Evaluation of in Vitro Reporter Gene Induction Assays for Use in a Rapid Prescreen for Compound Selection to Test Specificity in the Tg.AC Mouse Short-Term Carcinogenicity Assay

Karol L. Thompson*, Barry A. Rosenzweig*, Yi Tsong{dagger} and Frank D. Sistare*,1

* Division of Applied Pharmacology Research, OTR/OPS, and {dagger} Office of Biostatistics, ORM, Center for Drug Evaluation and Research, Food & Drug Administration, Laurel, Maryland 20708

Received January 19, 2000; accepted June 6, 2000


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Under ICH guidelines, short-term carcinogenicity assays such as the Tg.AC assay are allowed alternatives for one species in the 2-year rodent bioassay. The Tg.AC transgenic mouse, which carries the v-Ha-ras oncogene under control of the {zeta}-globin promoter, develops skin papillomas in response to dermal application of carcinogens and tumor promoters. The appropriate specificity of the Tg.AC model for testing pharmaceuticals has not been systematically evaluated. The selection of candidate test compounds among noncarcinogenic pharmaceuticals would be aided by a high-throughput in vitro prescreen correlative of activity in the in vivo Tg.AC assay. Here we describe the development of a prescreen based on correct response to 24 compounds tested previously in Tg.AC mice. The in vitro prescreens, chosen to reflect molecular pathways possibly involved in Tg.AC papilloma formation, consisted of a {zeta}-globin promoter-luciferase construct stably expressed in K562 cells (Zeta-Luc) and three of the stress-response element–chloramphenicol acetyltransferase (CAT) fusion constructs stably expressed in HepG2 cells that are part of the CAT-Tox (L)iver assay. The stress response elements chosen were the c-fos promoter, the gadd153 promoter, and p53 response element repeats. Of the four assays, the gadd153-CAT assay showed the strongest concordance with activity in the Tg.AC assay, correctly classifying 78% of Tg.AC positive and 83% of Tg.AC negative compounds. The correlation was further improved by adding the Zeta-Luc assay as a second-stage screen. These cell-based assays will be used in a novel approach to selecting candidate compounds that challenge the specificity of the Tg.AC assay toward pharmaceuticals.

Key Words: stress response genes; Tg.AC; in vitro assay; gadd153; c-fos; zeta-globin; p53; K562; HepG2.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
To evaluate the carcinogenic potential of certain pharmaceutical agents, U.S. Food and Drug Administration (FDA) regulations have required dosing of mice and rats for 2 years and a full histopathological evaluation to search for evidence of neoplasia. The FDA's Center for Drug Evaluation and Research has recently agreed through the International Conference on Harmonization (ICH) to allow the substitution of one species in the standard 2-year cancer bioassay with a short- or medium-term alternative model (ICH expert safety working group, 1996Go). The Tg.AC transgenic mouse model, which carries multiple copies of a transgene consisting of a {zeta}-globin promoter fused to a v-Ha-ras structural gene (Leder et al., 1990Go) and develops skin papillomas upon repeated dermal application of carcinogens or tumor promoters, has emerged as one of the more promising alternative models because it is capable of detecting both genotoxic and nongenotoxic carcinogens in less than 6 months (Tennant et al., 1996Go). The chemicals used in the initial validation of the Tg.AC model were predominantly industrial chemicals or environmental contaminants that had been tested in the 2-year National Toxicology Program (NTP) bioassay. A consortium of industry, academia, and government agencies organized by the International Life Sciences Institute (ILSI) has been formed to further scrutinize a number of alternative in vivo model systems, including Tg.AC, with particular emphasis on their response to diverse pharmaceuticals (Robinson, 1998Go). The compounds under evaluation by the ILSI consortium include genotoxic human carcinogens (cyclophosphamide, melphalan, phenacetin), an immunosuppressive human carcinogen (cyclosporin A), hormones, and rodent carcinogens that are putative human noncarcinogens (e.g., phenobarbital, methapyrilene, clofibrate). The noncarcinogens chosen by the ILSI consortium to test appropriate specificity of the Tg.AC assay (D-mannitol and sulfisoxazole) are more limited in mechanistic spectra.

Transgenic mice assays such as the Tg.AC skin paint assay, which predict carcinogenicity but work through incompletely understood mechanisms, present new challenges for regulatory agencies. Both sensitivity (the appropriate positive tumorigenic response to known carcinogens) and specificity (the appropriate lack of response to properly dosed noncarcinogens) must be demonstrated to fully understand and appreciate the predictive value of the Tg.AC model. The specificity of the Tg.AC assay towards pharmaceuticals might best be challenged by the selection and testing of potential false positives in the assay. There are a number of mechanisms through which false-positive responses could conceivably be generated in the Tg.AC assay, such as through effects on hair growth. Pharmaceuticals that are noncarcinogenic but induce gene expression patterns that correlate with activity in the Tg.AC assay may represent one type of false positive. In this paper, we focused on identifying common signal transduction pathways that may be linked to Tg.AC transgene activation through correlative analysis of Tg.AC-tested compounds in in vitro gene reporter assays. Reporter gene assays have been used as high throughput assays to detect early changes in gene expression linked to molecular toxicity (Beard et al., 1996Go; Todd et al., 1995Go). The assumption of an in vitro screening assay is that activation of a set of common signal transduction and response pathways is not limited to one cell type or species and thus could be replicated in vitro. It is not expected that all mechanisms involved in the induction of papillomas will be covered in the in vitro screens. The correlative in vitro assays, once identified, will be used to screen 100 pharmaceuticals that have tested negative for carcinogenicity in male and female rats and mice ("four cell negatives") and have been identified and catalogued in FDA databases (Contrera et al., 1997Go). For high throughput screening of pharmaceuticals, it was preferable to choose a prescreen that did not require or include exogenous metabolic activation, even though some drugs may not be correctly detected as active in its absence.

