REVIEW Variability in Gene Expression and Tumor Formation within Genetically Homogeneous Animal Populations in Bioassays 1

FUNDAMENTAL AND APPLIED TOXICOLOGY 29, 176-184 (1996) ARTICLE NO 0019 REVIEW Variability in Gene Expression and Tumor Formation within Genetically Homogeneous Animal Populations in Bioassays 1 GEORGE L. WOLFF 2 National Center for Toxicological Research, Food and Drug Administration, U.S. Department of Health and Human Services. Jefferson, Arkansas 72079; and Departments of Biochemistry/Molecular Biology and Pharmacology/Toxicology, University of Arkansas for Medical Sciences, Little Rock, Arkansas 72205 Variability in Gene Expression and Tumor Formation within Genetically Homogeneous Animal Populations in Bioassays. WOLFF, G. L. (1996). Fundam. Appl. Toxicol. 29, 176-184. Received April 14, I995; accepted August 21, 1995 Considerable variation in susceptibility to tissue-specific tumor formation in response to chronic treatment with low or intermediate dose levels of putative carcinogens is observed within populations of genetically homogeneous test animals under controlled environmental conditions. Experimental evidence from National Toxicology Program studies is reviewed, as are studies of differing degrees of carcinogenic response and tumor promotion among isoand congenic mice carrying the A" y (viable yellow) mutation. The data suggest that individual variations in carcinogenic response among genetically homogeneous animals may derive primarily from differences in regulation of gene transcription. Differences in posttranscriptional and posttranslational processing of gene products are probably also contributing factors. The viable yellow A"*Ia mouse model system is uniquely suited for investigating the developmental and molecular bases of this phenotypic variability in genetically homogeneous populations since various degrees of carcinogenic response and promotion of tumor formation can be predicted, a priori, at least as early as 7 days of age by correlation with coat color patterns. Ectopic expression of the agouti protein results in enhanced susceptibility to tumor formation in tissues which are already sensitized to neoplastic transformation by their strain genome. The differences in tumorigenic response and coat color pattern among A""I mice appear to be associated with different DNA methylation states of the promoter of an intracisternal A particle inserted into exon 1A of the agouti gene. 1 An earlier version was presented at the symposium Environmental Mutagens and Human Health held February 10 12. 1993 in Madras. India, under the sponsorship of the Environmental Mutagen Society of India. 2 Correspondence and reprint requests should be addressed to the author at Division of Nutritional Toxicology, NCTR. 3900 NCTR Road. Jefferson. AR 72079-9502. Fax: (501) 543-7635. E-mail: GWOLFF@FDANT.NC- TR.FDA GOV. 0272-0590/96 SI2 00 176 INTRODUCTION Data generated in chronic carcinogenicity studies with genetically homogeneous test animals (inbred strains and their F, hybrids) are major components in the estimation of health risks incurred by human populations exposed to particular xenobiotics. In such studies environmental conditions of the animals are closely controlled. Long-term exposure to the same low or intermediate dose level of test chemical is essentially the same among all animals in the same dose/sex/genotype group. Yet, not all animals in a given treatment group develop the expected neoplasm. Among those animals in which tumors do arise, there are differences in the time-to-tumor from the beginning of toxicant exposure. Obviously, considerable variation exists in the sensitivity of tissue-specific responses to prolonged tumorigenic stimuli among individuals whose transcribed DNA sequences are essentially identical. Before the advent of molecular biology, Wright (1949) stated "... genetic uniformity is compatible with great phenotypic variability, since accidents of development and environmental influences are, of course, not controlled in this way." Since such variation plays an important role in data analysis, interpretation and application to risk assessment, its likely source(s) should be understood and taken into consideration. The seeming paradox of phenotypic variability in spite of genetic homogeneity was discussed over 30 years ago (Wolff, 1961). Now, the explosion of knowledge regarding gene regulation in cell biology and carcinogenesis has made experimental approaches to this problem possible. Inbred and F, hybrid populations differ from each other in that the two alleles at each autosomal locus in an inbred population are identical, whereas in an F, hybrid population they may be different. Thus, while all autosomal loci in inbred organisms are homozygous. all F, hybrid offspring

