Mutagenesis vol. 19 no. 2 pp. 105-119,
March 2004
© 2004 UK Environmental Mutagen Society/Oxford University Press
The GreenScreen® genotoxicity assay: a screening validation programme
Department of Biomolecular Sciences, UMIST, Manchester M60 1QD, UK and 1Genetics Department, University of Wales Swansea, Singleton Park, Swansea SA2 8PP, UK
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
A yeast (Saccharomyces cerevisiae) DNA repair reporter assay termed the GreenScreen® assay (GSA) is described. This is a novel, cost-effective genotoxicity screen, developed to provide a pre-regulatory screening assay for use by the pharmaceutical industry and in other applications where significant numbers of compounds need to be tested. It provides a higher throughput and a lower compound consumption than existing eukaryotic genotoxicity assays and is sensitive to a broad spectrum of mutagens and, importantly, clastogens. We describe a simple, robust assay protocol and a validation study. The end-point of the test reflects the typically eukaryotic chromosomes and DNA metabolizing enzymes of yeast. The capacity for metabolic activation (MA) in yeast is limited compared with the mammalian liver or its extracts, but the assay does detect a subset of compounds that would require MA in existing genotoxicity tests. The GSA detects a different spectrum of compounds to bacterial genotoxicity assays and thus, together with an in silico structure–activity relationship (SAR) screen, and possibly a high throughput bacterial screen, would provide an effective preview of the regulatory battery of genotoxicity tests.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
The introduction of high-throughput bacterial screens such as the SOS umu and SOS lux tests (Oda et al., 1985
; Quillardet and Hofnung, 1993; van der Lelie et al., 1997; Verschaeve et al., 1999) have demonstrated that it is possible to apply genotoxicity testing to the small quantities of test chemical available at the early stages of the drug discovery process. This type of assay may also have an important role in such diverse disciplines as environmental monitoring, analysis of foodstuffs and military field testing. Although these bacterial assays have proved their effectiveness, they have some disadvantages, especially because they use non-eukaryotic cells and hence will not detect genotoxins that interact with eukaryote-specific targets.
The authors have previously reported the construction and preliminary validation of a yeast-based genotoxicity test system, RAD54–GFP (Walmsley et al., 1997; Billinton et al., 1998; Afanassiev et al., 2000; Knight et al., 2000) that will soon be commercially available as GreenScreen® from Gentronix Ltd. Induction of the RAD54 promoter due to DNA damage results in production of the extremely stable green fluorescent protein (GFP), which is fluorescent in the green spectrum when illuminated by blue light. Since the cell’s own DNA damage assessment apparatus is being monitored, the entire genome is used as the target for DNA damage. This is in contrast to mutation assays which monitor damage at a particular genetic locus, such as the Salmonella HIS operon in the Ames test. Many DNA-damaging agents affect other cellular targets and, depending on the exposure level, this can result in a reduction in proliferative potential as well as increasing mortality. In this assay relative total growth is assessed by comparing the extent of proliferation of treated cells with that of untreated cells. Growth data are primarily used to normalize fluorescence data collected in the assay, as it is necessary to distinguish a large concentration of weakly fluorescent cells from a small concentration of highly fluorescent cells. A detailed description of the handling of the data is presented in the Materials and methods section.
It appears that any agent capable of causing mutation in yeast will lead to RAD54 induction. Early studies using lacZ fusion reporter strains (Cole et al., 1987; Averbeck and Averbeck, 1994) revealed that RAD54 is induced above a constitutive level by a variety of different lesions induced by UV and ionizing radiation and chemical agents, including methyl methanesulfonate (MMS) and DNA cross-linking agents. The specificity of DNA damage, compared with other stressors, has been confirmed by Gasch et al. (2001), who studied the global transcriptional response to DNA-damaging agents using DNA microarrays. They found that RAD54 did not respond to non-genotoxic oxidative stress, heat shock, reductive stress, osmotic shock or amino acid starvation. Overall this is a very different pattern of regulation to that of the bacterial SOS pathway (Radman, 1974). The latter comprises a large set of cellular responses induced by a variety of both genotoxic and metabolic stresses, including starvation (Taddei et al., 1995).
RAD54 is a member of the RAD52 epistasis group of DNA repair genes and encodes a structural element of the homologous recombinational (HR) repair pathway (see Tan et al., 2003, for a review of the roles of Rad54p). It is transcriptionally up-regulated in response to a broad spectrum of genotoxins, suggesting that yeast DNA damage sensing pathways activate the HR pathway as a default for failure or saturation of the other repair pathways. This is supported by several lines of genetic evidence. For example, the loss of the non-homologous end-joining (NHEJ) pathway in yeast is undetectable unless the RAD52 (HR) pathway is also ablated (Boulton and Jackson, 1996a,b). Furthermore, the induction of RAD54 by MMS is insensitive to mutations in the ‘classical’ DNA damage sensing pathway controlled by the RAD9 and DDC1 checkpoints (Walsh et al., 2002). This suggests that there are additional DNA damage sensing pathways leading to RAD54 activation and that this may account for its broad response profile. The HR DNA repair process is well characterized (Paques and Haber, 1999; Mazin et al., 2000) and is apparently highly conserved within all eukaryotic organisms. Expression of the human homologue hRAD54 in yeast largely suppresses the phenotypes resulting from deletion of the native copy (Kanaar et al., 1996). Increased expression of RAD54 in yeast as a result of exposure to genotoxins is therefore likely to be predictive of the increased DNA repair activity in mammalian cells.
Yeast genotoxicity testing is not new. Previously developed tests targeted genetic end-points almost exclusively. The different approaches, including the detection of forward and reverse mutations, mitotic gene conversion and crossover, aneuploidy and respiratory competence, have been adequately described elsewhere (see review by Zimmermann et al., 1984). More recently deletion assays have been described (Schiestl et al., 1989). The results of these tests provide ample evidence of the utility of the yeast genome as a surrogate for the human genome as a target for compound testing.