In selecting which genes to test as potential surrogate indicators of activity in the Tg.AC assay, several molecular mechanisms of papilloma formation were considered. In the transgene, v-Ha-ras expression is under control of a {zeta}-globin promoter and is linked to papilloma formation in Tg.AC mice (Hansen et al., 1995Go). One assay was constructed to measure activation of the {zeta}-globin promoter in a cell line that is permissive for expression of erythroid-specific genes. The in vivo activation of the {zeta}-globin promoter by papillomagenic compounds is a more complex process that is still not fully understood but is thought to require palindromic orientation of {zeta}-globin promoter sequence for inducible transgene expression (Thompson et al., 1998Go; Thompson et al., 1999Go). Additionally, although the transgene is present in all cells in the Tg.AC mouse, measurable transgene expression is mostly limited to cells of hematopoietic origin and to skin papillomas (Leder et al., 1990Go). Establishment and maintenance of an in vitro cell system containing palindromically oriented {zeta}-globin promoter sequence would be technically very difficult because of the instability of palindromic DNA in vitro. Additionally, the role that the palindromic promoter region plays in the process of papilloma formation is not understood at this time. It may play more of a passive role, such as making the transgene locus permissive for activation, rather than as a direct target for the agents that induce papillomas. It is also not clear whether activation of the transgene locus is a direct effect of chemical treatment or occurs as the target cell population is induced to expand, such as after wounding.

The responsiveness of the Tg.AC assay to genotoxicants and the known inductive effect of genotoxicants on subsets of stress-response genes (Beard et al., 1996Go) was the basis for inclusion of candidate prescreens derived from components of the commercially available CAT-Tox assay, which is a battery of stress gene promoters and response elements coupled to the chloramphenicol acetyltransferase gene (CAT) and stably expressed in transformed human cell lines (Todd et al., 1995Go). To select which target genes might best detect compounds active in the Tg.AC assay, we examined the induction profile of compounds, identical or similar to those tested in Tg.AC mice, that were available in the Xenometrix database of results in the CAT-Tox (L) or CAT-Tox (D) assays. The best correlations were observed with the human c-fos promoter, the hamster gadd153 promoter, and repeats of p53 response elements. These three CAT-Tox (L) assays along with the {zeta}-globin promoter–luciferase construct were used to screen a set of 24 chemicals that had been previously tested in the Tg.AC mouse skin paint assay to try to identify a high throughput in vitro assay that is correlative with induction of papillomas in vivo.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Chemicals.
2,4-Diaminotoluene, tripropylene glycol diacrylate, ethyl acrylate, 2-butanone peroxide, 2-chloroethanol, and 2,6-diaminotoluene were purchased from Aldrich (Milwaukee, WI). Diethanolamine, 17-ß-estradiol, diethylstilbestrol, phorbol 12-myristate 13-acetate (PMA), benzoyl peroxide, p-anisidine, 8-hydroxyquinoline, acetone, benzethonium chloride, phenol, and dimethyl sulfoxide (DMSO) were purchased from Sigma (St. Louis, MO). Calbiochem (San Diego, CA) was the source of mezerein. Lauric acid diethanolamine, mirex, and o-benzyl-p-chlorophenol were obtained from the NTP chemical depository. Concentrated stock solutions of the chemicals in either ethanol, DMSO, or acetone were prepared on the day of dosing, except for PMA and mezerein, which were stored in aliquots at –20°C. Chemical solubility was determined by visual inspection in stock solutions and by microscopic examination of the wells of tissue culture plates after 48-h incubation of the chemicals with cells in culture media at 37°C.

DNA constructs.
A pGEM3-based plasmid containing the Tg.AC transgene (Zeta-ras) (Leder et al., 1990Go) was obtained from Dr. Philip Leder (Harvard Medical School) through a material transfer agreement. The {zeta}-globin promoter was isolated by digesting Zeta-ras with EcoRI, filling in the overhanging ends with DNA polymerase I Klenow fragment, and subsequently digesting with BamHI. The resulting 0.9 kb {zeta}-globin promoter fragment was gel purified and ligated into the multiple cloning region of a SmaI-BglII double-digested pGL3 vector (a promoterless luciferase reporter plasmid (Promega, Madison, WI)) to create the plasmid Zeta-Luc. The herpes simplex virus thymidine kinase (TK) promoter containing nucleotides –105 to +50 bp relative to the transcriptional start site was generated by PCR amplification and inserted into the pGL3 vector to create the plasmid TK-Luc. A PCR-based site-directed mutagenesis procedure (Stratagene, La Jolla, CA) was used to mutate a putative AP-1 site in the 3' coding region of the ampicillin resistance gene in the pGL3 plasmid, in which conservative substitutions were made in two amino acid codons. One inadvertent mutation resulted during the procedure, eliminating the last amino acid of the ß-lactamase gene, but had no effect on conference of ampicillin resistance. The identity of all constructs were confirmed by fluorescence-based DNA sequencing.

Transient transfection protocol.
K562 cells were maintained at densities below 106/ml in RPMI 1640 containing L-glutamine, 10% fetal calf serum (FCS), and antibiotics in a 5% CO2/95% air incubator. CV-1 cells were grown in Dulbecco's Minimal Essential Medium (DMEM) with 10% FCS and antibiotics. HepG2 cells were maintained in MEM supplemented with 2 mM glutamine, 1 mM pyruvate, 10% FCS, and antibiotics. The liposome reagent DOTAP (Roche Diagnostics, Indianapolis, IN) was used to transfect K562 cells. HepG2 and CV-1 cells were transfected using FuGene 6 (Roche Diagnostics). Luciferase and pSV-ß-galactosidase constructs were cotransfected at a 3:2 ratio. Inducers were added to cells on the day following transfection and 24 h later the cells were harvested for reporter gene measurements. K562 cells were washed twice with calcium-, magnesium-free phosphate-buffered saline (CMF-PBS) and resuspended in lysis buffer (0.1 M potassium phosphate pH 7.8, 1 mM DTT). The cell suspensions were frozen in a dry ice-ethanol bath and lysed by freeze-thawing three times. Adherent cells (CV-1 and HepG2) were washed thrice with CMF-PBS and lysed in a buffer containing 0.1 M potassium phosphate pH 7.8, 1% Triton X-100, 2 mM EDTA, 1 mM DTT for 15 min while rocking on ice. Cell debris was cleared by centrifugation and the supernatants were analyzed for luciferase and ß-galactosidase activity. Luciferase was measured in an Analytical Luminescence Laboratory luminometer (San Diego, CA) using flash kinetic conditions. ß-galactosidase activity was measured in a spectrophotometric plate reader using chlorophenolred-ß-D-galactopyranoside (Roche Diagnostics) as a substrate. Equivalent units of ß-galactosidase in the lysates was calculated from a standard curve of dilutions of E. coli ß-galactosidase.

Screening assay with stably transfected K562 cells.
K562 cells were cotransfected with Zeta-Luc and pcDNA3 (a plasmid containing the neomycin resistance gene) at a 2:1 ratio and selected for stable expression with 500 µg/ml GENETICIN®. Clones of GENETICIN®-resistant cells, isolated by limiting dilution, were screened for PMA-inducible luciferase activity. Clone ZL-9E was selected for compound-screening studies based on basal and induced luciferase activities that were similar to levels measured in both transiently transfected and stably transfected uncloned K562 cells.