GENE EXPRESSION AND TUMORIGENIC RESPONSE 177 of a cross between two inbred strains will carry a unique combination of homozygous and heterozygous loci characteristic for that F, hybrid. Therefore, all members of a F, hybrid population are just as genetically homogeneous as are those of an inbred population. A brief clarification of the degree of genetic identity within a population of inbred or F, animals may aid in consideration of the hypothesis proposed; Bailey (1978, 1982) provides a more detailed discussion. Residual heterozygosity in an inbred strain is defined as the proportion of loci that are not yet in the genetically fixed state (both parents homozygous for the same allele) at a specified generation of inbreeding (after Bailey, 1978). During full-sib inbreeding, complete freedom from heterogenic sections of mouse chromosomes (p = 0.99) is not attained until F^ (the 60th generation of full sib-mating) (Bailey, 1978). Opposing this decrease in residual heterozygosity is the spontaneous mutation frequency of about 10" 5 per gamete per locus. If there are 30,000 protein-coding genes, this rate would be about 0.3 per gamete (Bailey, 1978). After F 40, the rate of decrease in residual heterozygosity is balanced by the mutation rate. In practice, the inevitability of spontaneous mutations in inbred strains is handled by using a "bottleneck" of single full-sib mating pairs to propagate the strain every 3-5 generations. This results in the fixation in the subsequent generations of any new mutations carried by the pair, and the elimination of any new mutations which have occurred in the breeding population but are not carried by the bottleneck pair. Under this breeding scheme, there is minimal heterozygosity within each generation, but not necessarily between generations. In the present context this means that, for practical purposes, all inbred mice of any single generation are essentially genetically "identical." In the case of F, hybrids it is likely that new recessive mutations in either parent strain would be masked by dominant alleles from the other parent strain. Therefore, the practical effect of residual heterozygosity on carcinogenic responses in F, hybrids may be negligible. Since the animals are genetically homogeneous, phenotypic differences in response to potentially toxic stimuli must reflect differences in the temporal and/or quantitative expression of, at least, some of their genes. These differences presumably are traceable to factors in the exogenous environments of the animals which influence the presence and concentration of endogenous regulatory molecules, e.g., growth factors and hormones. Intra- and extracellular conditions, such as ph, also modulate the activities of genes and their products. These conditions and factors, in turn, modulate directly or indirectly the expression of genes encoding transcription factors and other molecular entities involved in the regulation of gene expression. An example of such external environmental factors which inevitably differ among individual animals, even under closely controlled and constant environmental conditions, is the behavioral dominance relationships among cage mates leading to neuroendocrine, nutritional, and immunological differences among isogenic and congenic animals. Gene-based differences in metabolic and other defensive processes appear to be masked at high toxicant levels. A rather striking example of this was observed in mammary tumor latency between congenic [same alleles at all loci except, in this case, the agouti locus] yellow A^'IA and agouti Ala (BALB/c X VY-A >T )F, female mice following treatment with either 1.5 or 6.0 mg dimethylbenz[a]anthracene. In the low dose group tumors appeared about 28% faster in the yellow than in the agouti mice, whereas in the high dose group the tumors in both genotypes appeared much more rapidly and there was only about 6% difference in tumor latency between the two genotypes (Wolff et ai, 1982). To provide background information for the later discussion of phenotypic variability observed in a number of studies, the viable yellow A vy mouse system used in most of them will be described first in some detail. The molecular aspects of the A"' mutation on which this system is based have been elucidated only recently (Duhl et ai, 1994; Michaud et al, 1994). MODEL SYSTEM FOR DEFINING DEVELOPMENTAL/ MOLECULAR BASES FOR DIFFERENTIAL RESPONSIVENESS TO XENOBIOTICS A model system for investigating the developmental and molecular bases of phenotypic variation is provided by congenic viable yellow (A vy ld) mice since the same genotype is expressed in several coat color phenotypes which are associated with different degrees of susceptibility to tumor formation. The coat color patterns range from clear yellow through various degrees of agouti/black mottling on a yellow background to pseudoagouti mice and can be identified from 7 days of age. Pseudoagouti mice are indistinguishable, grossly, from species-type agouti mice. Yellow and mottled yellow mice become obese as adults, but pseudoagouti mice remain lean. The A mutation is maintained congenically in the same inbred strains as the a allele by a mating system which results in "forced heterozygosis" at the agouti locus. In this system A^/a and a/a sibs are mated to produce litters in which mottled yellow (hereafter yellow) A^la, pseudoagouti A^la, and black a/a mice segregate. In the case of the VY/ WffC3Hf/Nctr-/r- and YS/AVffCSHf/Nctr-^"' strains, this has now continued for more than 100 generations. Therefore, it is reasonable to assume that the three mouse phenotypes in each strain have the same VY or YS background genome, respectively. The mottled yellow and pseudoagouti A^la mouse phenotypes appear to be associated with the state of methylation