In the development of early genetic end-point assays, considerable effort was put into discovering conditions which facilitated the detection of the activity of specific classes of compound. For example, it has been described how fast growing cells are more sensitive to chemicals requiring metabolic activation (Callen and Philpot, 1977; Callen et al., 1980). Furthermore, it has been shown that the genotoxic dose is sometimes detected over a narrow concentration range with toxic compounds (Callen et al., 1980). As a consequence, different subsets of compounds have been identified as genotoxic using different protocols with different test strains. Thus as a collection of tests yeast systems were effective, but they could not be reduced to a single, simple fixed protocol. There is a perception that yeast has an impermeable cell wall, although this is largely a result of experiences with the sporulating cultures used in studies of chromosome segregation in meiotic cells. Similarly, ‘bulky’ molecules are widely believed not to penetrate the cell wall despite the ease with which, for example, very bulky fragments of DNA can be taken up by cells treated with lithium salts. There is, however, little evidence that molecular size and cell permeability are generalized phenomena and the data presented in this paper provide evidence for both higher and lower sensitivity than, for example, the SOS response and reaction to quite bulky molecules (e.g. etoposide, mol. wt 589). These perceptions coupled to the time and compound consuming methodologies, such as ‘treat and plate’ protocols, probably explain why earlier yeast tests did not find favour in screening. Screening tests are for large numbers of compounds and are used to aid early candidate selection from the pools of compounds in the drug discovery pipeline. Multiple variants of tests using different conditions or different dilution regimes are more suited to analytical studies in which, for example, mechanisms of action are determined.
In a previous development of this assay we described a method to distinguish GFP fluorescence from any chemically derived emissions (Knight et al., 1999). Here we describe a protocol suited to the use of robotic liquid handling systems and present simple data handling protocols that provide clear graphical output. In an early validation study it is important to define the types of DNA-damaging agents detectable by a test. We report the results of a screening assay validation programme in which 102 diverse compounds were tested to assess the performance of GSA as an early (pre-regulatory) screen without investigating mechanisms.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
Strains and plasmids
The yeast strains, plasmids and growth media (F1) used in this study have been described previously (Walmsley et al., 1997
) and the use of GFP as a reporter for the DNA damage-induced gene RAD54 from Saccharomyces cerevisiae has also been described (Billinton et al., 1998). The Saccharomyces cerevisiae strain FF18984 (MATa, leu2-3,112 ura3-52 lys2-1 his7-1) was obtained from Francis Fabre (French Atomic Energy Commission, Fontenay-aux-Roses, France). The strain has been neither modified to increase permeability nor sensitized to DNA damage by mutations.
The reporter strain (GenT01) is FF18984 containing a nuclear, episomally replicating, multiple copy plasmid bearing the entire upstream non-coding DNA sequence of the S.cerevisiae RAD54 gene fused to a yeast codon-optimized derivative of the Aequorea victoria (jellyfish) GFP gene (Cormack et al., 1997). The control strain (GenC01) is FF18984 containing an identical plasmid except that 2 bp have been removed at the start of the GFP gene, such that no GFP is made. The plasmids are maintained during cell growth and division by selection of uracil prototrophy, conferred by the presence on both plasmids of the yeast URA3 marker gene.
Microplate preparation
Assays were carried out in 96-well, black, clear-bottomed microplates (Matrix ScreenMates, catalogue no. 4929; Matrix Technologies, Hudson, NH). A number of alternative microplates were assessed, although the variable background absorbance and fluorescence both within and between plates from individual manufacturers were generally unacceptable, leading to the conclusion that only Matrix or Corning (catalogue no. 3651; New York, NY) plates were appropriate for the assay at the time of writing. The assays were performed using a liquid handling robot (MicroLabS single probe; Hamilton GB Ltd, Birmingham, UK) in a protocol designed to test up to four compounds on a single 96-well microplate (see online supplementary material: Microplate layout). Set-up takes 30 min per plate. Results for a subset of compounds have been reproduced using a Genesis 8-probe robot (Tecan UK Ltd, Theale, UK), which can set up a similar microplate in <5 min. Microplates can also be rapidly and effectively filled using a multi-channel pipette. The arrangement of samples in the microplate is also well suited to manual filling. It is reasonable to test up to 50 compounds per day using the manual protocol and considerably more using a multi-probe robot.
A microplate version of the assay has been previously reported (Afanassiev et al., 2000), but different microplate layouts and controls were used for this study, so more detail is presented here. The following standard protocol was followed. A 1 mM stock of the test chemical was prepared in 2% (v/v) aqueous DMSO and used to make two identical dilution series across the microplate and a ‘control’ (see below). To achieve this, 150 µl of the test chemical solution were put into two microplate wells. Each sample was serially diluted by transferring 75 µl into 75 µl of 2% DMSO, mixing and then taking 75 µl out and into the next well. This produced nine serial dilutions of 75 µl each.
Controls were added as follows. (i) Compound alone, to provide information on compound absorbance/fluorescence. (ii) Yeast cultures diluted with 2% DMSO alone, to give a measure of maximum proliferative potential. (iii) MMS as a genotoxicity control: ‘high’ = 0.00125% (v/v), ‘low’ = 0.0001875% (v/v). (iv) Methanol as a cytotoxicity control: ‘high’ = 3.5% (v/v), ‘low’ = 1.5% (v/v). (v) Growth medium alone, to confirm sterility/lack of contamination.
Stationary phase cultures of GenT01 and GenC01 were diluted to an optical density (OD600 nm) = 0.2 in double strength F1 medium (Billinton et al., 1998). An aliquot of 75 µl of the yeast suspension was added to each well of the diluted chemical: GenT01 to one series and GenC01 to the second series of each compound, and to appropriate standards and controls (i.e. GenT01 to MMS-containing wells and GenC01 to methanol-containing wells). After the plates were filled, they were sealed using either a gas-permeable membrane (Breath-Easy; Diversified Biotech, USA) or a plastic lid and then incubated, without shaking, overnight (16–20 h) at 25°C.
Compounds chosen for the study
In order to compile a list of chemicals (Table I) unbiased by the authors’ laboratory, a number of pharmaceutical companies and contract research organizations were asked to suggest compounds for which results would be of interest to them in assessing a new genotoxicity screening test. To be successful in this context the results of the test should be of value in identifying compounds likely to score positive in one or more of the regulatory genotoxicity tests. The latter require substantially more material than is available at the early screening stage and are very time consuming and expensive. It is therefore desirable to remove such compounds. In addition to the compounds suggested to us, the basic dye crystal violet was added because of its role in the assessment of permeability in the Ames Salmonella test. Urethane, safrole and thiourea were selected from a list of mutagenic and non-mutagenic carcinogens detectable by the yeast DEL assay (selects for DNA deletion events) but not detectable in the Ames assay (Schiestl et al., 1989). All compounds were sourced at analytical purity where available (Sigma, Aldrich, Fluka, BDH and Avocado).