K562 clone ZL-9E cells were plated in 24-well plates at a density of 0.5 x 105 cells per ml in media without selective agent, and test chemicals were added at 2x concentrations in equal volumes of media. Appropriate dose ranges were determined by cell viability measurements (by MTS dye reduction) and compound solubility in culture media. Cells were harvested 48 h after dosing for luciferase measurements, as described above. Treatments with the positive control compound (PMA at 0.5 and 5 nM) were included in each experiment. The protein concentration in the lysates was measured with Coomassie Plus Protein Assay Reagent (Pierce, Rockland, IL) using bovine serum albumin as a standard. All compounds were tested in at least three independent experiments.

CAT-Tox (L) assay.
Each CAT-Tox (L) Assay was supplied by Xenometrix (Boulder, CO) as a customized kit for this study. Each kit is comprised of 3 stably transfected human HepG2 hepatoma cell lines with the capability of assaying four compounds in triplicate at five doses including a vehicle control and a positive control. The three cell lines were gadd153-CAT, fos-CAT, and p53RE-CAT. Just prior to each experiment, dose range–finding studies were performed on parental HepG2 hepatoma cells supplied separately by Xenometrix. The parental HepG2 hepatoma cell line was also included in each kit for a cellular viability assessment. Cells were plated by Xenometrix at a density of 55,000 cells per well in a 96-well plate according to our specified format, and shipped overnight. Upon receipt of the kit, the cell media was carefully aspirated and replaced with 100 µl of fresh media. The cells were then incubated overnight at 37° C in a 5% CO2/95% air incubator until the next morning when dosing was performed. Dosing solutions of each compound, at 2x concentrations, were prepared by diluting concentrated stock solutions, freshly prepared in the appropriate solvent, into fresh media; 100-µl aliquots of each solution were added to wells. Control wells received 100 µl of media alone or media containing the maximum concentration of solvent used for samples on the same 98-well plate (up to 4% at 2x concentrations). The plates were incubated for 48 h in a humidified 5% CO2/95% air incubator. After compound exposure, all the cells, excluding the parental HepG2 cells, were washed twice with 200 µl of Dulbecco's phosphate-buffered saline (D-PBS), and lysed with 100 µl of 1X lysis buffer for 30 min at room temperature. Duplicate wells that received the same chemical exposure were combined to maximize the sensitivity of the CAT ELISA assay. A 10-µl aliquot of the lysate was transferred to a fresh set of microtiter plates and assayed for total protein using Bradford Protein Dye supplied in the kit. After 30 min of incubation at room temperature, the absorbance at 600 nm was determined using a microtiter plate reader. The remaining 190 µl of lysate was then transferred to another set of microtiter strips coated with anti-CAT antibodies, and a CAT ELISA was performed according to the supplied protocol. After 45 min of incubation at room temperature, the absorbance at 405 nm was read using a microtiter plate reader. The cell viability assay was performed using the unlysed parental HepG2 cells. Fifty microliters MTT at 5 mg/ml in DMSO was added to each well and incubated at 37°C for 30 min. The wells were then aspirated and the formazan salt was solubilized in 200 µl DMSO. The absorbance at 550 nm was read using a microtiter plate reader.

Ratio of response measurement.
In the CAT-Tox (L) assays, CAT protein and total protein were measured at each dose level tested. CATk and TOTk denote the CAT protein and total protein measurements at the kth dose level, where k=0 is the measurement at baseline. The expression level of each promoter at the kth dose level is then represented by the ratio of response (CATk /TOTk). The adjusted ratio of response, (CATk /TOTk) / (CAT0 /TOT0), corrects for baseline levels of CAT expression and is a measurement of the induction of CAT protein at the kth dose level for a given compound. Similar calculations were performed for the Zeta-Luc assay using light units in place of CAT protein measurements.

Statistical analysis.
The objective of the statistical analysis is to use a logistic regression function to estimate the probability of activity in the Tg.AC assay of each compound based on the in vitro results. The discriminant value of the probability is then determined by the logistic probability value at which the Tg.AC results and the in vitro results have the maximum number of matches. To calculate the predictivity of each reporter gene assay for Tg.AC papillomagenicity, the largest ratio of response (limited to doses where TOTk > 40% of TOT0) was calculated for each compound from three separate data sets and linked using the following univariate logistic model:

where {theta}i is the probability of Tg.AC papillomagenicity of the ith compound, CP3i is the average ratio of response of the ith compound, and {alpha} and ß are parameters determined by the data for each reporter gene assay. As a result, the four estimated logistic probability functions are




From the data set with the 24 Tg.AC-tested compounds, the ratio of response that is discriminant for correct classification of Tg.AC-positive compounds was determined for each of the four assays.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Cell Type-Specific Induction of the {zeta}-globin Promoter
The expression profile of v-Ha-ras in the Tg.AC mouse shows apparent regulation by both the sequence-specific transcription factor binding sites in the {zeta}-globin promoter and by the palindromic orientation of {zeta}-globin promoter sequences in the transgene (Leder et al., 1990Go; Thompson et al., 1999Go). The tissue- and development-specific expression pattern of v-Ha-ras expression in Tg.AC mice reflects the expression pattern associated with the embryonic form of the {alpha}-globin gene family. In untreated Tg.AC mice, measurable transgene expression is detectable by ribonuclease protection assay only in embryonic erythroid cells, fetal liver, placenta, and, in low amounts, adult bone marrow (Leder et al., 1990Go). Transgene expression is undetectable in untreated skin, but it is highly expressed in the papillomas that arise after treatment with carcinogens and tumor promoters, UV light, or after wounding. These papillomas are thought to originate from expansion of a follicular stem cell population that is permissive for {zeta}-globin promoter expression (Hansen and Tennant, 1994Go). The technical difficulty of cloning and maintaining a palindromic orientation of {zeta}-globin promoter sequence in vitro precludes our examination of the effect of drug treatment on this aspect of regulation of transgene expression in vitro. In addition, the apparent progenitor cell of a Tg.AC papilloma (normal follicular stem cells) are not available in the quantities needed for an in vitro assay intended to screen hundreds of pharmaceuticals. We designed an in vitro assay that could measure activation of {zeta}-globin promoter activity in an immortalized cell line permissive for erythroid gene expression. For this assay, the 0.9 kb {zeta}-globin promoter from the Tg.AC transgene was transferred into a luciferase reporter plasmid to create the plasmid Zeta-Luc. We first assayed whether this promoter would display appropriate cell lineage–specific regulation. We compared the activity of Zeta-Luc relative to a control luciferase construct containing the herpes simplex virus thymidine kinase promoter (TK-Luc) in immortalized cell lines derived from three different tissues. These lines were K562 (a human erythroleukemia cell line that expresses inducible embryonic and fetal globin genes (Charnay and Maniatis, 1983Go; Dean et al., 1983Go; Rutherford et al., 1981Go), CV-1 (a green monkey kidney cell line), and HepG2 (a human hepatoma cell line). The cells were transiently transfected with either Zeta-Luc or TK-Luc along with a plasmid containing the ß-galactosidase gene to normalize for transfection efficiency. A subset of the transfected cells was treated with phorbol 12-myristate 13-acetate (PMA), which is a potent inducer of papillomas in Tg.AC mice (Spalding et al., 1993Go) and induces megakaryocytic differentiation of K562 cells (Tabilio et al., 1983Go).