178 GEORGE L. WOLFF of the promoter of the intracisternal A particle (IAP, a retrotransposon) inserted in exon 1A of the agouti gene. Hypomethylation of this heterologous promoter results in continuous transcription of agouti and production of phaeomelanin (yellow pigment) by hair follicle melanocytes. Methylation of this promoter results in resumption of control of agouti transcription by the normal promoter sequences of the agouti gene. When this has occurred in the progenitor cells of clones of hair follicle cells in different locations in the skin, islands of agouti/black hair result from the eumelanin synthesis in these clones resulting in the "mottled yellow" appearance of the animal. Methylation of the IAP promoter in all or most of the clones results in the pseudoagouti phenotype. In A^'la mice the agouti protein is expressed ectopically, more or less continuously, in most tissues, e.g., brain, liver, lungs, rather than being transcribed only during formation of the subapical phaeomelanic band in the hair as in agouti Al or A w l- mice (Duhl et ai, 1994). The agouti protein binds with high affinity to the a-melanocyte-stimulating hormone (a-msh) receptor MC1-R (melanocortin-1 receptor) and to another melanocortin receptor, MC4-R, located in the brain (Lu et al., 1994; Mountjoy et al., 1994). By preventing binding of a-msh to its receptors, the agouti protein prevents activation of adenylyl cyclase and the subsequent elevation of c-amp concentration in melanocytes and other cells. Ectopic expression of the agouti protein also has been reported to elevate intracellular Ca 2+ concentrations in soleus and gastrocnemius muscles of mottled yellow A^la mice compared with pseudoagouti A^'la and black a/a mice. This action of agouti protein was confirmed by the observation that it increased Ca 2+ influx in L6-cultured myocytes and in soleus myocytes from a/a mice (Zemel et al., 1995). The regulatory interrelations between adenylyl cyclase activation and intracellular cyclic AMP and Ca 2+ concentrations (Cooper et al., 1995) suggest that the two cellular effects of the agouti protein, identified so far, may produce interactive regulatory effect(s) on cellular metabolism dependent on cell type. The presence of the agouti protein in ectopic sites results not only in enhancement of growth, but also in obesity and diabetes (Yen et al., 1994) indicating an effect(s) on diverse cell types. Although no relevant data are available so far. it seems reasonable to postulate that the primary cellular targets) of the agouti protein is(are) similar in all tissues where it is expressed. The A^' mutation qualifies as a dominant oncogene because its constant and ectopic expression results in increased tumor formation. Fewer liver and lung tumors developed in the lean pseudoagouti A^'la mice treated with lindane than in the obese yellow A"I a mice, but still more than in their lean black a/a sibs (Wolff et al., 1987). Differential methylation of the promoter of an IAP inserted in exon 1A of the agouti gene provides a mechanistic explanation for this example of differential expression of dominant oncogenes in genetically homogeneous cells. Unpublished data (Dunkel et al.) indicate that A^la cells in early passage proliferate more rapidly than a/a cells. Spontaneous focus formation occurred in cell lines derived from A^'la neonates of two different inbred strains but not in cell lines derived from their a/a sibs (Hsiao et al., 1996). These observations strongly suggest that the constant presence of the agouti protein affects cell proliferation, and probably transformation. The specific steps in the cell cycle or in the transformation process putatively affected, directly or indirectly, by the activity of the agouti protein have not yet been identified. While no agouti locus mutation, equivalent to any found in the mouse, has yet been identified in Homo sapiens, the homolog of the agouti locus on mouse chromosome 2 has been located in the same syntenic segment on human chromosome 20q (Kwon et al., 1994). It has been designated agouti signaling protein (AS1P) (Wilson et al., 1995) and is expressed in various tissues; its function in human cells is unknown at present. DIFFERENTIAL RESPONSIVITY TO TUMORIGENS Compounds Tested under the National Toxicology Program Each inbred rodent strain or F, hybrid exhibits an array of constitutive tissue-specific prevalences of neoplasms. Treatment with a carcinogen increases these frequencies above the constitutive levels. Examples of increased tumor prevalence in B6C3F, mice fed test compounds for 24 months at two dose levels (Haseman, 1985) are shown in Table 1. In both treated and untreated groups, there were individuals of identical genetic constitution which were resistant to formation of the expected tumor. In Table 1 the proportions of "responsive" mice which developed a specific tumor type in response to a particular toxicant are listed. If these numbers are subtracted from the total mice treated, the proportion of mice resistant to 24 months' exposure to the given dose level of the toxicant is obtained. These F, hybrid rodent populations apparently consist of subpopulations of responsive and resistant animals of identical genotype. The data suggest that these apparent subgroups actually may represent a spectrum of degrees of responsiveness to a given tumorigenic stimulus. The animals with "spontaneous" neoplasms, i.e., control animals with tumors, are the most responsive phenotypes (Category 1), while those animals which develop a neoplasm in response to the low dose of carcinogen are the more responsive category (Category 2). Those animals responding only to the highest dose level are less responsive phenotypes (Category 3), while those animals which failed to develop a neoplasm in response to the highest dose are the least responsive (most resistant) phenotypes (Category 4).

