p53 induction as a genotoxic test for twenty-five chemicals undergoing in vivo carcinogenicity testing.

In vivo carcinogenicity testing is an expensive and time-consuming process, and as a result, only a relatively small fraction of new and existing chemicals has been tested in this manner. Therefore, the development and validation of alternative approaches is desirable. We previously developed a mammalian in vitro assay for genotoxicity based on the ability of cells to increase their level of the tumor-suppressor protein p53 in response to DNA damage. Cultured cells are treated with various amounts of the test substances, and at defined times following treatment, they are harvested and lysed. The lysates are analyzed for p53 by Western blot and/or enzyme-linked immunosorbent assay analysis. An increase in cellular p53 following treatment is interpreted as evidence for DNA damage. To determine the ability of this p53-induction assay to predict carcinogenicity in rodents and to compare such results with those obtained using alternate approaches, we subjected 25 chemicals from the predictive toxicology evaluation 2 list to analysis with this method. Five substances (citral, cobalt sulfate heptahydrate, D&C Yellow No. 11, oxymetholone, and t-butylhydroquinone) tested positive in this assay, and three substances (emodin, phenolphthalein, and sodium xylenesulfonate) tested as possibly positive. Comparisons between the results obtained with this assay and those obtained with the in vivo protocol, the Salmonella assay, and the Syrian hamster embryo (SHE) cell assay indicate that the p53-induction assay is an excellent predictor of the limited number of genotoxic carcinogens in this set, and that its accuracy is roughly equivalent to or better than the Salmonella and SHE assays for the complete set of chemicals. ImagesFigure 1Figure 2Figure 3Figure 4

hbp://ehpnetl.niebs.ni.gov/ldocs/l999/107p805-812duerksen-hughes/abstracst.bnl Because society has made human health and safety an important consideration, considerable resources have been and are being expended in efforts to identify and classify human carcinogens. Unfortunately, methods currently used, such as the Ames test (1) and in vivo animal testing, suffer from several shortcomings. These include the need to extrapolate from prokaryotes to humans (Ames test) and the need to extrapolate from rodents to humans, the high cost, and the long period before results are known (in vivo animal testing). With all current methods there is also limited predictivity, both when examining the agreement of various methods with each other (2)(3)(4)(5)(6) and when applying laboratory results to actual human populations.
To aid in the development and validation of improved ways to predict the carcinogenicity of compounds, the National Institute of Environmental Health Sciences (NIEHS; Research Triangle Park, NC) initiated the predictive toxicology evaluation project (PTE). This project listed two sets of chemicals undergoing in vivo carcinogenicity testing; the first set of 44 chemicals was listed in 1990 (PTE1) and the second set of 30 chemicals in 1994 (PTE2) (7,8). Researchers were invited to submit predictions regarding the ability of these chemicals to induce carcinogenesis in rodents, and the predictions and actual in vivo results were to be compared when the animal results became available. The results from PTE1 provided information regarding the features of chemicals most predictive of their ability to act as carcinogens, and the results from PTE2 will add to this dataset once all of the in vivo results are available for comparison with the predictions.
We previously developed a novel method for assessing the genotoxicity of substances that is simple, rapid, and cost-effective. This method is based on the biologic response of mammalian cells in culture to agents that damage their DNA (9). This assay is based on the well-documented observation that the cellular level of a tumor-suppressor protein called p53 increases following DNA damage (10)(11)(12)(13)(14)(15), primarily because of an increase in the stability of the protein and a resulting decrease in its rate of proteolysis (9,10). The p53 then acts to prevent replication of damaged DNA, either by causing the cell to undergo a reversible growth arrest (11,16,17) or by initiating a cell's apoptotic pathway (18)(19)(20)(21).
The assay itself is simple (9). Cultured mammalian cells (NCTC 929) are treated with various doses of the test agent, and at specified time points following treatment, the cells are harvested and lysed. The level of p53 in the lysates is measured by p53 enzymelinked immunosorbent assay (ELISA) and/or Western blot analysis, and compared to the level in untreated cells.