|
Data collection and handling
Following overnight incubation, GFP reporter fluorescence and yeast culture absorbance data were collected from the microplates. Three different microplate readers which combine fluorescence and absorbance functionality have been used, and comparable data were obtained. These were: a Tecan Ultra-384 (Tecan UK Ltd, Theale, UK), excitation 485 nm, emission 535 nm with an additional dichroic mirror (reflectance 320–500 nm, transmission 520–800 nm); a BMG PolarStar (BMG Labtechnologies, Offenburg, Germany), excitation 485 nm, emission 520 nm; a Wallac 1420 Victor2 (Perkin Elmer Life Sciences, Wellesley, MA), excitation 485 nm, emission 535 nm. Absorbance was measured through a 620 nm filter in both the Tecan and BMG instruments and through a 600 nm filter in the Wallac. The data were inserted into a Microsoft Excel spreadsheet and converted to graphical format (see typical data in Figure 1). Data processing requires only simple mathematical manipulations (see online supplementary material: Example data sheet, Data handling). Absorbance data give an indication of reduction in proliferative potential and these data were normalized to the untreated control (= 100% growth). Fluorescence data were divided by absorbance data to give ‘brightness units’, the measure of average GFP induction per cell. These data were normalized to the untreated control (= 1). In order to correct for induced cellular autofluorescence and intrinsic compound fluorescence, the brightness values for the GenC01 strain were subtracted from those of GenT01. This makes visual assessment of the data more reliable. All the data were checked with and without this correction, and the decision (see below) on whether or not a compound was classified as being genotoxic was not affected.
|
Decision thresholds
It is useful to have clear definitions of positive and negative results from routine assays and such definitions have been derived, taking into account the maximum noise in the system and data from chemicals where there is a clear consensus on genotoxicity and mechanism of action. Naturally, it is also possible for users to inspect the numerical and graphical data. Marks and/or damage to microplate bases, bubbles in individual wells and user errors can all introduce spikes in microplate data and these are recognized simply by inspection of the graphical data. However, visual inspection can also help in decision making. For example an upward trend in genotoxicity data that did not cross the threshold might distinguish two related compounds and allow a decision on selection. The decision thresholds used in data analysis for this paper were set as follows.
Two thresholds were set from the absorbance data. The first is the threshold at which there is a statistically significant reduction in the proliferative potential or relative total growth (RTG). This threshold is not used in data handling but is provided to give an indication to the user that the compound is causing some growth inhibition. It is set at 80% of the maximum extent of yeast cell proliferation on each microplate (i.e. the cell density reached by the untreated control cells). This is greater than 3 times the standard deviation of the background. Mortality is not measured in this assay: 80% RTG does not mean 20% of the cells are dead. In Table I + is recorded if one or two compound dilutions produce a final cell density below the 80% threshold, ++ is recorded if either (i) three or more compound dilutions produce a final cell density lower than the 80% threshold or (ii) at least one compound dilution produces a final cell density lower than 50% and – is recorded when no compound dilutions produce a final cell density lower than the 80% threshold. The lowest effective concentration (LEC) is the lowest test compound concentration that produces a final cell density below the 80% threshold.
The second threshold is set at 30% RTG. This is a rejection threshold for genotoxicity data and reflects two properties of the system. Firstly, this threshold recognizes the limits imposed by instrumentation: at cell densities lower than 30% RTG, interference in the optical measurements becomes significant due to variation in the background reflectance and absorbance of the microplate. Secondly, this level of growth means that the culture has been unable to complete even one doubling and as such is a toxicity threshold. A breakdown in cell integrity can lead to non-specific DNA damage, although if a cell is dead or dying this is clearly of little genetic consequence: apoptosis in mammalian cells is itself a deliberate manifestation of this. All the positive genotoxicity results presented in this paper are from concentrations of substance at which RTG was in an acceptable range (>30%).
The genotoxic threshold reflects a statistically significant increase in brightness compared with the untreated control. It is set at 1.3 (i.e. a 30% increase) and this is greater than 3 times the standard deviation of the background. A positive genotoxicity result (+) is concluded if one or two compound dilutions produce a relative GFP induction greater than the 1.3 threshold. A strong positive genotoxicity result (++) is concluded if either (i) three or more compound dilutions produce a relative GFP induction greater than the 1.3 threshold or (ii) at least one compound dilution produces a relative GFP induction greater than a 1.6 threshold. A negative genotoxicity result (–) is concluded where no compound dilutions produce a relative GFP induction greater than the 1.3 threshold. The LEC is the lowest test compound concentration that produces a relative GFP induction greater than the 1.3 threshold.
Compounds that tested negative for genotoxicity in the first assay were re-tested up to the 30% RTG threshold or the limit of solubility. Compounds that tested positive for genotoxicity were re-tested at the same compound concentrations to corroborate the initial result.
Cell proliferation and genotoxicity reporter controls
The cytotoxicity controls indicate that the yeast is behaving normally. The ‘high’ methanol standard should reduce the final cell density to below the 80% threshold and should be a lower value than the ‘low’ standard. If it is not, a warning can be generated automatically, indicating that the system is not detecting a dose-dependent effect on proliferation.
The genotoxicity controls demonstrate that the strains are responding normally to DNA damage. The ‘high’ MMS standard must produce a fluorescence induction >2 and be a greater value than the ‘low’ MMS standard, indicating a dose-dependent effect on the reporter. A warning can be generated automatically.
The growth inhibition and genotoxicity controls are in place for qualitative reasons to demonstrate that the assay is responding in a dose-dependent manner. The exact figures for relative total growth and fluorescence induction of the controls are not used in any calculations of toxicity/genotoxicity of the particular compounds being analysed. Controls are included on each microplate tested. Considering the results from 24 tests conducted over a period of 7 months, all controls ‘passed’ by the criteria prescribed. The average relative cell densities of the high and low methanol controls were 58.4 and 77.2, respectively. The average relative GFP inductions of the high and low MMS controls were 2.67 and 1.63, respectively. The average ratio of the high and low standard results, i.e. dose response, was 0.75 for methanol (relative standard deviation = 10.6%, n = 24, where relative standard deviation is the standard deviation expressed as a percentage of the mean value) and 1.64 for MMS (relative standard deviation = 6.7%, n = 24).
Test compound controls
The compound absorbance control allows a warning to be generated if a test compound is significantly absorbing. If the ratio of the absorbance of the compound control well to a well filled with diluent alone is >2, there is a risk of interference with interpretation.
