Molecular basis for effects of carcinogenic heavy metals on inducible gene expression.

Certain forms of the heavy metals arsenic and chromium are considered human carcinogens, although they are believed to act through very different mechanisms. Chromium(VI) is believed to act as a classic and mutagenic agent, and DNA/chromatin appears to be the principal target for its effects. In contrast, arsenic(III) is considered nongenotoxic, but is able to target specific cellular proteins, principally through sulfhydryl interactions. We had previously shown that various genotoxic chemical carcinogens, including chromium (VI), preferentially altered expression of several inducible genes but had little or no effect on constitutive gene expression. We were therefore interested in whether these carcinogenic heavy metals might target specific but distinct sites within cells, leading to alterations in gene expression that might contribute to the carcinogenic process. Arsenic(III) and chromium(VI) each significantly altered both basal and hormone-inducible expression of a model inducible gene, phosphoenolpyruvate carboxykinase (PEPCK), at nonovertly toxic doses in the chick embryo in vivo and rat hepatoma H411E cells in culture. We have recently developed two parallel cell culture approaches for examining the molecular basis for these effects. First, we are examining the effects of heavy metals on expression and activation of specific transcription factors known to be involved in regulation of susceptible inducible genes, and have recently observed significant but different effects of arsenic(III) and chromium(VI) on nuclear transcription factor binding. Second, we have developed cell lines with stably integrated PEPCK promoter-luciferase reporter gene constructs to examine effects of heavy metals on promoter function, and have also recently seen profound effects induced by both chromium(VI) and arsenic(III) in this system. These model systems should enable us to be able to identify the critical cis (DNA) and trans (protein) cellular targets of heavy metal exposure leading to alterations in expression of specific susceptible genes. It is anticipated that such information will provide valuable insight into the mechanistic basis for these effects as well as provide sensitive molecular biomarkers for evaluating human exposure. ImagesFigure 4Figure 5Figure 6