Our previous work (9) indicated that both Western blot and ELISA analyses yielded similar results. Therefore, our current work focused on the more quantitative ELISA assay. We also found that the NCTC 929 cell line was a useful model system. NCTC 929 cells were initially selected because previously published results (10) indicated that their levels of p53 rise following genotoxin treatment. We verified this, and found in addition that p53 in these cells can be immunoprecipitated by a monoclonal antibody specific for wild-type p53. We also found that the level of p53 in these cells rises significantly following treatment with indirect genotoxins such as aflatoxin BI and 2-acetylaminofluorene, indicating that they possess significant amounts of the metabolic activities necessary for biotransformation (9. To determine the ability of this assay to predict carcinogenicity in rodents and to compare its performance with other proposed alternatives, we subjected 25 of the 30 PTE2 substances to analysis by this method. Five substances (citral, cobalt sulfate heptahydrate, D&C Yellow No. 11, oxymetholone, and tbutylhydroquinone) tested positive in this assay and three substances (emodin, phenolphthalein, and sodium xylenesulfonate) tested as possibly positive. Comparisons between the results obtained with this assay and those obtained with the in vivo protocol, the Salmonella assay, and the Syrian hamster embryo (SHE) cells indicate that the p53induction assay is an excellent predictor of the limited number of genotoxic carcinogens in this set, and that its accuracy is roughly equivalent to or better than the Salmonella and SHE assays for the complete set ofchemicals.

Materials and Methods
Cell culture. These experiments were done using the NCTC 929 cell line derived from mouse fibroblasts and obtained from the American Type Culture Collection (Rockville, MD). Cells were cultured in modified Eagle media (Gibco/Life Sciences, Gaithersburg, MD) supplemented with 10% fetal calf serum (Gibco) (MEM 10). Cells were plated at a density of 0.5-1.5 x 106 cells per 100 mm plate 1 day prior to treatment. pAB 122 hybridoma cells, also obtained from the American Type Culture Collection, were propagated in Dulbecco's modified Eagle medium supplemented with 10% fetal calf serum, and served as the source for the primary anti-p53 antibody used in the ELISA analysis.
Unless noted otherwise, 10 mg/mL stock solutions of the test substances were prepared in either culture media (MEM 10) or in dimethylsulfoxide (DMSO). We have previously shown that DMSO alone at the concentrations used does not induce p53 in treated cells (9).
Cell treatment. Volumes of the stock solution corresponding to the doses tested were added to a series of plates (approximately 1 x 106 cells per 100 mm plate in a volume of 10-mL MEM 10). Control plates were treated with an amount of vehicle alone (media or DMSO) that corresponded to the highest volume added to the experimental plates. For each chemical, several doses between 1 and 100 pg/mL of the test substance were evaluated; the exact doses used are listed in the sections for the separate chemicals. Treated and control cells were incubated at 37°C and 5% CO2 until harvest. One series was harvested approximately 6 hr posttreatment, and a second series was harvested approximately 17 hr posttreatment.
Each chemical was tested, at the several dosages and at the two time points, between two and four times. In some cases, specific doses were repeated in additional trials.
To ensure that the p53 response of these cells to both direct-and indirect-acting genotoxic chemicals remained intact and relatively constant during these experiments, cultures of NCTC 929 cells from passages currently in use were periodically treated with known direct (N-methyl-N'-nitronitrosoguanidine) and indirect (mitomycin C) genotoxins (9) to verify that the cells could adequately respond to both types of chemicals. We found that the cellular response to both direct-and indirect-acting chemicals was stable throughout and beyond the course of these experiments.
Cell lysis andp53 ELISA. Cells were harvested, lysed, and analyzed for their protein concentration by the BCA assay (Pierce, Rockford, IL) and for their p53 levels by ELISA, as described previously (9). Briefly, cells were removed from the plate by trypsinization, concentrated by centrifugation, and suspended in lysis buffer. After lysis, lysates were stored at -80°C for no more than 1 week prior to analysis. The ELISA analysis used pAB 122 as the primary or capture antibody, biotinylated anti-p53 (Boehringer Mannheim, Annapolis, IN) as the detection antibody, and GST-p53 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) as a standard. Each point was measured in triplicate, and the results were normalized to the amount of protein present in each sample.
Analysis. Data were analyzed using oneway analysis ofvariance software (SuperAnova, Abacus Concepts, Berkeley, CA). The Scheffe S-test was used post hoc to identify significant differences between p53 levels in treated and untreated cells. Test results were interpreted as positive if one or more doses yielded a significant increase (p < 0.001) in the level of p53 in treated versus untreated cells, in two or more separate experiments, at one or both of the times tested (6 and 17 hr), and with no more than one of three or four experiments failing to show such an increase. Results were interpreted as possibly positive using the same criteria, except that the significance of the increase was less (p < 0.05). Data regarding the Salmonella and in vivo results (where available) were obtained from the PTE2 website (8).