The compound fluorescence control allows a warning to be generated when a compound is highly autofluorescent. If the ratio of the fluorescence of the compound control well to a diluent filled well is >5, there is a risk of interference with interpretation. In these cases (four in this study), fluorescence polarization can be used to distinguish GFP from other sources of fluorescence (Knight et al., 2000, 2002). The Tecan, BMG and Wallac instruments all have this facility. Occasionally, compounds induce cellular autofluorescence although not fluorescent themselves. This is apparent from dose-dependently increasing brightness in the control (GenC01) strain in the absence of fluorescence from the chemical-only control. The routine subtraction of GenC01 from GenT01 data removes this interference from the data.
Comparisons
Published results from other genotoxicity tests and cancer studies have been tabulated. They include the Ames test, mouse lymphoma assay (MLA), in vitro and in vivo cytogenetics (chromosome aberration assays), rodent micronucleus test (MNT) and rodent carcinogenicity. The data have been collected both from the peer-reviewed literature and from freely available Internet resources. These include CCRIS (Chemical Carcinogenesis Research Information System, http://toxnet.nlm.nih.gov), NTP (National Toxicology Program Reports, http://ntp-server.niehs.nih.gov), IARC (International Agency for Research on Cancer, http://www.iarc.fr), NIOSH (National Institute for Occupational Safety and Health, http://www.cdc.gov/niosh/homepage.html) and the USEPA (Environmental Protection Agency, http://www.epa.gov/iris/search.htm). Chromosome aberration test data were largely drawn from Ishidate et al. (1988).
There are several compounds where there are conflicting data from different sources (recorded as +/–). However, where comparisons have been drawn these are recorded as positive (+), to reflect the potential positive result in a regulatory test. The only exception to this is where NTP has more recently suggested a definitive designation. 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.
Several different indices have been used to assess the utility of new tests, and the relative value of these has been discussed elsewhere (Cooper et al., 1979). The appropriate terms and their definitions are taken from table 1 in the latter paper and are paraphrased as follows.
The test with GSA can have a positive outcome, for which there may be either a positive result (‘a’) or a negative result (‘b’) from another test (for example MLA). 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 ‘N’ is ‘a + b + c + d’. The following terms were calculated from these data: sensitivity, a/(a + c); specificity, d/(b + d); predictive value, a/(a + b); prevalence, (a + c)/N. The relevance of these terms to the current programme is discussed later in the paper.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
Validation screening programme
The results of the validation study incorporating 102 chemicals are shown in full in Table I. Even in a collection of this size, the results of comparisons must be considered as suggestive rather than definitive.
Typical results
Figure 1 shows representative results classified on the basis of the GSA, including a range from toxic genotoxins to non-toxic non-genotoxins. MMS causes a progressive reduction in cell yield, with two data points below the 80% RTG threshold (+). The GFP induction threshold is passed before there is any significant effect on RTG and there are six data points above it, resulting in a ++ score for genotoxicity. N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) has a greater effect on cell yield than MMS, giving a steeper fall in RTG with three data points below the 80% threshold (++). MNNG is strongly genotoxic (++), with three data points above the GFP induction threshold, all above the toxicity threshold of 30% RTG. Bleomycin has a marked affect on RTG (++), apparent over a wide concentration range, whereas GFP induction (++) rises sharply. N-nitroso-N-methylurea has far less effect on growth (–) than the previous examples, but like bleomycin causes a sharp increase in GFP induction (++). Benzaldehyde shows a sharp transition from no affect on RTG to below 50% at two concentrations (++). This is not untypical of the results with aldehydes in other genotoxicity assays. GFP induction shows a steadily climbing trend that crosses the threshold with two points (+). Aphidicolin shows a gradual decrease in RTG (+) and increase in GFP induction (+). The final two compounds do not induce GSA (–) and thus would be scored as non-genotoxic, but whereas 3,5-dichlorophenol causes a great reduction in RTG (++) and thus inhibits cell growth in the tested range, ampicillin has no effect (–) on RTG.
Comparisons of GSA with regulatory tests and carcinogenicity tests
With a new assay there is a need to provide some comparative data with existing tests to help provide some guidance on the relative value of the new assay and its potential place in testing strategies. Thus there has to be reliance on published data for this comparison. The published dataset for the test compounds in this study is not untypical of what is available for any dataset of compounds of interest, in that for most compounds the data on the standard battery of regulatory tests is incomplete: note particularly the variability in ‘N’, the number of compounds for which there was data for any given comparative test (see Table II). In addition, as discussed elsewhere in this paper, this dataset contains few compounds that are negative in all reported genotoxicity tests. This is reflected in the high prevalence figures, the proportion of test data from each comparative test that is positive. There are also conflicting data on the same tests in some cases (listed as +/–). It is apparent that for individual compounds there is discordance between the results of individual tests in the test battery, with few compounds being positive in all the reported tests. It is well known that there are bacteria-specific genotoxins, mammalian cell-specific genotoxins, in vitro specific genotoxins, in vivo specific genotoxins, etc., which will result in such discordance. Also, protocol variations, such as S9 conditions, choice of solvent, etc., can change the outcome of these tests on specific compounds. It is worth noting that even comparisons between tests that measure the same end-point can show concordance figures below 80% [e.g. in vitro chromosome aberration assay and MLA (Honma et al., 1999)].
|
The most striking figures (Table II) are for the predictive values. Fifty-five compounds were positive for genotoxicity in GSA, of which 50 have some regulatory test data. Forty-seven of these 50 GSA positive compounds were positive in at least one of the regulatory tests. There were only two compounds positive in GSA for which there was no corresponding positive data from other tests (considered in Discussion below). However, with all the caveats above the reader should be aware that the comparisons made here give limited guidance only. The coverage of compound types discussed below provide the more valuable information from the programme.