GENE EXPRESSION AND TUMORIGENIC RESPONSE 179 TABLE 1 Proportions of B6C3Fi Mice Responsive to Tumorigenesis at Different Dose Levels of the Test Compound in 24-Month Feeding Studies under the National Toxicology Program" Dose levels Test compound Sex Neoplasm Control Low High Di(2-ethylhexyl) adipate 2,6-Dichloro-p-phenylenediamine Zearalenone Ziram M F M F M F F Liver* Liver Liver adenoma Liver Pituitary adenoma Lung adenoma 26 r (13/50)" 6 (3/50) 8 (4/50) 12 (6/50) 0 (0/40) 6 (3/46) 4 (2/50) 41 (20/49) 38 (19/50) 14 (7/50) 12 (6/50) 9 (4/45) 5 (2/43) 10 (5/49) 55 (27/49) 37 (18/49) 30 (15/50) 32 (16/50) 14 (6/44) 31 (13/42) 20 (10/50) " Data from Haseman (1985). b Adenoma or carcinoma. c Percentage. d Tumor bearers/total treated mice. The strain and tissue-specific incidence of spontaneous tumor formation is presumably associated with endogenously generated carcinogens, as well as with naturally occurring carcinogens and tumor promoters in the feed. However, the strain and tissue-specific incidence of formation of such tumors is likely related to the relative sizes of the responsive and resistant subgroups in which differential gene transcription occurs. Definition of the molecular mechanisms and specific genes involved in the differential responsiveness to phenobarbital observed among A^'la (C3H-MTV"/HeN X VY/WffC3Hf/Nctr-A VT )Fl hybrid male mice (Wolff et al., 1991) would further our understanding of the mechanistic basis for different degrees of responsivity. Haseman et al. (1989) discussed environmental factors which may influence tumor prevalence in carcinogenicity studies. These factors, like the toxicant being assayed, induce responses from each animal, be they physiologic, metabolic, or behavioral. The qualitative and quantitative characteristics of each animal's responses to environmental stimuli ultimately are determined by the animal's genome. Therefore, if the genetic constitution of the animals is homogeneous, all members of the test population should respond similarly, if not identically, to each environmental stimulus. Why is this not the case? 7,12-Dimethylbenz[a\anthracene 7,12-Dimethylbenz[a]anthracene (DMBA), a complete genotoxic carcinogen, induced mammary adenocarcinomas and adenoacanthomas in female yellow A /A and agouti Ala (BALB/cStCrlfC3Hf/Nctr X VYAVffC3Hf/Nctr-^"0F, hybrid mice (Wolff et al, 1982). DMBA in corn oil was administered by gavage, the low dose group receiving 0.75 mg each week for 2 weeks, the high dose group receiving 1.0 mg each week for 6 weeks. Thus, in the low dose group there were two periods of DMBA exposure during which cellular transformation could be induced, while in the high dose group there were six periods of exposure during which initiation could occur. Consistent with the shorter time-totumor for yellow than agouti mice (p = 0.0007 at high dose, p = 0.0021 at low dose), there was a small, but consistent, excess of yellow mice bearing tumors compared to the agouti mice (Table 2). No tumors developed in control mice gavaged only with corn oil. Within genotypes different degrees of responsiveness to the carcinogen were observed, since, even when treated with the high dose, there were 13 yellow and 28 agouti mice which failed to develop any mammary tumors (Table 2). Additionally, the latency period to tumor detectability, which reflects the relative rates of proliferation, differentiation, and apoptosis of the transformed cells, varied from less than 100 days to more than 300 days after the first DMBA dose; this was true for both the obese yellow and lean agouti mice. Lindane (y-hexachlorocyclohexane) Female mice of the (YS/ChWffC3Hf/Nctr X VY/ WffC3Hf/Nctr-A^F hybrid were used in this study to mini-

180 GEORGE L. WOLFF TABLE 2 Proportions of Yellow A""IA and Agouti Ala (BALB/c x VY- A vy )F, Hybrid Female Mice Developing Mammary Tumors after Two Weekly Gavages of 0.75 mg 7,12-Dimethylbenz[a]anthracene (DMBA) or Six Weekly Gavages of 1.0 mg DMBA" Dose Vehicle control 1.5 mg DMBA 6.0 mg DMBA " Data from Wolff et al. b Tumor bearers. " Percentage. Yellow responsive mice*/ total treated mice (1982). 0 c (0/48) 43 (41/95) 86 (83/96) Agouti responsive mice*/ total treated mice 0 (0/48) 33 (32/96) 71 (67/95) mize inherent susceptibility to hepatocarcinogenesis as much as possible. In addition to differing in coat color patterns, the alternative yellow and pseudoagouti A vy la phenotypes also differ in body weight, mottled yellow A vy la mice becoming obese, while the pseudoagouti A vy la mice remain lean. Within both the obese mottled yellow and the lean pseudoagouti populations there was variation with respect to inducibility of lindane-associated hyperplasia and neoplasia (Wolff et al, 1987). Lindane ingestion resulted in lung tumor formation in 15% (treated, 19%; control, 4%) of the yellow, in 8% (treated, 14%; control, 6%) of the pseudoagouti mice, but in only 1% (treated, 3%; control, 2%) of the congenic black a/a mice (Table 3). Since the A vy gene may potentiate sponta- TABLE 3 Prevalence of Hepatocellular Adenomas and Lung Tumors in Yellow, Pseudoagouti, and Black Female (YS x VY-A" y )Fi Hybrid Mice Fed 160 ppm Lindane in Purina 5010M Diet for 24 Months" Phenotype Yellow A"la Pseudoagouti A"/a Black a/a Yellow A "la Pseudoagouti A"la Black a/a Treated Control Hepatocellular adenomas 35 12 3 19 14 3 1 Data from Wolff et al. (1987). 