Results
Cobalt sulfate heptahydrate and D& C Yellow No. 11 induce p53 strongly by 6 hr. Two substances, cobalt sulfate heptahydrate and D&C Yellow No. 11, caused increases in cellular p53 by 6 hr. The increases for cobalt sulfate heptahydrate were significant at both 50 and 100 pg/mL, and those for D&C Yellow No. 11 were significant at 5, 10, and 25 pg/mL. On the basis of these results, they are classified as positive. Figure 1 depicts representative experiments for cells treated with each of these substances and harvested at 6 hr. For these two substances, the increase in the cellular level of p53 in cells treated with the statistically significant doses is robustgreater than a 4-fold increase over that seen in untreated cells.
Citral cobalt sulfate heptahydrate, D& C Yellow No. 11, oxymetholone, andt-butylhydroquinone induce p53 strongly by 17 hr. In addition to the two substances that induced p53 by 6 hr, three additional substances caused significant increases in the cellular level of p53 by 17 hr (citral, oxymetholone, and t-butylhydroquinone). The increase for citral was significant at doses of 10, 15, 20, 25, and 30 pg/mL; some cytotoxicity was noted at 25 and 30 pg/mL and increased cytotoxicity at 50 pg/mL. The increases for cobalt sulfate heptahydrate were significant at doses of 20 and 50 pg/mL; significant cytotoxicity was noted in cells treated with 100 ig/mL for 17 hr. Cells treated with 25 Volume 107, Number 10, October 1999 * Environmental Health Perspectives ig/mL D&C Yellow No. 11 for 17 hr showed a significant increase in their cellular p53 level; cells treated with a higher level displayed cytotoxicity. Oxymetholone caused significant increases in the p53 level at doses of 10 and 25 pg/mL, with increasing cytotoxicity noted at doses of 25 and 50 pg/mL. Finally, t-butylhydroquinone significantly increased cellular p53 levels at doses of 1, 5, and 10 pg/mL, with significant cytotoxicity noted at higher doses. On the basis of these results, these five substances are classified as positive. Figure 2 depicts representative experiments for cells treated with each of these substances and harvested at 17 hr. For each of the substances that tested positive, the increase in the cellular level of p53 in cells treated with the statistically significant doses is robust-greater than a 4-fold increase over that seen in untreated cells. a.~~~Doe(#L Emodin, phenolphthalein, xylenesulfonate inducep53 less s, hr. Emodin increased cellular p dose of 20 pg/mL, phenolphtha of 10 pg/mL by 17 hr, and s nesulfonate at doses of 1 and 5 p case of these three chemicals, th of the increase was less than that five chemicals previously disci they are dassified as possible po% 3 depicts representative experim treated with each of these substa vested at 17 hr. The extent of td p53 was reduced (approximately over that seen in untreated cells) to that of the previous five substa Seventeen chemicals did cellular levels ofp53. The rem; teen chemicals (1-chloro-2 cinnamaldehyde, codeine, diet I a Dose (ig/mL) and sodium 1,2-dihydro-2,2,4-trimethylquinoline, ethyltrongly by 17 benzene, ethylene glycol monobutyl ether, 53 levels at a furfuryl alcohol, gallium arsenide, isobulein at a dose tyraldehyde, methyleugenol, nitromethane, ;odium xyleprimaclone, pyridine, scopolamine hydrobroig/mL. In the mide trihydrate, sodium nitrite, and tetrahye significance drofuran) were unable to reproducibly and noted for the significantly increase cellular p53 levels at the ussed; hence, times and doses tested. Based on these sitives. Figure  results, they are classified as negative. ients for cells Substances classified on the basis oftheir nces and har-ability to increase cellular p53. Table 1 lists he increase in the substances that tested positive (p < 0.001) K 2to 3-fold or possibly positive (p < 0.05) in this assay, as compared along with the time(s) and dose(s) at which mces.
positive results were achieved. -| i1 X k induce p53 at at least one tested dose, although the increase was less significant than seen with the previous five substances.
Summary. Table 2 shows the summary data for the substances that we tested as well _ as previously reported results from the 20 50 Salmonella and SHE assays (22). It  Specific notes regarding the individual substances tested follow. Classifications (positive, possibly positive, and negative) refer to the results of this study.