Table III contains a subset of the data for which there are Ames data and highlights the distinctively different end-points for GSA and bacterial genotoxicity results. In this group of 93 compounds, 18 that are positive in GSA are negative for the Ames test (with or without S9). Seventeen of the 18 are, however, positive in animal carcinogenesis or mammalian cell tests. Six of 31 GSA positives that are Ames positive required metabolic activation for the Ames result. Fifteen Ames positive compounds were negative in GSA, of which six are only Ames positive with metabolic activation. Results were also compared with the SOS umuC test (Table II), conceived as a screening preview of the Ames test (Reifferscheid and Heil, 1996).
|
Robustness and reproducibility
In an assessment of reproducibility, the assay was performed using MMS (32.5 µg/ml), MNNG (2.5 µg/ml) and methanol (3.5% v/v) as test compounds, 20 times over a period of 7 months (17 times in the case of methanol). The genotoxicity results for MMS and MNNG and the cytotoxicity results for methanol are shown in Table IV, along with the lowest effective concentration (LEC) which produced a positive effect. The results encompass variability from a variety of sources. These include two different operators, different stock yeast cultures grown freshly every week, each experiment carried out on a different day, different stocks of growth medium made approximately monthly and each test chemical solution made up daily by a series of dilutions from the pure chemical. The assay gave the same qualitative strong positive genotoxicity result for MMS and MNNG in each case and all LEC results were within one serial dilution of the most frequently occurring (modal) value. The assay gave the same qualitative positive cytotoxicity result for methanol in each case, with the exception of two runs in week 9 where a strong positive was recorded. Again, all LEC results were within one serial dilution of the modal value.
|
In a second series of experiments the reliability of the RTG threshold was demonstrated. Two compounds were selected: ampicillin, which is negative in this test and not toxic in the range tested (top concentration = 200 µg/ml); 3,5-dichlorophenol (3,5-DCP), which is negative in this test and highly toxic (top concentration = 25 µg/ml). Over a 5 month period ampicillin was tested 15 times and 3,5-DCP 11 times. Neither compound gave a positive genotoxicity result, allowing the constitutive reporter activity data to be analysed statistically. The 10 genotoxicity data points (one from each concentration, including untreated) for each time point were used to calculate an average value (AV) and standard deviation (SD). An average for AV and SD for the whole set was calculated for each compound and the SD multiplied by 3 and added to the average figure. For ampicillin the value was 1.13 (n = 15, confidence 99.7%) and for 3,5-DCP the figure was 1.22 (n = 11, confidence 99.7%). Since neither compound shows sufficient variation to pass the 1.3 threshold it was concluded that the genotoxicity threshold is robust; false positive results will be very rare, but some weakly active agents may fail to cause an increase as large as the 1.3-fold threshold. Since there is no fall in RTG for ampicillin, the growth data for each of the 15 time points provides 10 data points whose variation reflects noise in the OD measurement. The AV and SD for each of the 15 data sets were calculated, as well as the AV and SD for all 15 combined. The AV (100.33%) minus 3 SD (10.83) gives a value of 89.17%, which is well within the set threshold, demonstrating that the RTG threshold is robust. A similar calculation cannot be performed for 3,5-DCP as its toxicity leads to falling RTG with increasing concentration.
Alternative solvents
Table V lists chemicals/solvents that are often used with test samples, to promote dissolution, etc. All of these compounds are toxic at higher concentrations due to osmotic effects and/or effects on membranes and interfere with data interpretation. If these chemicals are used to dissolve compound stock solutions, they should also be included in the diluent used to make the compound dilution series. The routine use of DMSO in this study revealed its predictable interference with certain compounds (Gebel and Koenig, 1999). For example, cisplatin was only positive when re-tested using water in place of 2% DMSO as the diluent. Brennan et al. (1994) have previously reported interference of DMSO, a free radical scavenger, in the recombinogenic properties of several oxidizing agents. It has been noted previously that yeast is relatively solvent tolerant: it is also pH tolerant (Parry, 1999) and although this property has not been exploited in the present study, it is nevertheless of relevance when compound solubility is an issue.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
Induction of the yeast RAD54 gene has been monitored using a promoter–GFP fusion. In a previous study (Afanassiev et al., 2000
) a small sample of chemicals (12) was selected to provide an indication of the range of DNA-damaging agents that could be detected. That study used a slightly different experimental protocol, although the results demonstrated a good coverage of DNA-damaging agents as well as enzyme inhibitors, including DNA polymerase (aphidicolin) and topoisomerase 1 (camptothecin) inhibitors. Validation of new assays is not a simple or short exercise as comparisons have to be made with published data on established assays that are invariably incomplete and of variable quality. In addition, it is difficult to choose a limited validation set of chemicals that is not biased to one degree or another. This is true of the set of 102 chemicals used in this exercise. The set of chemicals in Table I was chosen to be of specific interest, after discussion with potential collaborators. It includes a variety of compounds that are well-known genotoxins as well as compounds that are less well studied. One drawback of the selection is that there are only eleven compounds that do not possess a known positive result in one or more of the established genotoxicity/carcinogenicity tests compared here (from the available information) and a further six with no such data at all.
The paucity of compounds with only negative data and the large number of compounds with conflicting data from the regulatory tests in this data set serve to illustrate the incomplete agreement between any of the regulatory genotoxicity assays, in terms of the compounds detected as genotoxic. A proportion of this discordance is due to differences in genetic target for the genotoxins in question. This is why a battery of tests is included in international guidelines for genotoxin detection. The reasons for incomplete agreement are manifold. Some genotoxins are positive solely in bacteria due to bacteria-specific metabolism (see for example Arenaz et al., 1989; Gocke and Albertini, 1996). Genotoxins that are also bactericidal are by definition difficult to detect in bacteria-based assays. Some genotoxins induce chromosome damage by interacting with eukaryote-specific targets and thus do not induce mutations in bacteria. Recent analysis of regulatory submissions have shown that chromosome aberration assays and the MLA yield about 4-fold higher incidences of positive test results than the bacterial reverse mutation test or in vivo bone marrow tests (Müller and Sofuni, 2000). This discrepancy is possibly due to the results of tests carried out at high cytotoxic doses in the latter assays resulting in ‘false positives’. Analysis of published test data using the same genetic end-point, chromosome damage, but with different target cells also shows <80% concordance in genotoxin detection (Honma et al., 1999), although this is probably due to protocol differences (Galloway et al., 1997). These reasons for discordance all add to the complexity of the assessment of the validity of a new assay.
In addition to the above considerations on available data from regulatory tests, it is necessary to comment on the evaluation of the GSA in the context of its intended application in drug discovery screening. The assessment criteria applied here are different to those employed in evaluating a carcinogen screening test as the consequence associated with missing a carcinogen is likely to be more serious (exposure to animals or human recipients) than misclassifying a non-carcinogen (loss of a potential drug). To be successful in the context of an early warning genotoxicity screening test, the results should be of value in identifying compounds likely to score positive in one or more of the later regulatory genotoxicity tests. These require substantially more material than is available at the early screening stage and are very time consuming and expensive. It is therefore of value to reduce the number of compounds that reach this stage. The penalty for misclassifying a genotoxic compound (false negatives that reflect low sensitivity) is low as it would be picked up in the later regulatory test. The penalty for misclassifying a non-genotoxin (false positives that reflect low specificity) is of greater concern, as this could contribute to the decision to abandon a potentially useful compound. This has to be an accepted consequence of the regulatory tests as it is acknowledged that there is a general tendency to ‘overcall’ genotoxicity in certain tests (discussed above). This is exemplified by the data set as a whole: 82 of the compounds have at least one positive result for a regulatory test, however, only 31 of these have only positive regulatory test data. Twenty-three of those 31 are also positive in GSA and only two of the GSA positives have no associated positive regulatory data.