9 5 6 Lung tumors 4 6 2 Associated with different agouti mrna tissue concentrations (%) 26 7 0 15 8 1 TABLE 4 Proportions of Phenotypes with Different Degrees of Response Among Yellow A""IA and Agouti Ala (C3H X VY-A vy )F, Hybrid Male Mice to 18 Months' Exposure to NIH-31 Diet with or without 0.5% Sodium Phenobarbital 0 Control Yellow Agouti Treated Yellow Agouti Single adenoma 11 16 18 19 Multiple adenomas 1 0.5 36 5 1 f \ i~\ O f l_ \ Data from woirt et al. (1986b). Adenoma + carcinoma 4 1 10 4 Carcinoma 13 15 3 8 None neous transformation in fibroblasts (Hsiao et al, 1996), these data suggest that the A allele may have facilitated, directly or indirectly and in an unknown manner, pulmonary cell transformation. These transformed cells were then stimulated to proliferate by lindane or its metabolites resulting in lung tumor formation. The two phenotypes also exhibited a marked difference in hepatocellular adenoma formation in response to the liver tumor promoter lindane with an incidence of 26% (treated, 35%; control, 9%) among the yellow mice and 7% (treated, 12%; control, 5%) among the pseudoagouti mice. No effect of lindane on liver tumor prevalence among black a/a siblings was observed (Table 3). As in the case of pulmonary cells, it seems likely that more initiated hepatocytes were present in A /a than in a/a livers, with different levels of agouti mrna/protein being associated with different proportions of initiated hepatocytes. Under this scenario, lindane or its metabolites promoted these initiated cells to proliferate and form tumors and did not, itself, take part in the transformation process. Sodium Phenobarbital Body weight and tumor promotion. Sodium phenobarbital (PB, 0.05%), a nongenotoxic tumor promoter, was mixed in NIH-31 diet and fed to male yellow A vy /A and agouti Ala (C3H-MTV"/HeN X VYAVffC3Hf/NctM"')F, hybrid mice for 18 months (Wolff et al, 1986a). The parent C3H/ HeN strain has a high incidence of spontaneous liver tumors which is, in part, related to expression of the Hcs gene (Drinkwater, 1994). Some control and treated animals developed only a single hepatocellular adenoma, some developed multiple adenomas, still others developed both an adenoma and a carcinoma, while some exhibited only carcinoma or no liver neoplasm at all (Table 4). Thus, within each of these populations (A^'/A and Ala) of congenic mice, there were several different levels of response to endogenous carcino- 70 68 32 65

181 GENE EXPRESSION AND TUMORIGENIC RESPONSE 600 i i Light mice EH3 Heavy mice "? 400 c <L> o 200 n^ flft cyt P450 EROD T60Hy<J TCT 7GT PROO DCNB ONCB FIG. 1. Relative inducibility by sodium phenobarbital of hepatic drugmetabolizing enzymes in light and heavy A"IA (C3H X VY-/^'T)F, hybrid male mice, cyt P450, total cytochrome P450; EROD, 7-ethoxyresorufin-Odeethylase; T6/?Hyd, testosterone-6 /3-hydroxylase; TGT, activated UDPglucuronyltransferase activity toward testosterone: -ygt, -y-glutamyltranspeptidase; PROD, 7-pentoxyresorufin-O-dealkylase; DCNB. glutathione Stransferase activity toward l,2-dichloro-4-nitrobenzene; DNCB, glutathione 5-transferase activity toward l-chloro-2,4-dinitrobenzene. Values are expressed as {PB-treated activity/control activity - control activity} X 100. Reproduced with permission of the publisher, from Wolff et al. (1991). genie stimuli, as well as to the treatment with a single dose level of phenobarbital. Examination of the body weight curves revealed that the mice with multiple adenomas had a higher rate of body weight gain, beginning about 5 months after the start of the study, than those which had only one tumor or had failed to develop any tumors (Wolff et al., 1986b). Isozyme activity and inducibility. Did these subgroups, which differed in the response to PB with respect to tumorigenesis and body weight gain, also differ in the inducibility of PB-dependent enzymes by PB? Differential inducibility would reflect differential transcription of the genes coding for the enzymes in question and would suggest that, while the DNA coding sequence of a gene might be identical in all individuals, its transcription might be differentially regulated in different individuals. To explore this possibility, a follow-up study (Wolff et al., 1991) was undertaken with yellow mice since the largest variation in body weight and tumor development had been observed among mice of this genotype in the original study. Male yellow A^IA (C3H-MTV~/N X VY/WffC3Hf/NctrAvy) ] hybrid mice were fed NIH-31 diet, containing 0.05% sodium PB, for 7 months beginning at 7-8 weeks of age. At the end of that period, a battery of Phase I and Phase II isozyme activities was assayed in the livers of the heaviest and lightest mice in the control and treated groups. Differences between the heavier and lighter mice in responsiveness to PB were demonstrated by the assay for PBinducible P450IIB-selective 7-pentoxyresorufin-O-dealkylase activity (Fig. 1). A sixfold increase in the treated "heavy" mice compared to the control "heavy" mice was observed. In contrast, in the "light" mice there was only a threefold increase in activity due to PB induction. By comparison, the non-pb-inducible P450IA-selective 7ethoxyresorufin-Odeethylase activity was increased only 70 and 84%, respectively, in the treated heavy and light mice. Some constitutive isozyme activities, not dependent on P450IIB, likewise differed between the heavy and light mice, e.g., testosterone 6/3-hydroxylase (P450IIIA), 7-ethoxyresorufin-O-deethylase (P450IA) (Table 5). Also associated with greater body and carcass (body minus liver) weights was increased responsiveness to PB induction of glutathione-5-transferase isozymes of the Nl:l gene family. Increased responsiveness to PB promotion of body weight gain and of hepatocellular adenoma formation was also associated with repression of cytochrome P450IIE isozymes. In contrast, decreased constitutive expression, but increased PB inducibility, of cytochrome P450 isozymes of the IIIA, and possibly