1-Chloro-2-propanol. A 10-mg/mL stock of 1 -chloro-2-propanol was prepared in 10 25 DMSO, and cells were treated with doses of Cinnamaldehyde did not induce p53 at any of the doses or times tested, and is therefore classified as negative. The Salmonella results were weakly positive. Citral. A 10-mg/mL stock of citral was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 15, 20, 25, 30, and 50 pg/mL. Cells treated with citral displayed increasing cytotoxicity at the highest doses tested (25,30, and 50 mg/mL) by 6 and 17 hr. Citral induced p53 by 17 hr, and is therefore classified as positive. This contrasts with the results from the Salmonella test, which was negative.
Cobalt sulfate heptahydrate. A 10mg/mL stock of cobalt sulfate heptahydrate was prepared in H20, and cells were treated with doses of 1, 10, 20, 50, and 100 pig/mL. Cells treated with cobalt sulfate heptahydrate displayed cytotoxicity when treated with the substance at 100 pg/mL for 17 hr. Cobalt sulfate heptahydrate strongly induced p53 at both the 6 and 17 hr time points, and is therefore dassified as positive. The Salmonella test gave a weakly positive response, and the SHE assay was also positive. Animal tests have shown some evidence for carcinogenicity in male rats and dear evidence for carcinogenicity in female rats, male mice, and female mice.
Codeine. A 10-mg/mL stock of codeine was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 Pg/mL. Cells treated with codeine did not display cytotoxicity at any of the times or doses tested, and codeine did not induce p53 at any of the doses or times tested. Codeine is therefore classified as negative. This is in agreement with the negative test results seen in the Salmonela and SHE assays, and a lack of evidence for carcinogenicity seen in the animal tests. D&C Yellow No. 11. A 5-mg/mL stock of D&C Yellow No. 11 was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with D&C Yellow No. 11 displayed some cytotoxicity at 50 pg/mL by 17 hr. The actual doses received by the cells were less than those administered, as some precipitation of the substance was observed after its addition to the media. Cells took up the dye; cell pellets were yellow. This color partitioned with the membrane fraction during centrifugation of the lysate. D&C Yellow No. 11 strongly induced p53 at both 6 and 17 hr, and is therefore classified as positive. The Salmonella test gave a wealdy positive result, the SHE test results were positive, and in animal studies, male and female rats showed some evidence of carcinogenicity.
Diethanolamine. A 10-mg/mL stock of diethanolamine was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with diethanolamine did not display cytotoxicity at any of the times or doses tested, and diethanolamine did not induce p53 at any of the doses or times tested. Diethanolamine is therefore classified as negative. This is in agreement with a negative result in the Salmonella assay, and contrasts with a positive result from the SHE test. During animal testing, no evidence for carcinogenicity was ._ seen in male or female rats, whereas clear evidence was observed in male and female mice.
1 ,2-Dihydro-2,2,4-trimethylquinoline (DTQ). A 10-mg/mL stock of DTQwas prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 jig/mL. Cells treated with DTQ did not display cytotoxicity at any of the times or doses tested, and DTQ did not induce p53 at any of the doses or times tested. DTQ is therefore dassified as   negative. This is in agreement with a negative result from the Salmonella assay, and contrasts with a positive result from the SHE test. The animal results list some evidence for carcinogenicity in male rats and no evidence in female rats, male mice, and female mice.
Emodin. A 10-mg/mL stock of emodin was prepared in DMSO, and cells were treated with doses of 0.5, 1, 10, 20, and 50 pg/mL. Ethylbenzene. A 10-mg/mL stock of ethylbenzene was prepared in DMSO, and cells were treated with doses of 1, 10, 20, and 50 gg/mL. Cells treated with ethylbenzene did not display significant cytotoxicity at any of the times or doses tested, and ethylbenzene did not induce p53 at any of the doses or times tested. Ethylbenzene is therefore dassified as negative. This is in agreement with a negative result from the Salmonella assay and contrasts with a positive result from the SHE test. The animal results list dear evidence for carcinogenicity in male rats, and some evidence in female rats, male mice, and female mice.
Ethylene glycol monobutyl ether (EGMBE). A 10-mg/mL stock of EGMBE was prepared in MEM 10, and cells were treated with doses of 1, 5, 10, 25, and 50 pig/mL. Cells treated with EGMBE did not display cytotoxicity at any of the times or doses tested, and EGMBE did not induce p53 at any of the doses or times tested. EGMBE is therefore classified as negative. This is in agreement with a negative result from the Salmonella assay and contrasts with a positive result from the SHE assay. The animal results list some evidence in male and female mice, no evidence in male rats, and equivocal evidence in female rats.
