GSA is coupled with relatively low sensitivity, which is lower than Ames and other mammalian cell tests for this compound set. However, as previously mentioned, GSA positives are not simply a subset of the Ames positives. The Ames test (performed with and without S9 activation) identifies some compounds which are missed by GSA (performed without S9) largely as a consequence of its incomplete metabolic competency. However, GSA identifies some compounds missed by the Ames test, largely as a consequence of the lack of eukaryotic targets in bacterial cells (chromatin and DNA-metabolizing enzymes acting on chromatin, etc.). It is of course because of the prokaryote/eukaryote difference that there is an ICH battery of tests and not just the bacterial Ames test. It is therefore informative to look at combinations of assays with GSA and in this context, Figure 1 reveals that the combination of two high-specificity tests (Ames and GSA) maintains specificity while increasing sensitivity, in contrast to combinations including lower specificity mammalian cell tests.
|
Conclusions |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
The current ICH battery of genotoxicity tests exhibits high sensitivity such that few genotoxic carcinogens are missed. Unfortunately, their collective lack of specificity means that professional genetic toxicologists are forced to spend their time deciding how to interpret or further investigate positive results for which there is a high chance of irrelevance to human risk assessment. In this study, GSA has been shown to exhibit high specificity and a very low rate of false positive results, while combining results with a bacterial Ames test provides improved sensitivity for the detection of genotoxic carcinogens. Thus, when incorporated in a small screening battery, including the Ames test with and without S9 metabolic activation, GSA provides a mechanism to help reduce the number of hazardous compounds needlessly proceeding to later stage regulatory genotoxicity assays and animal tests, without producing additional false positive cells which can lead to the rejection of valuable compounds or necessity for extra investigative testing. At the very least, GSA can be used to allow effective prioritization of compounds in the drug development process, and the high-throughout nature of the assay coupled with the sparing use of valuable drug compound means that GSA can be applied early in the process.
This data review has focussed on the yeast genotoxicity assay, but some more general conclusions can be drawn. The ‘relative predictivity’ metric does indeed appear to have utility in the assessment of genotoxicity testing strategies. GFP reporter systems are sensitive to coloured and fluorescent compounds, but in general, by use of a non-fluorescent control strain the intensity of optical interference is not significant at the concentrations of test compounds currently required by the regulators. Recently, a genotoxicity reporter system exploiting the expression of the GADD45a gene in the human lymphoblastoid cell line TK6 has been described (19). It too uses GFP as the reporter fluorophore and demonstrates greater sensitivity than the yeast test but uniquely not at the cost of reduced specificity. The analysis of data from the yeast test provides a sound basis for the definition of similar operating limits for the human cell test.
|
Supplementary data |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
Supplementary data are available at Mutagenesis Online.
|
Acknowledgments |
|---|
Conflict of interest statement: R.M.W. is the Founder and Chief Scientific Officer, D.J.T. is a Consultant and A.W.K., N.B., P.A.C. and P.O.K. are or have previously been employed by Gentronix Ltd. The remaining authors have no conflicts to declare.
|
Notes |
|---|
* To whom correspondence should be addressed. Tel: +44 (0) 161 606 7266; Fax: +44 (0) 161 606 7337; Email: andrew.knight{at}gentronix.co.uk
|
References |
|---|
|
TopIntroductionMaterials and methodsComparative resultsTechnical issues of compound...Protocol development with S9...Discussion of resultsConclusionsSupplementary dataReferences |
|---|
-
1. Müller L, Kikuchi Y, Probst G, Schechtman L, Shimada H, Sofuni T, Tweats D. ICH-Harmonised guidances on genotoxicity testing of pharmaceuticals: evolution, reasoning and impact. Mutat. Res. (1999) 436:195–225.[CrossRef][Web of Science][Medline]
2. Kirkland D, Aardema M, Henderson L, Muller L. 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. (2005) 584:1–256.[Web of Science][Medline]
3. Averbeck D, Averbeck S. Induction of the genes RAD54 and RNR2 by various DNA damaging agents in Saccharomyces cerevisiae. Mutat. Res. (1994) 315:123–138.[Web of Science][Medline]
4. Cole GM, Schild D, Lovett ST, Mortimer RK. Regulation of RAD54- and RAD52-lacZ gene fusions in Saccharomyces cerevisiae in response to DNA damage. Mol. Cell. Biol. (1987) 7:1078–1084.
5. Paques F, Haber JE. Multiple pathways of recombination induced by double-strand breaks in Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. (1999) 63:349–404.