IA, gene families and 17-hydroxysteroid UDP-glucuronyltransferase were associated with decreased susceptibility to PB promotion of body and carcass weight and, by implication, decreased promotion of hepatocellular adenoma formation. Thus, the patterns of constituent levels of isozyme activities (Table 5), as well as the patterns of their inducibilities (Fig. 1), differed between the light and heavy phenotypic subgroups, even though the mice were genetically homogeneous. Whether any of these enzymatic differences are actually involved in tumor promotion is unknown; however, the enzyme activity patterns do appear to be markers for differential susceptibility to hepatic tumorigenesis. Retrospectively, we found that those mice which were heavy and those which were light at the end of the study could have been identified by body weight at least as early as 3-4 weeks of age. Therefore, it seems likely that the potential for increased responsiveness to PB was acquired some time during prenatal or early postnatal development, i.e., before any exposure to the toxicant. TABLE 5 Differences in Body Weight and Constitutive Hepatic Enzyme Activities Associated with Differential Responsiveness to Sodium Phenobarbital" Light A"/A mice (n = 6) Initial body weight (g) Final body weight (g) Monooxygenases (nmol/min/mg microsomal protein) 7-Ethoxyresorufin-O-deethylase Testosterone 6/3-hydroxylase Data from Wolff ei al. (1991) 25.6 43 2 ±0.8 ±0.6 0.198 ± 0 014 0 96 ± 0 34 Heavy A"'IA mice (n = 5) 30.9 51.0 ±1.7 ± 0.7 0.277 ± 0.030 2.08 ± 0.43

182 GEORGE L. WOLFF DISCUSSION Differential Responsiveness to Xenobiotics and Regulation of Gene Expression Existence of a spectrum of phenotypes exhibiting differential responsivity to xenobiotics or their metabolites among genetically homogeneous test animals is suggested by the failure of tissue-specific neoplasms to develop in all members of inbred or F, hybrid test animal populations exposed to low or intermediate toxicant dose levels in chronic bioassays. Gartner (1990) studied the basis of variability in quantitative traits, such as body weight, among genetically identical animals and concluded that "Reduction of genetic variability by using inbred strains and reduction of environmental variability by highly standardized husbandry did not remarkably reduce the range of random variability in quantitative biological traits." He found that 70-80% of body weight variation in inbred mice was due to a component, possibly maternal ooplasmic factors, active at or before fertilization, which created random biological variability; only 20-30% of the variability could be ascribed to environmental variation. Local variations in microenvironmental conditions in the maternal reproductive tract, e.g., different sites of placental attachment and vascularization, could alter the regulatory programs for gene expression imprinted in different embryos of identical genotype. For example, the proportions of the A vy /a concepti which develop into the alternative lean pseudoagouti or obese mottled yellow phenotypes are correlated with the dam's strain genome, the sire's genome apparently playing no role (Wolff, 1978). These alternative phenotypes reflect differential methylation of the promoter of an intracisternal A particle (IAP) inserted in exon 1A of the agouti locus (Duhl et al., personal communication); this is also the case with the similar A mpy allele (Michaud et al., 1994). The mechanisms involved in this differential DNA methylation are still unknown. However, the solely maternal association of the ratio of pseudoagouti:yellow phenotypes strongly suggests that expression of the maternal genome is involved, possibly by influencing microenvironmental conditions in the reproductive tract which modulate DNA methylation in embryos. In the phenobarbital study reviewed, the mice which developed or failed to develop multiple hepatocellular adenomas differed from each other, not only in body weight, but also in the pattern of activities of a battery of Phase I and Phase II isozymes. Not only did the patterns of constitutive activities differ, but inducibilities of different isozymes also differed between these subgroups. These phenotypes could already be identified by body weight at weaning. Therefore, the potential for differential responsiveness to PB must have been determined before any exposure to the toxicant, i.e., at some time during prenatal or early postnatal development. These observations suggest that the programs for regulation of the levels of gene transcription and isozyme inducibility are integral to the phenotypic differentiation of individual organisms and, presumably, are subject to the influence of maternal environmental factors. According to the phenoclone hypothesis of embryonic differentiation and development (Mintz and Bradl, 1991), "... virtually all mammalian cell types comprise phenotypically different developmental clones, or phenoclones; and virtually all loci have some alleles subject to such ambiguous cis-acting controls." This hypothesis is based on the occurrence of melanocyte clones which differ from each other in the degree of expression of specific coat color genes, even though the clones are genetically homogeneous. These authors suggest that "... phenotypically variant but genetically identical clones of the same cell type that express a dominant oncogene to different extents could have significantly different degrees of propensity for tumorigenesis. Or different clonal expressions of a recessive gene suppressing tumorigenesis could in some situations simulate the loss of one allele in the underexpressing clones without any actual deletion having occurred." The