Furfuryl alcohol. A 10-mglmL stock of furfuryl alcohol was prepared in MEM 10, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with furfuryl alcohol did not display cytotoxicity at any of the times or doses tested, and furfuryl alcohol did not induce p53 at any of the doses or times tested. Furfuryl alcohol is therefore classified as negative. This is in agreement with negative results from the Salmonella and the SHE assays. The animal results list equivocal evidence for carcinogenicity in female rats, some evidence in male rats and mice, and no evidence in female mice. Gallium arsenide. A 1-mg/mL stock of gallium arsenide was prepared in DMSO, and cells were treated with doses of 1, 10, 20, and 50 pg/mL. The actual dose received by the cells is lower than that listed, as some  Cells treated with gallium arsenide displayed some cytotoxicity at the highest doses tested (20 and 50 pg/mL), and gallium arsenide did not induce p53 at any of the doses or times tested. Gallium arsenide is therefore classified as negative. This is in agreement with a negative result from the Salmonella assay and contrasts with a positive result from the SHE assay.
Isobutyraldehyde. A 10-mg/mL stock of isobutyraldehyde was prepared in DMSO, and cells were treated with doses of 1, 10, 20, 50, and 100 ,glmL. Cells treated with isobutyraldehyde did not display cytotoxicity at any of the times or doses tested, and isobutyraldehyde did not induce p53 at any of the doses or times tested. Isobutyraldehyde is therefore classified as negative. The Salmonella results were incondusive, the SHE results were negative, and no evidence for carcinogenicity was observed in male or female rats or in male or female mice. Methyleugenol. A 10-mg/mL stock of methyleugenol was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with methyleugenol did not display cytotoxicity at any of the times or doses tested, and methyleugenol did not induce p53 at any of the doses or times tested. Methyleugenol is therefore classified as negative. This is in agreement with a negative result from the Salmonella assay and contrasts with a positive result from the SHE assay. In vivo test results showed clear evidence for carcinogenicity in male and female rats and male and female mice. Nitromethane. A 10-mg/mL stock of nitromethane was prepared in DMSO, and cells were treated with doses of 1, 10, 20, and 50 pg/mL. Cells treated with nitromethane did not display cytotoxicity at any of the times or doses tested, and nitromethane did not induce p53 at any of the doses or times tested. Nitromethane is therefore classified as negative. This is in agreement with a negative result from the Salmonella assay, and contrasts with a positive result from the SHE assay. The animal results list no evidence for carcinogenicity in male rats and clear evidence in female rats, male mice, and female mice.
Oxymetholone. A 1 0-mg/mL stock of oxymetholone was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with oxymetholone displayed cytotoxicity at 50 pg/mL by 6 hr and at 25 and 50 pg/mL by 17 hr. Oxymetholone strongly induced p53 by 17 hr, and is therefore classified as positive. This is in contrast to a negative result from the Salmonella assay, but in agreement with a positive result from the SHE assay.
Environmental Health Perspectives * Volume 107, Number 10, October 1999 Articles a Duerksen-Hughes et al.
Animal results showed equivocal results in male rats and clear evidence in female rats. Mice were not tested. Phenolphthalein. A 10-mglmL stock of phenolphthalein was prepared in DMSO, and cells were treated with doses of 1, 10, 20, 40, and 50 pg/mL. Cells treated with phenolphthalein did not display cytotoxicity at any of the times or doses tested, and phenolphthalein induced p53 weakly by 17 hr. Phenolphthalein is therefore classified as possibly positive. This is in contrast to a negative result from the Salmonella assay, and in agreement with a positive result from the SHE assay. The animal results list clear evidence for carcinogenicity in male rats, male mice, and female mice, and some evidence in female rats.
Primaclone. A 10-mg/mL stock of primaclone was prepared in DMSO, and cells were treated with doses of 1, 10, 20, and 50 pg/mL. Cells treated with primaclone did not display cytotoxicity at any of the times or doses tested, and primaclone did not induce p53 at the doses and times tested. Primaclone is therefore classified as negative. This is in contrast with a positive result from the Salmonella assay, but in agreement with a negative result from the SHE assay. The animal results list equivocal evidence for carcinogenicity in male rats, no evidence in female rats, and clear evidence in both male and female mice.