The use of the predictive value (PV) is limited to screening programmes such as this, where a screening test is applied to a specified group of compounds or substances. Even in programmes such as this, PV is profoundly influenced by prevalence (Cooper et al., 1979). This particular group of compounds has high prevalence values (
69%; see Table II) for all but the two bacterial tests and it is unsurprising that the tests with the highest specificities (chromosome aberration, 83%; rodent carcinogenicity, 76%) have the highest PVs (95 and 89%, respectively). Thus this validation exercise can only be regarded as preliminary and any comparative information such as specificity, sensitivity, accuracy, etc., needs to be regarded as illustrative and is not definitive. In addition, it is not helpful to deduce false positive and false negative rates from these data.
This broader study of 102 compounds (Table I) confirms that the RAD54 gene is induced by a variety of DNA damage lesions caused by direct-acting agents. These include examples of compounds which amongst other effects can cause base alterations (MMS and EMS), clastogenicity (i.e. chromosome breakage) [bleomycin, catechol, 1,2-dimethylhydrazine, methyl viologen, methapyrilene, diethylhexylphthalate (DEHP) and phleomycin] and cross-linking (mitomycin C and cisplatin). It is interesting that the photomutagen psoralen was GSA positive as the test was not intentionally exposed to UV light. However, no attempt was made to protect the assay from normal laboratory lighting during its preparation and testing. The reporter also responds to a variety of genotoxins that may not act directly on DNA. These include compounds that target topoisomerases (etoposide and ellipticine) as well as anti-mitotic, spindle targeting compounds (aneugens), such as colchicine and econazole. The latter results are interesting because they suggest that some aneugens can also lead indirectly to DNA damage.
The spectrum of metabolic enzymes expressed by yeast is less complex than that of mammalian cells. There are, for example, the CYP (cytochrome P450) type enzymes [ERG5/CYP61 (Kelly et al., 1995), DIT2/CYP56 (Briza et al., 1994) and P45014-DM/ERG11/CYP51 (Aoyama et al., 1984)], the P450 oxidoreductase NCP1/CPR1 (Yabusaki et al., 1988) and glutathione S-transferases GTT1 and GTT2 (Choi et al., 1998). This proficiency is reflected in the identification of six compounds producing positive data with GSA that require S9 for a positive Ames result (Neutral red, 2-amino-4-nitrophenol, proflavin hemisulfate, ethidium bromide, benzo[a]pyrene and 1-naphthylamine). S9 was not used in any of the GSA tests reported in this paper because one aim of this preliminary study was to provide an overview of the inherent capacity for metabolic activation of promutagens in this yeast strain under these conditions.
Safrole, urethane and thiourea are carcinogens not detected by the Ames test. All of these require metabolism to form genotoxic metabolites in vivo. They are all difficult to detect as active in most well-used in vitro genotoxicity assays. All three are positive in the GSA at concentrations >500 µg/ml, although toxicity was not sufficient to reduce cell proliferation below the 30% threshold. All three have been shown to be active in a variety of yeast-based screens and there is some evidence that urethane and thiourea may induce free radicals in yeast cells (Brennan and Schiestl, 1998). Safrole is genotoxic in vivo, possibly through the formation of an epoxide (Guenthner and Luo, 2001). This metabolite appears to be difficult to form in vitro using rat liver S9, thus safrole is negative in most in vitro genotoxicity assays. However, there are several reports that safrole is active in yeast-based systems, including those designed to detect induced DNA repair (Simmon, 1979; Sharp and Parry, 1981; Schiestl et al., 1989). Urethane is known to require metabolism by P450 CYP2E1 and possibly esterases to produce clastogenic and mutagenic metabolites in vivo, such as vinyl carbamate epoxide. As with safrole, this appears difficult to model in vitro. However, like safrole there are a number of reports that urethane is detectable as a genotoxin in different yeast models (Zimmermann and Mohr, 1992; Hubner et al., 1997; Brennan and Schiestl, 1998). With regard to thiourea, it appears that genotoxicity may be due to the formation of S-oxygenation products (Ziegler-Skylakakis et al., 1998). Thiourea has been shown to be active in other yeast assays (Brennan and Schiestl, 1998). Finally in this group, DEHP is positive in the GSA. This compound is categorized as a non-genotoxic mouse hepatocellular carcinogen that acts via peroxisome proliferation (Melnick et al., 1987). The literature shows that most, but not all, genotoxicity assays are negative for DEHP (see for example Elliott and Elcombe, 1987). However, there is some evidence that this compound can induce or influence the induction of DNA double-strand breaks (Kawai, 1998), which is consistent with the positive response seen in the GSA assay.
Importantly for a proposed screening test, only two positive results were unique to this yeast test (‘false positives’), tritolyl phosphate and cimetidine. The result with cimetidine (cyano-methylguanidine) is intriguing, since whilst cimetidine itself does not appear to be genotoxic, there are reports of positive UDS results in rat, but not human, hepatocytes (Martelli et al., 1986) that may be artifactual (Lefevre and Ashby, 1985). Cimetidine and in particular N-nitrosocimetidine is a methylating agent chemically related to MNNG. N-nitrosocimetidine is formed under acidic conditions in the presence of nitrite and is clearly genotoxic in vitro (De Flora and Picciotto, 1980; Ichinotsubo et al., 1981). Neither cimetidine nor N-nitrosocimetidine induce tumours in rodents, as nitrosocimetidine rapidly denitrosates in blood (Jensen et al., 1987) and thus does not express its genotoxicity and potential carcinogenicity in vivo (Habs et al., 1982).