The activity of the {zeta}-globin promoter, relative to the activity of the control TK promoter, differed markedly between the K562 cells and the nonerythroid lines (Fig. 1Go). In K562 cells, the basal activities of the {zeta}-globin and TK promoters were similar, but the activity of the {zeta}-globin promoter was highly inducible by PMA. A two-day exposure to 1 nM PMA (a dose suboptimal for induction of megakaryocytic differentiation) increased Zeta-Luc activity 20- to 60-fold in K562 cells. Increased megakaryocytic differentiation of K562 cells at higher doses of PMA was associated with decreased expression of {zeta}-globin promoter–directed luciferase activity. A smaller induction (< 10-fold) of TK-Luc activity was also observed. In contrast, in the nonerythroid lines, the basal activity of the {zeta}-globin promoter was only a fraction (~5%) of the TK promoter basal activity. Neither promoter was inducible by 0.5–10 nM PMA in CV-1 or HepG2 cells. The Zeta-Luc construct was also uninducible by PMA when transfected into a cell line derived from a Tg.AC skin papilloma (data not shown). In summary, the {zeta}-globin promoter from the Tg.AC transgene was inducible by PMA in K562 cells, but not in cells of nonerythroid origin. Additionally, the mouse {zeta}-globin promoter, after transfection into K562 cells, was responsive to agents [e.g., hydroxyurea, phenylacetate, arabinosylcytosine (araC); unpublished observations] that are reported to induce endogenous fetal globin expression in K562 cells (Jeannesson et al., 1984Go; Samid et al., 1992Go; Xu and Zimmer, 1998Go). This result indicated that trans-acting factors present in human erythroid cells could activate a mouse globin promoter. A clonal population of K562 cells stably transfected with the Zeta-Luc construct (clone ZL-9E) was generated for the in vitro prescreen of Tg.AC-tested compounds.



View larger version (20K):
[in this window]
[in a new window]
 
FIG. 1. Cell-type specific induction of the {zeta}-globin promoter. CV-1, HepG2, and K562 cells were transiently transfected with luciferase reporter plasmids containing either 150 bp of the thymidine kinase (tk) promoter or 0.9 kbp of the {zeta}-globin promoter ({zeta}) along with a control plasmid containing 419 bp of SV40 early promoter and enhancer sequences fused to the ß-galactosidase gene. After a 48-h incubation with or without 1 nM PMA, the cells were harvested and analyzed for luciferase and ß-galactosidase activities (n = 3). Note that the y axis is discontinuous.

 
Selection of Stress-Induced Gene Promoter Assays
In vitro toxicity assays that monitor stress-induced transcriptional changes can provide valuable information on the toxic potential of a compound and the molecular mechanisms involved in its toxic effect (Beard et al., 1996Go; Todd et al., 1995Go). The CAT-Tox (L)iver system, which utilizes HepG2 human hepatoma cells, was well characterized and available for analysis of multiple endpoints in a high throughput format. The utility of the Tg.AC assay is a result of its ability to develop skin papillomas in response to compounds that are carcinogenic upon oral or other systemic routes of administration and have a range of target organ specificities. The unique transgenic composition of the Tg.AC mouse allows skin to serve as a surrogate target for more general inferences regarding compound tumorigenicity. In vitro screens that assay the induction of different gene promoters by Tg.AC-tested compounds do not necessarily need to be epithelial in origin, as many Tg.AC carcinogens are not skin specific in standard carcinogenicity assays.

To select among the set of 21 CAT-Tox promoter assays for potential correlation with activity in the Tg.AC skin paint assay, we reviewed a database from Xenometrix containing the responses of each promoter assay to compounds identical or similar to those previously tested in Tg.AC mice, e.g., dimethylbenzanthracene, hydrogen peroxide, PMA, reserpine, phenol, and acetone. The best correlation was found with three stress gene promoter-CAT fusion gene assays that contain the gadd153 promoter, the c-fos promoter, and p53 response element repeats upstream of the thymidine kinase promoter.

Assaying Reporter Gene Inducibility by Tg.AC-Tested Compounds
Twenty-four compounds that had already been tested in the Tg.AC skin paint assay were selected to be screened in the in vitro reporter gene assays. Due to the short duration of the Tg.AC assay (i.e., 26 weeks), the database of compounds tested in this assay has quickly expanded. However, this study was initiated at a time when only a limited number of drugs had been tested in the Tg.AC assay and fully reported on. With this limitation, we tried to select compounds that represented a range of activities, including carcinogens, tumor promoters, and compounds of unknown carcinogenicity. Twelve of the compounds selected were positive in the Tg.AC assay and the remaining dozen did not induce papillomas in Tg.AC mice. Most of the compounds we selected for testing were not genotoxic, as defined by positive activity in the Ames Salmonella mutagenicity assay, because the majority of four-cell negative drugs that the in vitro assays are being designed to screen will also not be Ames test positive. Overall, for this group of 24 compounds, activity in the Tg.AC assay was highly correlative with carcinogenic or tumor-promoting activity, regardless of genotoxicity, as has been previously observed (Tennant et al., 1996Go). Among the 24 chemicals assayed are two false negatives in the Tg.AC assay. These compounds are ethyl acrylate, a forestomach carcinogen in mice and rats when administered by gavage (NTP, 1986), and diethanolamine, a mouse liver and kidney carcinogen (NTP, 1998).