Pyridine. A 10-mg/mL stock of pyridine was prepared in MEM 10, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with pyridine did not display cytotoxicity at any of the times or doses tested, and pyridine did not induce p53 at any of the doses or times tested. Pyridine is therefore classified as negative. This is in agreement with negative results from the Salmonella and SHE assays. The animal results list some evidence for carcinogenicity in male rats, equivocal evidence in female rats, and clear evidence for both male and female mice.
Scopolamine hydrobromide trihydrate. A 10-mg/mL stock of scopolamine hydrobromide trihydrate was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 jig/mL. Cells treated with scopolamine hydrobromide trihydrate did not display cytotoxicity at any of the times or doses tested, and scopolamine hydrobromide trihydrate did not induce p53 at any of the doses or times tested. Scopolamine hydrobromide trihydrate is therefore classified as negative. This is in agreement with a negative result from the Salmonella assay and contrasts with a positive SHE result. The animal results list no evidence for carcinogenicity. Sodium nitrate. A 10-mg/mL stock of sodium nitrate was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 jig/mL. Cells treated with sodium nitrate did not display cytotoxicity at any of the times or doses tested, and sodium nitrate did not induce p53 at any of the times or doses tested. Sodium nitrite is therefore classified as negative. The Salmonella and SHE results were positive. Sodium xylenesulfonate. A 10-mg/mL stock of sodium xylenesulfonate was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with sodium xylenesulfonate did not display cytotoxicity at any of the times or doses tested, and sodium xylenesulfonate induced p53 weakly by 17 hr. Sodium xylenesulfonate is therefore classified as possibly positive. The Salmonella and SHE results were negative, and no evidence of carcinogenicity was observed in male or female rats or male or female mice. t-Butylhydroquinone. A 1 0-mg/mL stock of 5-butylhydroquinone was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 25, and 50 pg/mL. Cells treated with t-butylhydroquinone displayed cytotoxicity at 25 and 50 pg/mL by 17 hr, and t-butylhydroquinone induced p53 by 17 hr. t-Butylhydroquinone is therefore classified as positive. This contrasts with negative Salmonella and SHE assay results. No evidence of carcinogenicity was observed in male or female rats or male or female mice. Tetrahydrofuiran. A 10-mg/mL stock of tetrahydrofuran was prepared in DMSO, and cells were treated with doses of 1, 5, 10, 20, 25, 50, and 100 ,ug/mL. Cells treated with tetrahydrofiran did not display cytotoxicity at any of the times or doses tested, and did not induce p53 at any of the times or doses tested. Tetrahydrofuran is therefore dassified as negative. The Salmonella and SHE results were negative. In animal studies, some evidence for carcinogenicity was found in male rats, no evidence in female rats and male mice, and clear evidence in female mice.

Discussion
The purpose of this study was to determine the ability of the newly developed p53-induction assay to identify genotoxic substances and to predict carcinogenicity in rodents. To this end, we subjected 25 of the 30 PTE2 substances to this test. Five substances strongly and reproducibly increased cellular p53 levels and therefore were dassified as positive, three substances increased cellular p53 levels, but to a less significant extent and therefore were identified as possibly positive, and seventeen substances did not significantly increase cellular p53 levels and were therefore classified as negative.
Five of the PTE2 substances were not tested. As currently configured, our assay system is capable of testing solid or liquid substances that can be dissolved in culture media to the required concentrations and which will remain in solution for several hr at 37°C. Anthraquinone, molybdenum trioxide, and vanadium pentoxide were not tested because of the limited solubility of these substances in water, culture media, and DMSO.
It may be possible in the future to identify conditions that will allow us to deliver the substances to the cells at the required concentrations. Isobutene is a gas at room temperature and was not tested; it may be possible in the future to set up a delivery system that would deliver a constant concentration of the gas to cells in culture. Chloroprene was not tested because it cannot be shipped.
The level of p53 detected in untreated cells varied between experiments from a low of undetectable (0) to a high of approximately 10 ng p53/mg total cellular protein. Therefore, each experiment was analyzed separately and the results of replicate experiments compared with each other. We found good agreement between experiments; for most chemicals, all experiments demonstrated the same increase or lack of increase, and with the same or a similar statistical significance. For a few chemicals, one experiment of three or four either failed to show the same trend, or did show the same trend but at a lower level of significance.