About half the compounds in this study were negative in the GSA and largely expected to be so on the basis of other test results (e.g. acetylsalicyclic acid, nicotine and 3,5-DCP). Further ‘true negatives’ not present in Table Ib have subsequently been confirmed as such in the GSA, but using a slightly higher (2%) DMSO concentration (EDTA, glucose, sucrose, D-mannitol, Triton X-100, Tween 20, amoxicillin and vanillin). Several compounds, discussed below, were negative for genotoxicity in the GSA and had either positive or equivocal data in some or all of the regulatory tests. All tests have their own ‘false negatives’ and although at one level these reflect genetic diversity, they are an important part of the definition of a new test. They are perhaps less important in a screening test where the few compounds that are missed will be identified further down the development pathway.
Cumene hydroperoxide is negative despite other oxidizing agents being positive in GSA. There was no attempt to reduce evaporative loss from volatile compounds and the exposure will have been lower than that calculated. However, it has recently been reported that yeast glutaredoxins are active as glutathione reductases in yeasts and they have been demonstrated to directly reduce cumene hydroperoxide (Collinson et al., 2002). A similar explanation is plausible for hydroquinone, which is negative in GSA but an in vivo mutagen in some other tests (Adler and Kliesch, 1990). The Committee on Mutagenicity of Chemicals in Food, Consumer Products and the Environment concluded that ‘hydroquinone is an in vivo mutagen in somatic cells’ but that the ‘risk to human health would be likely to be greatly reduced by rapid conjugation and detoxification via the glutathione pathway’ (Parry, 2000).
The metabolic capability of yeast contrasts with the enhanced activities in S9 mix required for the activation of most promutagens and explains the most obvious group of compounds not detected by GSA: primary aromatic amines and aromatic amides. Nine such compounds were tested. Three were positive in GSA: 2-amino-4-nitrophenol, 1-naphthylamine and 9-aminoacridine (which is direct acting). The remaining six were negative in GSA and positive in other in vitro tests: 2-acetamidofluorene, 2-aminoanthracene, 4,4-oxydianiline, o-anisidine, aniline and 4-aminophenol. The last two of these were also negative with and without S9 in the Ames test. None of the six GSA negatives are particularly bulky (see below), and other investigators (Sengstag et al., 1994; Imaoka et al., 1996) have recognized this particular deficiency as a metabolic problem. Sengstag’s group demonstrated that the metabolic competence of yeast strains could be enhanced by the expression of genes encoding selected components of the metabolic activation system present in human liver (Phase I: cytochrome P450 CYP1A2, the recycling enzyme, NADPH cytochrome P450 oxidoreductase; Phase II: N-acetyltransferase) (Paladino et al., 1999). This resulted in strong recombinagenic and mutagenic activity in cells exposed to the two aromatic amines most abundant in cooked meats, 2-amino-3-methyl-imidazo[4,5-f]quinoline (mol. wt 198) and 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (mol. wt 224), and weaker activities in several others. In preliminary experiments using plasmids from the Sengstag group in the yeast strains used in this study, the RAD54–GFP reporter has also increased its spectrum of response. A full assessment of these strains is underway (including an assessment of the S9 protocol) and these will be reported when complete. Two other negatives, isobutyl nitrite and N-nitrosodimethylamine, also require metabolic activation. It is worth noting that the aromatic amine motif is well known and as a consequence it is recognised by SAR programmes, so whilst missed by GSA, compounds containing such moieties would be unlikely new drug candidates.
The yeast cell wall forms a physical barrier to some external solutes, largely due to its mannoprotein content (Zlotnik et al., 1984). In addition, yeast cells are able to actively expel a wide range of drugs and chemical compounds by a complex network of pleiotropic ATP binding cassette transporters, e.g. the Pdr5, Yor1 and Snq2 membrane proteins (see for example Kolaczkowski et al., 1998). The GreenScreen assay is performed using haploid cells, which do not present the same permeability barrier as that presented during meiosis and sporulation of a diploid cell, when a protective dityrosine layer is formed around developing spores (Miller, 1989; Briza et al., 1990). Permeability in the Ames strains is increased by the rfa mutation and routine confirmation of the mutant phenotype in tester strains is confirmed by determining sensitivity to bulky molecules such as crystal violet (Ames et al., 1973). In the present study bulky molecules such as etoposide (mol. wt 589), streptonigrin (mol. wt 506) and mitomycin C (mol. wt 334) scored positive. Crystal violet (mol. wt 408) was subsequently tested in 2% DMSO and also scored positive. ICR191 acridine mutagen is the bulkiest false negative (mol. wt 451) and has not been reported as positive in other yeast tests. The related mutagenic compound 9-aminoacridine (Young et al., 1981) tested positive in GSA, although it has been reported as negative in some yeast tests (Iwamoto et al., 1985): it is a smaller molecule (mol. wt 231) than ICR191 and thus it is plausible that size was an issue in this case. It does not appear generally, however, that with this yeast strain and protocol the cell wall/membrane was necessarily significant in producing negative results in the GSA with those specific compounds expected to give positive results. However, as suggested in a previous study (Afanassiev et al., 2000), we are currently assessing mutants defective in different transporter systems (e.g. YOR1, SNQ2 and PDR5) and cell wall metabolism (PSA1 and SEC53) and preliminary results show these strains to be more effective in the detection of some DNA-damaging agents. Chlorambucil, a weak alkylating agent and potent clastogen, was a surprising negative result in this screen and may be a candidate for expulsion from the cell by such transport mechanisms.
To assess the relative compound requirement of the GSA when compared with bacterial screening tests, the LECs of a small sample of the validation chemical set (where corresponding figures have been published; Reifferscheid et al., 1991; Verschaeve et al., 1999) are compared in Table VI. This comparison indicates that the GSA is apparently of comparable sensitivity to the SOS umuC chromotest and the SOS lux bioluminescence assay (based on a very limited data set). Therefore, the total compound requirement of the GSA is comparable with other commonly used bacterial short-term screens. Snyder (2003) has recently drawn attention to the expanding role of genetic toxicology in drug discovery, noting how the requirement for multigram quantities of compound for regulatory tests has historically relegated genotoxicity testing to the later stages of drug development, a stage when there has already been huge resources expended. The compound requirement for this test (0.5 mg) contrasts favourably even with that of the miniaturized versions of the Ames assay and in vitro micronucleus assay that require 5–30 mg of compound (Snyder, 2003).
|
The assay is not intended for more analytical studies (such as mechanism of action), however, where a genotoxin is known to give a positive response in the assay, its easy read-out lends itself to use in monitoring applications such as environmental contamination, as well as compound screening. The GSA can be performed rapidly, with small amounts of compounds and has a simple and fixed microplate protocol with clear decision making thresholds. Together these properties were conceived to give a high-throughput capability and to minimize the generation of conflicting data by different users. These are important requirements of any useful screen (Redman, 1981
). There are several conclusions that can be drawn from this study. Analysis of the data in Table I indicates that the GSA screen detects a different spectrum of genotoxins to any of the individual assays used in regulatory test batteries. This is not surprising as the assay measures a different end-point (DNA repair induction), in a different cell type (yeast). DNA repair induction is often an early response to genotoxins, before genetic damage becomes fixed as a gene mutation, chromosome damage, recombinational events, etc. The chemicals testing positive in this validation exercise covered a wide mix and breadth of chemicals, acting via a variety of end-points and mechanisms, thus DNA repair induction in yeast appears to be a reliable indicator of genotoxicity. Furthermore, it is clear that the assay is capable of detecting some genotoxins that are positive in eukaryotic cells only, including a significant proportion of known clastogens.