6. Tan TLR, Kanaar R, Wyman C. Rad54, a Jack of all trades in homologous recombination. DNA Repair (Amst.) (2003) 2:787–794.[CrossRef][Medline]
7. Cahill PA, Knight AW, Billinton N, Barker MG, Walsh L, Keenan PO, Williams CV, Tweats DJ, Walmsley RM. The GreenScreen genotoxicity assay: a screening validation programme. Mutagenesis (2004) 19:105–119.
8. Van Gompel J, Woestenborghs F, Beerens D, Mackie C, Cahill PA, Knight AW, Billinton N, Tweats DJ, Walmsley RM. An assessment of the utility of the yeast GreenScreen assay in pharmaceutical screening. Mutagenesis (2005) 20:449–454.
9. Bartos T, Letzsch S, Skarek M, Flegrova Z, Cupr P, Holoubek I. GFP assay as a sensitive eukaryotic screening model to detect toxic and genotoxic activity of azaarenes. Environ. Toxicol. (2006) 21:343–348.[CrossRef][Web of Science][Medline]
10. Cooper JA, Saracci R, Cole P. Describing the validity of carcinogen screening tests. Br. J. Cancer. (1979) 39:87–89.[Web of Science][Medline]
11. Snyder RD, Green JW. A review of the genotoxicity of marketed pharmaceuticals. Mutat. Res. (2001) 488:151–169.[CrossRef][Web of Science][Medline]
12. Knight AW, Goddard NJ, Billinton N, Cahill PA, Walmsley RM. Fluorescence polarisation discriminates green fluorescent protein from interfering autofluorescence in a microplate assay for genotoxicity. J. Biochem. Biophys. Methods. (2002) 51:165–177.[CrossRef][Web of Science][Medline]
13. Kelly SL, Lamb DC, Corran AJ, Baldwin BC, Parks LW, Kelly DE. Purification and reconstitution of activity of Saccharomyces cerevisiae P450 61, a sterol delta 22-desaturase. FEBS Lett. (1995) 377:217–220.[CrossRef][Web of Science][Medline]
14. Briza P, Eckerstofer M, Breitenbach M. The sporulation-specific enzymes encoded by the DIT1 and DIT2 genes catalyze a two-step reaction leading to a soluble LL-dityrosine-containing precursor of the yeast spore wall. Proc. Natl Acad. Sci. USA (1994) 91:4524–4528.
15. Aoyama Y, Yoshida Y, Sato R. Yeast cytochrome P-450 catalyzing lanosterol 14 alpha-demethylation. II. Lanosterol metabolism by purified P-450(14)DM and by intact microsomes. J. Biol. Chem. (1984) 259:1661–1666.
16. Yabusaki Y, Murakami H, Ohkawa H. Primary structure of Saccharomyces cerevisiae NADPH-cytochrome P450 reductase deduced from nucleotide sequence of its cloned gene. J. Biochem. (1988) 103:1004–1010.
17. Choi JH, Lou W, Vancura A. A novel membrane-bound glutathione S-transferase in the stationary phase of the yeast Saccharomyces cerevisiae. J. Biol. Chem. (1998) 273:29915–29922.
18. Flückiger-Isler S, Baumeister M, Braun K, Gervais V, Hasler-Nguyen N, Reimann R, van Gompel J, Wunderlich H-G, Engelhardt G. Assessment of the performance of the Ames II assay: a collaborative study with 19 coded compounds. Mutat. Res. (2004) 558:181–197.[Web of Science][Medline]
19. Hastwell PW, Chai L-L, Roberts KJ, Webster TW, Harvey JS, Rees RW, Walmsley RM. High-specificity and high-sensitivity genotoxicity assessment in a human cell line: validation of the GreenScreen HC GADD45a-GFP genotoxicity assay. Mutat. Res. (2006) 607:160–175.[Web of Science][Medline]
Received on June 28, 2007; revised on July 24, 2007; accepted on August 14, 2007.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  X. Liu, J. A. Kramer, Y. Hu, J. M. Schmidt, J. Jiang, and A. G. E. Wilson Development of a High-Throughput Human HepG2 Dual Luciferase Assay for Detection of Metabolically Activated Hepatotoxicants and Genotoxicants International Journal of Toxicology, May 1, 2009; 28(3): 162 - 176. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. W. Knight, L. Birrell, and R. M. Walmsley Development and Validation of a Higher Throughput Screening Approach to Genotoxicity Testing Using the GADD45a-GFP GreenScreen HC Assay J Biomol Screen, January 1, 2009; 14(1): 16 - 30. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||