In this study we identified chemicals that induced p53 but to a lesser extent and to a lower level of significance as possibly positive. It may be that weakly positive would be a more appropriate classification.
To distinguish between the two, it may be helpful to perform additional trials of these substances to confirm the reproducibility of the observed increases. The strength of this p53-induction assay is expected to be in identifying genotoxic compounds, of which this PTE2 group contains only a few. We therefore drew comparisons between results from the p53-induction assay and the in vivo results for both the entire set of tested compounds and separately for the genotoxic substances. Figure 4 summarizes the comparisons between the Salmonel4, p53 induction, SHE, and in vivo test results. Figure 4A compares the overall in vivo test results with each of the three in vitro protocols separately, and with a combination of predictions from the three in vitro protocols. In this analysis, chemicals were dassified as in vivo positives if some or dear evidence of carcinogenicity was obtained in either or both species. Chemicals were classified as in vitro positives if they tested either positive, possibly positive, or weakly positive. Analyses included only those chemicals where both assays have been done; or, in the case of the "any of the three" analysis, all four assays have been done. Volume 107, Number 10, October 1999 * Environmental Health Perspectives This accounts for the differences in the totals for the induded analyses. Also, for the "any of the three" analysis, a chemical is scored as in vitro positive if any of the three tests (Salmonella, p53 induction, or SHE) gave a positive result. By these criteria, the Salmonella test gave a concordance with the in vivo test results of 32%, the p53-induction assay a concordance of 40%, the "any of the three" results a concordance of 59%, and the SHE assay a concordance of 67%. Therefore, by this analysis, the SHE test yielded the highest concordance with the in vivo results, followed by the composite, then the p53 induction, and finally the Salmonella assays. These results also reflect the higher likelihood of the SHE test to produce positive results for this set of chemicals, most of which were also positive in in vivo testing.
A comparison of the result from the p53induction assay and the in vivo results for only the genotoxic compounds in this group (those that tested positive in the Salmonella assay) shows that two substances were positive in both (cobalt sulfate heptahydrate and D&C Yellow No. 11), one substance was positive in vivo but negative in the p53-induction assay (primaclone), and one substance was negative in both assays (1-chloro-2-propanol). Of this limited number of genotoxic compounds present in the PTE2 list, the p53-induction assay has a concordance of75%.
Because the primary motivation in performing these types of studies is the protection of human health, transspecies carcinogens may be of greater concern than single-species carcinogens. Figure 4B compares the in vivo test results for trans-species carcinogens with each of the three in vitro protocols separately, and with a combination of predictions from the three in vitro protocols. In this analysis, chemicals were classified as in vivo positives only if some or clear evidence of carcinogenicity was obtained in both tested species; otherwise, the analysis was as described for Figure 4A. By these criteria, the "any of the three" results gave a concordance of 35%, the Salmonella test a concordance of 50%, the SHE assay a concordance of 53%, and the p53-induction assay a concordance of 56%. Therefore, by this analysis, the p53-induction assay yielded the highest concordance with the in vivo results, followed by the SHE analysis, then the Salmonella, and finally the composite assays. Because fewer of the in vivo results are classified as positive by the transspecies criteria, the concordance of the two assays less likely to test positive for this set of chemicals (Salmonella and p53 induction) with the in vivo test results was improved, and the concordance of the two data sets more likely to test positive for this set of chemicals (SHE and "any of the three") decreased.
A comparison of the p53-induction assay results and the transspecies in vivo results for only the genotoxic compounds in this group (those that tested positive in the Salmonella assay) shows that one substance was positive in both assays (cobalt sulfate heptahydrate) and two substances were negative in both assays (1-chloro-2-propanol and primaclone). Of this limited number of genotoxic compounds present in the PTE2 list, the p53-induction assay has a concordance with the in vivo results of 100%. Figure 4C compares results from the three in vitro assays with each other. In this analysis, only those chemicals for which both listed assays have been reported are induded. The concordance of Salmonella and SHE assays is 44%, that of the p53-induction and SHE assays is 48%, and the p53-induction and Salmonella assays is 63%, indicating that of the three, the p53 induction and the Salmonella assays are the most like each other, although still significantly different. Based (22). In contrast, the p53-induction assay takes advantage of a normal cellular response to DNA damage; namely, the increase of the tumor-suppressor protein p53. It is therefore not dependent on a permanent change in DNA; rather, it works by eavesdropping on the events that normally follow DNA damage. An advantage of this assay, therefore, is that it may be capable of detecting transitory DNA damage that is repaired too quickly to be detected in tests which require heritable changes in the DNA. Diethylstilbestrol, which produces DNA lesions with a short biologic half-life, induced p53 [(9) and references therein] while giving a negative or weak response in the Salmonella assay. This feature is one possible explanation for substances (such as sodium xylenesulfate or t-butylhydroquinone) testing positive or possibly positive in this assay while testing negative in the in vivo tests; the substance may have caused DNA damage that was quickly repaired before DNA replication.