Dose selection for in vitro testing was based on the 48-h viability dose-response curve determined for each compound in K562 and HepG2 cells or based on compound solubility limitations in cell culture media. In each experiment, 3 to 5 doses of each drug were tested and, where solubility was not limiting, the highest dose was set as the dose that decreased cell viability by 60%. The percent solvent (acetone, DMSO, or ethanol) in the dosing solutions was limited to 2% or lower, which was a level that by itself did not induce reporter gene activity (see Table 1Go). For most of the compounds, HepG2 cells were able to tolerate higher doses than K562 cells. This disparity could be due to the higher density of the HepG2 cells at the time of treatment, which was preset by Xenometrix. The K562 clone ZL-9E line were kept at cell densities permissive for logarithmic growth throughout the 3 days of the assay because higher cell densities were associated with increased basal level of {zeta}-globin promoter-driven luciferase activity. If toxic concentrations could not be reached due to solubility limitations and no induction in reporter gene activity was observed, the compound was considered not to have been conclusively evaluated and was excluded from the calculations of assay concordance. Two compounds, mirex and benzoyl peroxide, were excluded because of solubility limitations. Mirex did not induce toxicity or promoter-driven reporter gene activity at the maximal soluble dose in any of the four reporter gene assays. The maximal dose of benzoyl peroxide tested, which was limited by solubility in tissue culture media, was toxic to K562 cells but not to HepG2 cells. Therefore, no conclusive in vitro results were obtained with benzoyl peroxide in the three CAT-Tox (L)-derived assays.


View this table:
[in this window]
[in a new window]
 
TABLE 1 Maximum Inductions Observed in the Zeta-Luc, gadd153-CAT, fos-CAT, and p53RE-CAT Assays with 24 Compounds Previously Tested in the Tg.AC Assay
 
The maximal response ratios observed in the four reporter gene assays for each of the 24 compounds and the dose at which the maximal induction was seen, as determined from triplicate determinations, are listed in Table 1Go. For the CAT-Tox and Zeta-Luc assays, the maximum induction is the response observed at the dose that did not decrease the concentration of protein in lysates made from treated cells by more than 60%. In Figure 2Go, the dose-response curves for the nine compounds that induced relative CAT protein expression maximally by 2-fold or greater in the gadd153-CAT assay are shown. The values for CAT protein, total protein, and the ratio of the two (CAT/total) are graphed for each dose relative to control values. For these compounds, CAT protein levels either show an increasing trend or no significant change compared to control levels over the dose range that ends with 60% toxicity. Figure 3Go shows the dose-response curves for two compounds (diethanolamine and phenol) that were toxic to HepG2 cells. Neither induced an increase in unnormalized CAT protein levels or a 2-fold or greater increase in relative CAT protein levels in the gadd153-CAT assay at doses that did not decrease total protein more than 60%.



View larger version (25K):
[in this window]
[in a new window]
 
FIG. 2. Dose-response curves for nine Tg.AC-tested compounds in the gadd153-CAT assay. HepG2 cells stably transfected with a gadd153-CAT construct, supplied by Xenometrix, were treated with a range of doses of each compound for 48 h. CAT (open squares) and total protein (closed diamonds) were determined as outlined in Materials and Methods. The values for CAT/total protein are represented by closed boxes. All nine compounds elevated CAT/total protein levels by two-fold or greater, at doses that did not decrease total protein more than 60%.

 


View larger version (9K):
[in this window]
[in a new window]
 
FIG. 3. Dose-response curves for two Tg.AC-tested compounds in the gadd153-CAT assay. HepG2 cells stably transfected with a gadd153-CAT construct, supplied by Xenometrix, were treated with a range of doses of each compound for 48 h. CAT (open squares) and total protein (closed diamonds) were determined as outlined in Materials and Methods. The values for CAT/total protein are represented by closed boxes. Neither compound increased CAT/total protein levels by two-fold or greater, at doses that did not decrease total protein more than 60%.

 
To determine how correlative each of the four reporter gene assays were of papillomagenic activity in the Tg.AC assay, a univariate logistic model was used to link the Tg.AC activity of each compound with the maximum response ratios observed in the in vitro assays. Using this model, the optimal ratio of response cutoff point that best correlated with positive Tg.AC activity was calculated for each assay. From this analysis, the lowest maximum response ratios that designate a positive test result for each assay are 2.0 for gadd153-CAT, 2.23 for fos-CAT, 1.85 for Zeta-Luc, and 1.10 for p53RE-CAT. Three of the reporter gene assays have calculated ratio of response cutoff values close to or equal to 2, which is the minimum fold induction level that is often designated a positive response in in vitro tests. The ratio of response cutoff value of 1.10 calculated for the p53RE-CAT assay indicates that this assay is a poor discriminator of activity in the Tg.AC assay. Using a ratio of response of 2-fold or greater as the discrimination value, each compound in the Zeta-Luc, gadd153-CAT, and fos-CAT assays was assigned either a positive or negative result. The induction results that were positive by this criteria are marked with an asterisk in Table 1Go. To highlight the comparatively low response of the p53RE-CAT assay to Tg.AC papillomagenic compounds, the p53RE-CAT inductions that were 2-fold or greater are also marked with an asterisk in Table 1Go. Using the positive or negative test result assignments, several parameters for evaluating the association between Tg.AC activity and activity in each in vitro assay were calculated (Table 2Go). Overall concordance measures the number of Tg.AC-positive compounds that are also positive in vitro plus the number of Tg.AC negative compounds that are also negative in vitro as a percent of the total number of compounds tested. Sensitivity measures the percent of Tg.AC positives that are also positive in vitro. Specificity denotes the percent of Tg.AC negatives that are also negative in vitro. Positive predictivity is a measurement of the percent of compounds positive in vitro that are also Tg.AC-positive (or true positives). Negative predictivity measures the percent of compounds negative in vitro that are also Tg.AC negative (or true negatives).


View this table:
[in this window]
[in a new window]
 
TABLE 2 Operational Characteristics of In Vitro Reporter Gene Induction Assays
 
The highest overall concordance (81%), highest sensitivity (78%), and highest specificity (83%) were observed using the gadd153-CAT assay (Table 2Go). This assay also had the highest positive and negative predictivity values. The only two Tg.AC papillomagenic compounds that were not positive in the gadd153-CAT assay were LADA and mezerein.