difference between hepatocellular adenoma formation in yellow and pseudoagouti A^'la mice fed the lindane-supplemented diet is a relevant example of a dominant oncogene affecting tumorigenesis differentially when expressed to different extents. Gene expression is induced, modulated, and repressed by environmental stimuli. Environment, as defined here, includes both intra- and extracellular microenvironments, as well as the macroenvironment of the organism. All mice within each of the congenic F, hybrid populations discussed were genetically homogeneous, i.e., the expressed DNA sequences of each gene were essentially the same among all the mice in each population. Some of the major processes involved in gene expression are transcription, posttranscriptional processing, translation, and posttranslational modification. Each of these environmentally modulated processes is complex. For example, gene transcription is regulated by specific proteins (transcription factors) which themselves are specified by other genes whose transcription is also regulated. Transcription factors act by binding to specific DNA regulatory elements located in the promoter regions of the target genes (Papavassiliou, 1995). Recent reviews (Papavassiliou, 1994; Vallejo, 1994) of the transcriptional control of numerous genes by the camp response element binding/activating transcription factor proteins and of the regulation of its own activation illustrate the complexity of the process and the possibility for variations among even genetically identical individuals. Whether any particular genes or gene patterns are expressed in the cells of an individual tissue depends on the relative concentrations of the proteins involved in the regulation of their expression. The probability that relevant regulatory proteins will be present in sufficiently close proximity to bind to

GENE EXPRESSION AND TUMORIGENIC RESPONSE 183 the specific DNA sequences is dictated by their individual concentrations in each cell. These concentrations are modulated by conditions in the cell's immediate microenvironment. Since certain critical interactions between protein and DNA may depend on their exact concentrations, particular genes or gene patterns will not be expressed in all of the genetically homogeneous cells of the organism. Indeed, this is the ultimate basis for embryonic differentiation of cells and tissues. Synthesis of a polypeptide by translation of the mrna produced by transcription of a particular gene can be affected by ribosomal proteins and alteration of mrna stability. Functioning of the protein can also be affected by its immediate microenvironment via effects on its conformation. The rate of degradation of a protein, which influences its cellular concentration, is likewise modulated by environmental conditions. Phenotypic variability due to multiple functions of gene products has been considered by Hershman (1991) who stated that "the variability of biological response resulting from the employment of genes whose products are used in many contexts occurs as a consequence of (a) differential quantitative induction of primary response genes by distinct ligands, (b) differing kinetics of induction of the primary response genes and their products, (c) formation of changing patterns of heterodimeric transcription factors, with alternative transcriptional capacities, during the course of ligandstimulated responses, (d) cell-type specific restriction of the expression of subsets of primary response genes, (e) ligandand cell-specific post-translational modification of primary response gene products and consequent alterations in the biological properties of these proteins, and (/) differential production of autocrine and paracrine factors that modulate initial inductive responses." Multiple functions of gene products, together with the complex interactions of the many factors involved in the regulation of gene expression, and the complex relationships among the metabolic and physiological processes in which the gene products participate predict that phenotypic variation among isogenic organisms is inevitable. Since there is phenotypic variability within genetically homogeneous animal populations, why should the cheaper, genetically heterogeneous, "outbred" or "random-bred" animals NOT be used for chronic bioassays? Because the essentially undefined genetic variation in these populations is confounded with phenotypic variability so that meaningful data interpretation becomes difficult, if not impossible. Requirements for controlled genetic and phenotypic variation would be satisfied and statistical power would be increased by a design using equal sample sizes of several carefully chosen inbred strains without increasing the total number of animals (Festing, 1995). CONCLUSION Phenotypic variation in responsiveness to toxicants among isogenic organisms is inevitable. Although its developmental and molecular parameters have not been defined, this phenotypic variation is probably associated with environmentally induced variation in the regulation of gene expression. Carcinogenicity bioassay data obtained with a versatile experimental animal model system, using congenic A^la and a/a or Ala Fi hybrid mice, have been reviewed. In this system phenotypes with different degrees of responsiveness to tumorigenic stimuli are associated with specific coat color patterns identifiable as early as 7 days of age. Therefore, it seems uniquely adapted for identification and delineation of developmental pathways and molecular processes underlying phenotypic variation in responsivity to xenobiotics. ACKNOWLEDGMENTS Consultative advice by Michael F. W. Festing and Gregory S. Barsh and preparation of the tables by N. Ruth York are gratefully acknowledged. REFERENCES Bailey. D. W (1978) Sources of subline divergence and their relative importance for subhnes of six major inbred strains of mice. In Origins of Inbred Mice (H. C. Morse III, Ed.), pp. 197-215, Academic Press, New York. Bailey, D. W. (1982). How pure are inbred strains of mice? Immunol. Today 3, 210-214. Cooper, D. M. F., Mons, N., and Karpen, J. W. (1995). Adenylyl cyclase and the interaction between calcium and camp signalling. Nature 374, 421-424. Drinkwater, N. R. (1994). Genetic control of hepatocarcinogenesis in C3H mice. Drug. Metab Rev. 26, 201-208. Duhl. D. M. J., Vneling, H., Miller, K. A., Wolff, G. L., and Barsh, G. S. (1994) Neomorphic agouti mutations in obese yellow mice. Nature Genet. 