There are examples of genotoxic carcinogens that for sound mechanistic reasons are positive in only some of the tests in the battery, e.g. specific clastogens and mammalian cell-specific genotoxins. Therefore, in regulatory terms an unequivocal positive result in one or more genotoxicity tests in the test battery can be sufficient to stop drug development or, at the very least, cause a delay in development where further testing is required to determine if the test compound is a confirmed genotoxic hazard. Positive regulatory genotoxicity results will also appear on the drug label regardless of the overall conclusion on safety. From this data set and the comparisons made, it appears that positive results in the GSA can be used to identify drug candidates that are very likely to give positive results in the regulatory test battery. It is suggested that compounds that are negative in the current GSA screen should undergo SAR evaluation and possibly high throughput bacteria-based screens. This combination of rapid pre-regulatory screens is very likely to be effective in minimizing the number of genotoxins that fail the regulatory test battery.
|
Supplementary material |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
Supplementary material can be found at:http://www.mutage.oupjournals.org/
|
Acknowledgements |
|---|
The authors would like to thank those in the pharmaceutical industry who helpfully suggested the names of compounds to be tested in this validation exercise, especially Neal Cariello (GlaxoSmithKline, NC) and Patricia Collins (Sequani Ltd, Ledbury, UK). This project was supported at UMIST by a research contract with Gentronix Ltd. The RAD54–GFP reporter system is protected by patents in the USA and Australia and pending in the EU, Canada and Japan.
|
Conflict of interest statement |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
During this work P.A.Cahill, A.W.Knight, N.Billinton and M.G.Barker were Research Associates employed at UMIST through research contracts with Gentronix Limited; L.Walsh was employed as a Research Associate at Manchester University, supported by a Teaching Company Scheme Project with Gentronix Limited; P.O.Keenan was a PhD student at UMIST, whose PhD funded by the BBSRC, also had a case award, funded by Gentronix Limited; C.V.Williams and D.J.Tweats acted as consultants to Gentronix Limited and received payment to help in the writing of this paper. R.M.Walmsley is a founder and director of Gentronix Limited.
|
Notes |
|---|
2To whom correspondence should be addressed. Tel: +44 161 200 4174; Fax: +44 161 236 0409; Email: richard.walmsley{at}umist.ac.uk
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionSupplementary materialConflict of interest statementReferences |
|---|
-
Adler,I.D. and Kliesch,U. (1990) Comparison of single and multiple treatment regimes in the mouse bone marrow micronucleus assay for hydroquinone and cyclophosphamide. Mutat. Res., 234, 115–123.[CrossRef][ISI][Medline]
Afanassiev,V., Sefton,M., Anantachaiyong,T., Barker,M.G., Walmsley,R.M. and Wölfl,S. (2000) Application of yeast cells transformed with GFP expression constructs containing the RAD54 or RNR2 promoter as a test for the genotoxic potential of chemical substances. Mutat. Res., 464, 297–308.[ISI][Medline]
Ames,B.N., Lee,F.D. and Durston,W.E. (1973) An improved bacterial test system for the detection and classification of mutagens and carcinogens. Proc. Natl Acad. Sci. USA, 70, 782–786.
Aoyama,Y., Yoshida,Y. and Sato,R. (1984) 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., 259, 1661–1666.
Arenaz,P., Hallberg,L., Mancillas,F., Gutierrez,G. and Garcia,S. (1989) Sodium azide mutagenesis in mammals: inability of mammalian cells to convert azide to a mutagenic intermediate. Mutat. Res., 227, 63–67.[CrossRef][ISI][Medline]
Averbeck,D. and Averbeck,S. (1994) Induction of the genes RAD54 and RNR2 by various DNA damaging agents in Saccharomyces cerevisiae. Mutat. Res., 315, 123–138.[ISI][Medline]
Billinton,N., Barker,M.G., Michel,C.E., Knight,A.W., Heyer,W.-D., Goddard,N.J., Fielden,P.R. and Walmsley,R.M. (1998) Development of a green fluorescent protein reporter for a yeast genotoxicity biosensor. Biosens. Bioelectron., 13, 831–838.[CrossRef][ISI][Medline]
Boulton,S.J. and Jackson,S.P. (1996a) Identification of a Saccharomyces cerevisiae Ku80 homologue: roles in DNA double-strand break rejoining and in telomeric maintenance. Nucleic Acids Res., 24, 4639–4648.
Boulton,S.J. and Jackson,S.P. (1996b) Saccharomyces cerevisiae Ku70 potentiates illegitimate DNA double-strand break repair and serves as a barrier to error-prone DNA repair pathways. EMBO J., 15, 5093–5103.[ISI][Medline]
Brennan,R.J. and Schiestl,R.H. (1998) Free radicals generated in yeast by the Salmonella test-negative carcinogens benzene, urethane, thiourea and auramine. Mutat. Res., 403, 65–73.[ISI][Medline]
Brennan,R.J., Swoboda,B.E. and Schiestl,R.H. (1994) Oxidative mutagens induce intrachromosomal recombination in yeast. Mutat. Res., 308, 159–167.[CrossRef][ISI][Medline]
Briza,P., Breitenbach,M., Ellinger,A. and Segall,J. (1990) Isolation of two developmentally regulated genes involved in spore wall maturation in Saccharomyces cerevisiae. Genes Dev., 4, 1775–1789.
Briza,P., Eckerstofer,M. and Breitenbach,M. (1994) 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, 91, 4524–4528.
Callen,D.F. and Philpot,R.M. (1977) Cytochrome P-450 and the activation of promutagens in Saccharomyces cerevisiae. Mutat. Res., 45, 309–324.[ISI][Medline]