The fos-CAT assay had the second highest predictivity for Tg.AC activity, with an overall concordance of 71%. The Zeta-Luc assay was the least sensitive and specific of the three assays. Of the five false positives in the Zeta-Luc assay, three were genotoxic noncarcinogens and one was the mouse carcinogen diethanolamine. The protein kinase C (PKC) activators PMA (the positive control compound for the Tg.AC mouse assay) and mezerein were strong inducers of both Zeta-Luc and fos-CAT activity.

A multivariate screening procedure using a hierarchical binary tree approach was performed next to see if the correlation with Tg.AC activity could be improved by using the in vitro test results in combination. The strategy for this analysis was to use the best single screener (gadd153-CAT) in the first-stage screening and to rescreen the false negatives with one of the other three in vitro assays. The most optimal screening factor in the second stage was Zeta-Luc, with a ratio of response > 8. With this second screen, mezerein is now detected as a Tg.AC positive. This multivariate approach correctly detects 89% of Tg.AC-positive compounds and has a correct classification rate of 86%. No other combination of screens was found to improve the correlation between in vitro and in vivo activities.


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Greater acceptance and application by the scientific community of alternative methodologies involving genetically altered mice for evaluating the carcinogenic potential of pharmaceuticals requires an understanding of the limits of each assay's sensitivity and specificity. Selection of which positive and negative control compounds can best test the Tg.AC assay's strengths and limitations is a critical part of the evaluation process. Noncarcinogenic pharmaceuticals that, due to their pharmacologic mechanism of action, have the potential to produce a false-positive response in the Tg.AC assay should be identified and tested in vivo to add supportive evidence of appropriate specificity for the Tg.AC assay.

Although the mechanism of papillomagenesis in Tg.AC mice is not completely understood, it is associated with focal overexpression of the transgenic v-Ha-ras gene in follicular epidermal cells (Hansen and Tennant, 1994Go). The expression of v-Ha-ras in the Tg.AC mouse is under apparent control of palindromically oriented {zeta}-globin promoter sequence (Thompson et al., 1998Go; Thompson et al., 1999Go), with possible contribution from cis elements that are specific to the site of transgene integration. The diverse stimuli that induce papillomas in Tg.AC mice (wounding, UV light, genotoxic and nongenotoxic carcinogens, tumor promoters) probably converge on a limited number of signal transduction pathways that are directly involved in transgene activation, as well as others that are only correlative. These remain unidentified to date. Although previous studies suggest that TGF{alpha} is involved in mouse skin carcinogenesis, a recent study with Tg.AC transgenic mice null for the TGF{alpha} gene indicates that the TGF{alpha} growth factor signaling pathway is not required for papilloma induction or progression to malignancy in this system (Humble et al., 1998Go). Because the mechanism of transgene activation in follicular stem cells is presently not understood, it is important to increase our knowledge base about the specificity of the response of the Tg.AC assay to pharmacologically active, noncarcinogenic drugs.

In this paper, we screened four in vitro assays for correlation with Tg.AC responsiveness that were chosen to reflect possible biochemical pathways involved in Tg.AC papilloma formation. One assay measured induction of the {zeta}-globin promoter and the other three assays (gadd153-CAT, fos-CAT, and p53RE-CAT) contain transcriptional elements that respond to DNA damage, oxidative stress, mitogenic stimulation, or other cellular events characteristic of exposure to carcinogens or tumor promoters. Twenty-four compounds that had previously been tested in the Tg.AC assay and represent different classes of carcinogens, tumor promoters, and noncarcinogens were tested in vitro to assess the correlation between activity in the four assays and in vivo activity. At the time the compound selection was completed, only a limited number of compounds had been assessed in the Tg.AC assay. Currently, due in part to the short duration of the assay and the increased interest in its range of responsiveness, the database of Tg.AC tested compounds is more extensive.

Of the four assays, the gadd153-CAT assay had the highest overall correlation rate (81%) with Tg.AC responsiveness (Table 2Go). The gadd153-CAT assay also had high sensitivity (78%) and specificity (83%). Gadd153 is a widely expressed mammalian transcription factor that is strongly induced by genotoxic agents that damage DNA through a variety of mechanisms, including addition of bulky adducts, DNA-crosslinking, alkylation, intercalation, and topoisomerase inhibition (Beard et al., 1996Go; Luethy and Holbrook, 1992Go). Other stress pathways that induce gadd153 expression include oxidative stress, media depletion, hypoxia, and impairment of endoplasmic reticulum function (Price and Calderwood, 1992Go; Wang et al., 1996Go). Gadd153 is not reported to be significantly inducible by the PKC pathway (Luethy and Holbrook, 1992Go), which agrees with our failure to observe inducibility with mezerein. Although mezerein is a potent activator of PKC and an effective second-stage promoter, it is a weak complete tumor promoter (Slaga et al., 1980Go). The PMA induction of gadd153-CAT, which we observed only at high doses, may result from PMA-induced oxygen radical formation rather than PKC activation. The inductive response of the gadd153 promoter to cellular events characteristic of exposure to carcinogens and tumor promoters such as DNA damage or oxidative stress may account for the high correlation of this assay with Tg.AC activity.

Following a first-stage screen with the gadd153-CAT reporter with a second-stage screen consisting of a {zeta}-globin promoter construct expressed in K562 cells improved the correlation with Tg.AC responsiveness of the in vitro screen. In this hierarchical binary tree approach, 89% of the Tg.AC positive compounds tested either induced gadd153-CAT by greater than 2-fold, or induced Zeta-Luc activity assay by 8-fold or more. Fetal globin expression is typically enhanced in K562 cells by compounds that induce erythroid differentiation, such as hemin, hydroxyurea, phenylacetate, and antineoplastic drugs (Jeannesson et al., 1984Go; Rutherford et al., 1981Go; Samid et al., 1992Go; Xu and Zimmer, 1998Go) By itself, the {zeta}-globin promoter–driven luciferase assay had a lower correlation with Tg.AC activity than gadd153-CAT due to the greater than 2-fold induction seen for Zeta-Luc with most of the genotoxic compounds tested (both carcinogens and mutagenic noncarcinogens). Zeta-Luc activity was induced by lower doses of both of the PKC activators tested than gadd153-CAT or fos-CAT, but was not induced by any of the non-PKC activating tumor promoters in the study. The relatively poor correlation of Zeta-Luc and in vivo Tg.AC activity could be explained by a mechanism of papilloma induction in Tg.AC mice that occurs through reversal of the silencing of the transgene locus in cell compartments that already express proteins, such as GATA transcription factors, necessary for basal expression of erythroid specific genes.