8, 59-65. Festing, M. F. W. (1995). Use of a multistrain assay could improve the NTP carcinogenesis bioassay. Environ. Health Perspect. 103, 44-52. Gartner, K. (1990). A third component causing random variability beside environment and genotype: A reason for the limited success of a 30 year long effort to standardize laboratory animals? Lab. Anim. 2A, 1\-11. Haseman, J. K. (1985). Issues in carcinogenicity testing: Dose selection. Fundam. Appl. Toxicol. 5, 66-78. Haseman, J. K., Huff, J. E., Rao, G. N., and Eustis, S. L. (1989). Sources of variability in rodent carcinogenicity studies Fundam. Appl. Toxicol. 12, 793-804. Hershman, H. R (1991). Primary response genes induced by growth factors and tumor promoters. Annu. Rev. Biochem. 60, 281-319. Hsiao, W.-L. W., Wolff, G. L., North, B. M., Ollmann, M. M., Barsh, G. S., and Fan, H. Y. (1996). Differential spontaneous transformation in vitro of newly established mouse fibroblast lines carrying or lacking the viable yellow mutation {A"") of the mouse agouti locus. Mol. Carcinogen. 15, 70-80. Kwon, H. Y., Bultman, S. J., Loffler, C, Chen, W-J., Furdon, P. J., Powell, J. G., Usala, A-L., Wilkison, W., Hansmann, I., and Woychik, R. P. (1994). Molecular structure and chromosomal mapping of the human homolog of the agouti gene. Proc. Natl. Acad. Sci. USA 91, 9760-9764. Lu, D., Willard, D., Patel, I. R., Kadwell, S., Overton, L., Kost, T., Luther, M., Chen, W., Woychik. R. P., and Cone, R. D. (1994). Agouti protein

184 GEORGE L. WOLFF is an antagonist of the melanocyte-stimulating-hormone receptor. Nature 371, 799-802. Michaud, E. J., van Vugt, M. J., Bultman, S. J., Sweet, H. O., Davisson, M. T., and Woychik, R. P. (1994). Differential expression of a new dominant agouti allele (A Uvy ) is correlated with methylation state and is influenced by parental lineage. Genes Dev. 8, 1463-1472. Mintz, B., and Bradl, M. (1991). Mosaic expression of a tyrosinase fusion gene in albino mice yields a heritable striped coat color pattern in transgenic homozygotes. Proc. Natl. Acad. Sci. USA 88, 9643-9647. Mountjoy, K. G., Mortrud, M. T., Low, M. J., Simerly, R. B., and Cone, R. D. (1994). Localization of the melanocortin-4 receptor (MC4-R) in neuroendocrine and autonomic control circuits in the brain. Mol. Endocrinol. 8, 1298-1308. Papavassiliou, A. G. (1995). Transcription factors. N Engl. J. Med. 332, 45-47. Papavassiliou, A. G. (1994). The CREB/ATF family of transcription factors: Modulation by reversible phosphorylation. Antkancer Res. 14, 1801-1806. Vallejo. M. (1994) Transcriptional control of gene expression by campresponse element binding proteins. J. Neuroendocrinol. 6, 587-596 Wilson, B. D., Ollmann, M. M., Kang. L., Stoffel, M., Bell, G. I., and Barsh, G. S (1995). Structure and function of ASP, the human homolog of the mouse agouti gene. Hum. Mol. Genet. 4, 223-230. Wolff, G. L. (1961). Some genetic aspects of physiological variability. Cancer Res 21, 1119-1123. Wolff, G. L (1978). Influence of maternal phenotypeon metabolic differentiation of agouti locus mutants in the mouse. Genetics 88, 529-539. Wolff, G. L., Kodell, R. L., Cameron, A. M., and Medina, D (1982). Accelerated appearance of chemically induced mammary carcinomas in obese yellow (A"IA) (BALB/c X VY) F, hybrid mice. /. Toxicol. Environ. Health 10, 131-142. Wolff, G. L., Morrissey, R. L., and Chen, J. J. (1986a). Amplified response to phenobarbital promotion of hepatotumorigenesis in obese yellow A"l A (C3H X VY) F-l hybrid mice, Carcinogenesis 7, 1895-1898. Wolff, G. L., Morrissey, R. L., and Chen, J. J. (1986b). Susceptible and resistant subgroups in genetically identical populations: Response of mouse liver neoplasia and body weight to phenobarbital. Carcinogenesis 7, 1935-1937. Wolff, G. L.. Roberts, D. W., Morrissey, R. L., Greenman, D. L., Allen, R. R., Campbell, W. L., Bergman, H., Nesnow, S., and Fnth, C. H. (1987). Tumorigenic responses to lindane in mice: Potentiation by a dominant mutation. Carcinogenesis 8, 1889-1897. Wolff, G. L, Leakey, J. E. A., Bazare, J., Harmon, J. R., Webb, P. J., and Law, M. G. (1991). Susceptibility to phenobarbital promotion of hepatocarcinogenesis: Correlation with differential expression and induction of hepatic drug metabolizing enzymes in heavy and light male (C3H X VY) F, hybrid mice. Carcinogenesis 12, 911-915. Wright, S. (1949). Differentiation of strains of guinea pigs under inbreeding. Proc. Natl Cancer Conf. 1, 13-27. Yen, T. T., Gill, A. M., Frigeri, L. G., Barsh, G. S., and Wolff, G. L (1994). Obesity, diabetes, and neoplasia in yellow A"*/ mice: Ectopic expression of the agouti gene. FASEB J. 8, 479-488. Zemel, M. B., Kim, J. H., Woychik, R. P., Michaud, E. J., Kadwell, S. H., Patel, 1. R., and Wilkison, W. O. (1995). Agouti regulation of intracellular calcium: Role in the insulin resistance of viable yellow mice. Proc. Natl. Acad. Sci. USA 92, 4733-4737.