Another possible explanation for chemicals testing positive in this assay and negative Environmental Health Perspectives * Volume 107, Number 10, October 1999 t in the in vivo protocol is that they could cause increases in cellular p53 levels by a mechanism independent of DNA damage (23). For example, a substance that could block proteosome function would be expected to increase cellular p53 (24,25). We have found, however, that the ability of substances to increase p53 and to cause cellular cytotoxicity is not correlated; indeed, most substances do not induce p53 even when clearly inducing stress and cytotoxicity [(9) and this study]. Hence, it seems unlikely that p53 induction is a general nonspecific response to chemical or cytotoxic stress. The p53-induction assay used in this study utilized NCTC 929 cells, which are a mouse fibroblast cell line. Although we have shown that these cells can respond to indirect genotoxins such as aflatoxin B1 by increasing their level of p53, demonstrating the presence of activating metabolic enzymes, it is possible that differences in the array of metabolic activities in these cells as opposed to liver extract could create a different combination of metabolites. This altered range of activities could then either increase or decrease DNA damage. This could account for differences in the results noted with the p53-induction assay and those noted for other assays. It may be interesting to test a subset of these substances for their ability to induce p53 in a different cell line, perhaps one derived from liver.
We have previously shown that the increase in p53 is transient with respect to time and is maximal at certain substancespecific doses, and that a higher dose does not necessarily lead to a higher level of cellular p53. In this study, only a limited number of times and doses could be tested. Based on our previous work (9), direct-acting genotoxic chemicals induced p53 at early times (2-8 hr), whereas indirect-acting genotoxic chemicals induced p53 at later times (12-24 hr). Hence, we tested each of these 25 PTE2 chemicals at both early (6 hr) and late (17 hr) time points in an attempt to detect both direct and indirect genotoxins. We found that one chemical exerted significant effects at more doses at the early time point than at the latter (D&C Yellow No. 11), and that others showed an effect at only the latter time point (citral, oxymetholone, t-butylhydroquinone; compare Figure 2 and Figure  1). It is possible, however, that a substance capable of weakly inducing p53 could exert its maximal effect at a time other than those tested, and that any induction seen at the tested times was not statistically significant.
Previously tested genotoxins yielded maximal induction at doses between 1 and 100 pg/mL; therefore, for each of the 25 substances analyzed, we tested several doses between 0.5 and 100 gg/mL. We found that most of the chemicals that tested positive did so only at one or a few of the tested doses, suggesting that the dose range at which an effect can be observed may be rather narrow. For a number of the substances, the higher doses (25, 50, and 100 j.g/mL) induced extensive cytotoxicity, compromising the validity of the p53 assay at those dosages. It is possible that a substance testing negative in our assay would have tested positive at some other untested dose. Our results were analyzed with the Scheffe S-test, which is a relatively conservative post hoc test for differences.
The expected strength of this p53-induction assay is in the identification of genotoxic compounds. There were few of these compounds in this PTE2 group, and with these few compounds, the p53-induction assay did an excellent job of identifying them. The most likely explanation for the inability of this assay to identify many of the carcinogenic compounds in this group is that they are epigenetic substances that cause cancer by nongenotoxic mechanisms. Additionally, factors such as absorption, distribution, metabolism, and excretion, which operate in whole animals but do not operate in in vitro test systems, could play a role.
It has been suggested by Ashby and Tennant (26) that carcinogenicity may result from a specific interaction between the chemical and the tissue rather than it being an intrinsic and unique property of the chemical; therefore, it may well be that no one chemical feature or assay will provide the predictivity desired. It may therefore be helpful to add this new set of information to that considered when evaluating the carcinogenic potential of substances. For example, the results of the p53-induction assay could be considered as an input in the induction of rules for predicting chemical carcinogenicity in rodents (27).