The fos-CAT assay had the second highest overall concordance with Tg.AC responsiveness (71%) of the in vitro assays we tested. Although fos is typically induced during development, cell differentiation, and by mitogenic stimulation, it is also involved in the cellular stress response to heat shock and ultraviolet radiation (Devary et al., 1992Go). Fos is highly inducible by PKC activation and to a lesser extent by DNA-damaging agents (Hollander and Fornace, 1989Go). The larger number of non–stress-response mechanisms that can activate fos expression may account for the lower correlation of this assay compared to the gadd153-CAT assay with this group of compounds.

The p53RE-CAT assay was the least responsive of the four in vitro assays tested. Only 4 of 10 Tg.AC positive compounds tested induced p53 response element–driven CAT expression 2-fold or greater (Table 1Go). The p53 gene is inducible by IR, UV, DNA damaging agents, and medium depletion (Zhan et al., 1993Go), but unlike the other three promoter elements in the screen, p53RE-CAT is not induced by PMA. The p53RE-CAT construct, which contains three repeats of the p53 response element and is stably expressed in HepG2 cells, was only weakly responsive to the DNA-methylating agent MMS in one published study (Todd et al., 1995Go). In contrast, a p53RE-CAT construct containing 13 repeats of the p53 response element linked to CAT and stably expressed in RKO cells were highly induced by similar doses of MMS in a different report (Zhan et al., 1993Go). It is therefore not clear whether the lack of response to Tg.AC positive compounds in the p53RE-CAT assays conducted for this report is due to limited mechanistic overlap between p53 and Tg.AC transgene activation pathways, suboptimal culture conditions (Deffie et al., 1995Go), or an inherently low dynamic range of response of this assay. As all four of the p53RE-CAT–positive compounds were also positive in the gadd153-CAT assay, the p53RE-CAT assay contributed no apparent additional predictive value.

The usefulness of an in vitro prescreen for selecting potential false positives in the Tg.AC assay among a large set of noncarcinogenic pharmaceuticals may be limited by the failure of an in vitro assay in replicating in vivo conditions. Drugs that act through mechanisms dependent on tissue-specific factors or metabolic capabilities not present in a homogenous cultured cell lines will not be detected in this screen. Drug activities which were observable in vitro may not be detected in vivo because the drug is not sufficiently soluble in a solvent suitable for skin paint application to the back of a Tg.AC mouse. Additionally, cultured cells may be more sensitive to the toxic effects of pharmaceuticals compared to the follicular stem cells of the Tg.AC mouse. Alternately, organ-specific toxicities may limit in vivo dosing levels prior to achievement of application site cytotoxicity or genotoxicity.

The in vitro promoter assays that we have shown here to be correlative with Tg.AC activity will be used in a novel approach to selecting candidate compounds that challenge the specificity of the Tg.AC assay towards pharmaceuticals. In subsequent studies, 100 noncarcinogenic pharmaceuticals will be screened for potential inductive activity of reporter gene constructs fused to the promoters for gadd153, {zeta}-globin, and fos. Compounds with robust induction activity and good dermal availability will be candidates for testing in the Tg.AC skin paint assay as potential false positives. The outcome of these studies will help in the assessment of the specificity of the Tg.AC assay, which is one of the short-term carcinogenicity assays that has the potential to accelerate the process of evaluating the carcinogenic potential of pharmaceuticals.


    NOTES
 
1 To whom correspondence should be addressed at Division of Applied Pharmacology Research, OTR/OPS/CDER/FDA, 8301 Muirkirk Rd., Laurel, MD 20708. Fax: (301) 594-3037. E-mail: sistare{at}cder.fda.gov. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Beard, S. E., Capaldi, S. R., and Gee, P. (1996). Stress responses to DNA damaging agents in the human colon carcinoma cell line, RKO. Mutat. Res. 371, 1–13.[ISI][Medline]

Charnay, P. and Maniatis, T. (1983). Transcriptional regulation of globin gene expression in the human erythroid cell line K562. Science 230, 1281–1283.

Contrera, J. F., Jacobs, A. C., and DeGeorge, J. J. (1997). Carcinogenicity testing and the evaluation of regulatory requirements for pharmaceuticals. Regul. Toxicol. Pharmacol. 25, 130–145.[ISI][Medline]

Dean, A., Ley, T. J., Humphries, R. K., Fordis, M., and Schechter, A. (1983). Inducible transcription of five globin genes in K562 human leukemia cells. Proc. Natl. Acad. Sci. U.S.A. 80, 5515–5519.[Abstract]

Deffie, A., Hao, M., Montes de Oca Luna, M., Hulboy, D. L., and Lozano, G. (1995). Cyclin E restores p53 activity in contact-inhibited cells. Mol. Cell Biol. 15, 3926–3933.[Abstract]

Devary, Y., Gottlieb, R.A., Smeal, T., and Karin, M. (1992). The mammalian ultraviolet response is triggered by activation of src tyrosine kinases. Cell 71, 1081–1091.[ISI][Medline]

Eastin, W. C., Haseman, J.K., Mahler, J. F., and Bucher, J. R. (1998). The National Toxicology Program evaluation of genetically altered mice as predictive models for identifying carcinogens. Toxicol. Pathol. 26, 461–473.[ISI][Medline]

Hansen, L. A., Spalding, J.W., French, J. E., and Tennant, R. (1995). A transgenic mouse model (TG.AC) for skin carcinogenesis: Inducible transgene expression as a second critical event. Prog. Clin. Biol. Res. 391, 223–235.[Medline]

Hansen, L. A., and Tennant, R. (1994). Focal transgene expression associated with papilloma development in v-Ha-ras-transgenic Tg.AC mice. Mol. Carcinog. 9, 143–154.[ISI][Medline]

Holden, H. E., Stoll, R. E., Spalding, J. W., and Tennant, R. W. (1997). Evaluation of oxymetholone in the Tg.AC transgenic mouse model for accelerated carcinogenicity detection. Fundam. Appl. Toxicol. 36(1), 178–179.

Hollander, M. C., and Fornace, A. J., Jr. (1989). Induction of fos RNA by DNA-damaging agents. Cancer Res. 49, 1687–1692.[Abstract]

Humble, M. C., Szcesniak, C. J., Luetteke, N. C., Spalding, J. W., Cannon, R. E., Hansen, L. A., Lee, D. C., and Tennant, R. W. (1998). TGF{alpha} is dispensable for skin tumorigenesis in Tg.AC mice. Toxicol. Pathol. 26, 562–569.[ISI][Medline]

International Agency for Research on Cancer (1987). Overall evaluations of carcinogenicity: an update of IARC monographs volumes 1 to 42. IARC Monographs on the evaluation of the carcinogenic risk of chemicals to humans, Supplement 7. IARC, Lyon, France.

International Agency for Research on Cancer (1979). Sex Hormones (II). IARC Monographs on the evaluation of the carcinogenic risk of chemicals to humans, Vol. 21, p. 279. IARC, Lyon, France.

ICH Expert Safety Working Group (1996). Testing for carcinogenicity of pharmaceuticals. Federal Register 61, 43298–43300.

Jeannesson, P., Ginot, L., Manfait, M., and Jardillier, J. C. (1984). Induction of hemoglobin synthesis in human leukemic K562 cells by adriamycin. Anticancer Res. 4, 47–51.[ISI][Medline]

Leder, A., Kuo, A., Cardiff, R. D., Sinn, E., and Leder, P. (1990). V-Ha-ras transgene abrogates the initiation step in mouse skin tumorigenesis: Effects of phorbol esters and retinoic acid. Proc. Natl. Acad. Sci. U.S.A. 87, 9178–9182.[Abstract]

Luethy, J. D., and Holbrook, N. J. (1992). Activation of the gadd153 promoter by genotoxic agents: a rapid and specific response to DNA damage. Cancer Res. 52, 5–10.[Abstract]

National Toxicology Program (1986). Carcinogenesis studies of ethyl acrylate (CAS No. 140885) in F344/N rats and B6C3F1 mice (gavage studies). Technical Report Series No. 259, U.S. Department of Health and Human Services, Public Health Service, Bethesda, MD.

National Toxicology Program (1998). Toxicology and carcinogenesis studies of diethanolamine (CAS No. 111–42–2) in F344/N rats and B6C3F1 mice (dermal studies). Technical Report Series No. 478, U.S. Department of Health and Human Services, Environmental Health Information Service, Research Triangle Park, NC.

Nylander-French, L. A., and French, J. E. (1998). Tripropylene glycol diacrylate but not ethyl acrylate induces skin tumors in a twenty-week short-term tumorigenesis study in Tg.AC (v-Ha-ras) mice. Toxicol. Pathol. 26, 476–483.[ISI][Medline]

Oh, H-S., and Smart, R. C. (1995). Estradiol arrests hair growth and promotes tumors in mouse skin. Proc. Amer. Assoc. Cancer Res. 36, 180.

Price, B. D., and Calderwood, S. K. (1992). Gadd45 and Gadd153 messenger RNA levels are increased during hypoxia and after exposure of cells to agents which elevate the levels of the glucose-regulated proteins. Cancer Res. 52, 3814–3817.[Abstract]

Robinson, D. (1998). The International Life Sciences Institutes's role in the evaluation of alternative methodologies for the assessment of carcinogenic risk. Toxicol. Pathol. 26, 474–475.[ISI][Medline]

Rutherford, T., Clegg, J. B., Higgs, D. R., Jones, R. W., Thompson, J., and Weatherall, D. J. (1981). Embryonic erythroid differentiation in the human leukemic cell line K562. Proc. Natl. Acad. Sci. U.S.A. 78, 348–352.[Abstract]

Samid, D., Yeh, A., and Prasanna, P. (1992). Induction of erythroid differentiation and fetal hemoglobin production in human leukemic cells treated with phenylacetate. Blood 80, 1576–1581.[Abstract]

Slaga, T. J., Fischer, S. M., Nelson, K., and Gleason, G. L. (1980). Studies on the mechanism of skin tumor promotion: evidence for several stages in promotion. Proc. Natl. Acad. Sci. USA 77, 3659–3663.[Abstract]

Spalding, J. W., Momma, J., Elwell, M. R., and Tennant, R. W. (1993). Chemically induced skin carcinogenesis in a transgenic mouse line (Tg.AC) carrying a v-Ha-ras gene. Carcinogenesis 14, 1335–1341.[Abstract]

Tabilio, A., Pelicci, P. G., Vinci, G., Mannoni, P., Civin, C. I., Vainchenker, W., Testa, U., Lipinski, M., Rochant, H., and Breton-Gorius, J. (1983). Myeloid and megakaryocytic properties of K-562 cell lines. Cancer Res. 43, 4569–4574.[Abstract]

Tennant, R. W., French, J. E., and Spalding, J. W. (1995). Identifying chemical carcinogens and assessing potential risk in short-term bioassays using transgenic mouse models. Environ. Health Perspect. 103, 942–950.[ISI][Medline]

Tennant, R. W., Spalding, J., and French, J. E. (1996). Evaluation of transgenic mouse bioassays for identifying carcinogens and noncarcinogens. Mutat. Res. 365, 119–127.[ISI][Medline]

Thompson, K. L., Rosenzweig, B. A., and Sistare, F. D. (1998). An evaluation of the hemizygous transgenic Tg.AC mouse for carcinogenicity testing of pharmaceuticals. II. A genotypic marker that predicts tumorigenic responsiveness. Toxicol. Pathol. 26, 548–555.[ISI][Medline]

Thompson, K. L., Rosenzweig, B. A., Honchel, R., Blanchard, K. T., Stoll, R. E., and Sistare, F. D. (1999). Evidence for a critical palindromic orientation of transgene promoter sequence for tumorigenic responsiveness among Tg.AC transgenic mice. The Toxicologist 48, 255.

Todd, M. D., Lee, M. J., Williams, J. L., Nalezny, J. M., Gee, P., Benjamin, M. B., and Farr, S. B. (1995). The CAT-Tox (L) assay: a sensitive and specific measure of stress-induced transcription in transformed human liver cells. Fundam. Appl. Toxicol. 28, 118–128.[ISI][Medline]

Wang, X-Z., Lawson, B., Brewer, J., Zinzser, H., Sanjay, A., Mi, L., Boorstein, R., Kreibich, G., Hendershot, L., and Ron, D. (1996). Signals from the stressed endoplasmic reticulum induce C/EBP homologous protein (CHOP/GADD153). Mol. Cell. Biol. 16, 4273–4280.[Abstract]

Xu, J. and Zimmer, D. B. (1998). Differential regulation of A{gamma} and G{gamma} fetal hemoglobin mRNA levels by hydroxyurea and butyrate. Exp. Hematol. 26, 265–272.[ISI][Medline]

Zhan, Q., Carrier, F., and Fornace, A. J. (1993). Induction of cellular p53 activity by DNA-damaging agents and growth arrest. Mol. Cell. Biol. 13, 4242–4250.[Abstract]