Introduction
The heavy metals chromium(VI) and arsenic(III) are considered human carcinogens and share several properties in regard to their carcinogenicity. They are both associated with certain tumors in humans based principally on epidemiologic evidence that is quite strong and that preceded animal carcinogenicity data by many years (1)(2)(3). Both are considered lung carcinogens whose effects are synergistic with cigarette smoking or exposure to other lung carcinogens such as benzo[a]pyrene or other carcinogenic metals. Human exposure to these metals occurs principally in occupational settings and through environmental contaminations such as from toxic waste sites, including numerous Superfund sites, or from naturally contaminated drinking water. Chromium and arsenic are listed as 2 of the 8 heavy metals that are among the top 22 compounds of particular concern to the U.S. Environmental Protection Agency (U.S. EPA) Office of Solid Waste and Emergency Response concerning Superfund sites. Thus, there is now increasing concern about human health effects from exposure to heavy metals found in high concentrations at these sites, such as the high levels of chromium and arsenic at waste sites in Woburn, Massachusetts, and northern New Jersey. However, potential human exposures and health risk from these sites are less well defined. Thus, understanding the overall human health impact of environmental exposure to these heavy metals is of paramount concern. Chromium(VI) is a known human lung carcinogen and may also contribute to nasopharyngeal cancers (1,2,4,5). A strong association between chromium exposure and an increase in lung cancer was first observed in the chrome ore industry in Germany in the 1920s. Since then, there have been numerous studies linking inhalation exposure to chromium, particularly chromium(VI) compounds, with increases in lung and nasal cancers. However, early recognition of this association and significant decreases in exposure through industrial hygiene practices have reduced these incidences. Arsenic is considered to be a probable human lung, skin, and bladder carcinogen (3,6). Lung exposure has been principally through occupational settings in the mining, processing, and smelting of arsenic-containing ores, and in the manufacture of arsenic-containing pesticides, whereas skin exposure has been through many of these same occupational settings as well as topical application of arsenical medications. There is also an association between environmental arsenic exposure, e.g., through arsenic in well water, and an increased risk of certain skin and possibly bladder cancers (3,7). Arsenic exposure has also been implicated in certain other tumor incidences including scrotal cancer (copper smelters) and hepatic angiosarcomas (vineyard workers) (3,7). As with the other carcinogenic metals, there is a strong synergistic association between arsenic exposure and cigarette smoking for risk of lung cancer. However, in the case of arsenic, this is complicated by the fact that there were high levels of arsenic in tobacco prior to the 1970s as a result of use of arsenical pesticides in tobacco agriculture (3). Arsenic has only recently been shown to be carcinogenic in animal models (3).
The form and valence of chromium is a major factor in assessing its biologic effects, and chromium(VI) is the principal form responsible for the mutagenic and carcinogenic properties, as first proposed in the uptake-reduction model by Wetterhahn and co-workers (1,4,8,9). Chromium(VI) has been shown to be genotoxic, mutagenic, and carcinogenic in a large number of short-and long-term assay systems using various end points and systems (1,2,4,8,9). Like many organic carcinogens such as benzo[a]pyrene and aflatoxin B1, chromium(VI) appears to act as a classic initiator in these test systems, and this is believed to be the principal mechanism by which chromium(VI) increases carcinogenic risk. Our laboratory has hypothesized that genotoxic carcinogens, including chromium(VI), exert preferential effects on the expression of a specific dass of genes, i.e., inducible genes, as a result of nonrandom DNA damage targeted to members of this gene class (8)(9)(10)(11)(12). Previous studies in our laboratory have shown that a number of different genotoxic chemical carcinogens that induce different types of DNA damage, including the genotoxic metals chromium and nickel, significantly alter both basal and inducible expression of several model inducible genes but have no effect on expression of constitutive genes. Assuming that the effects on gene expression are a result of carcinogen-induced DNA damage, what is their molecular basis?
We have postulated that inducible genes are strongly affected by chemicallyinduced DNA damage as a result of both targeting of DNA damage and the intrinsic structural and biochemical properties of those genes, and have proposed a model in which these effects occur as a consequence of chromatin structure and/or nuclear architecture (10,13). This nuclear model provides a framework for understanding how two different genes in the same nucleus may reside in very different physical and chemical environments, and may therefore have different sensitivities to chemical attack. Thus, even low levels of DNA damage in vivo might be sufficient to have significant effects on expression of certain sensitive genes. We previously demonstrated that treatment of 14-day chick embryos with a single administration of chromium(VI), at a dose that produced no overt toxicity but that caused significant levels of DNA damage (14), had profound effects on expression of several inducible genes while having no effect on expression of several constitutively expressed genes (10). Both the basal and inducible expression of the inducible 5'-aminolevulinate synthase, cytochrome P450 CYP2HI, and phosphoenolpyruvate carboxykinase (PEPCK) genes were markedly affected by the chromium treatment, whereas the albumin, transferrin, and ,-actin genes were refractory to this treatment. The effects on expression of the inducible genes were seen at both the steady-state mRNA and transcriptional levels, and the time courses for these effects closely matched the time course for chromium-induced DNA damage and repair (10,12,14). Interestingly, certain effects most closely correlated with chromium-DNA monoadduct formation, whereas other effects were more closely associated with chromium-DNA cross-link formation (10,12,14).
In contrast to chromium, the mechanistic basis for the carcinogenic effects of arsenic is still poorly understood. Arsenic has generally been shown to be negative in bacterial and mammalian mutagenesis assays, and there is little evidence of DNA damage induced by arsenic, although it has been demonstrated to be a clastogen and weak inducer of sister chromatid exchanges in some systems (3,6). Arsenic causes morphologic transformation, anchorage independence, aneuploidy, and gene amplification in mammalian cell culture assays (6). These effects may be related to its ability to alter DNA synthesis and DNA repair, as well as fidelity of chromosome segregation. Arsenite [As(III)] is approximately 4-fold more toxic than arsenate [As(V)], although these two forms can be inter-converted by redox pathways in vivo (3). Arsenic(III) is readily methylated in vivo, and the methylated forms have a low interaction with macromolecules and are readily excreted. However, the rate of methylation differs among tissues and species. The primary mechanism of arsenic(V) toxicity is uncoupling of oxidative phosphorylation, by substitution for inorganic phosphate. Arsenic(III) also can cause oxidative bursts in endothelial and other cells, and is heat shock-mimetic, inducing the classic heat shock and stress responses of cells (15). Thus, its biologic effects are more closely associated with cytoplasmic rather than nuclear effects, and DNA does not appear to be the primary target for arsenic toxicity or carcinogenesis.
In this regard arsenic appears to act more as a classic promoter such as the phorbol ester 12-O-tetradecanoylphorbol-13acetate or hydrogen peroxide, although its specific effects may be more complex. Arsenic displays distinct interactions with specific proteins in vivo, which may be the basis for much of its specific toxic effects as well as its carcinogenicity.
Most of the intracellular effects of arsenic are attributed to arsenic(III) rather than arsenic(V). Arsenic(III) binds avidly to thiols in cells, both to small molecule thiols such as glutathione as well as thiolcontaining proteins (3,16,17). Arsenic has a particular avidity for hydrolases and can compete with phosphate to inhibit enzymes such as phosphoraldehyde and glucose dehydrogenase (3). Many of the cellular effects of arsenic can be ascribed to its ability to elicit a heat-shock response (15,16). Arsenic induces phosphorylation of hsp27, increases expression of hsp27, hsp70, and hsp90 through the heat-shock factor, induces heme oxygenase, mdrl, and quinone reductase gene expression, and induces metallothionein expression by both a transcriptional and posttranscriptional mechanism, although arsenic is not a ligand for metallothionein protein binding (3,18,19). Arsenic exhibits a strong preferential binding to the vicinal dithiol of the glucocorticoid receptor, inhibiting binding of glucocorticoid hormone but not altering hsp9O binding (16,17). This effect is extremely specific, as there is not a similar effect on the closely related receptors for androgen, estrogen, mineralocorticoids or progesterone. Because at least one of the model inducible genes we have used in our studies, i.e., PEPCK is regulated by glucocorticoids, we predicted that this induction pathway would also be markedly affected by arsenic treatment. Other effects on PEPCKexpression are also predicted to be likely, as a result of arsenic effects on other Environmental Health Perspectives * Vol 106, Supplement 4 * August 1998 specific response pathways and regulatory proteins within the cell.
Thus, there is reason to believe that both chromium and arsenic can have very selective effects on activation of regulatory pathways and expression of certain genes in vivo, albeit through different mechanisms. We are particularly interested in testing the hypotheses that specific DNA regulatory regions within the promoters of targeted genes are responsible for mediating the cis effects, and that specific transcriptional factors are responsible for mediating the trans effects of these heavy metals on inducible gene expression. Such effects may be very important in understanding the overall effects of these and other heavy metals on the carcinogenic process. The purpose of these studies is to explore these hypotheses in detail.

Materials and Methods
Animal and Cell Treatment and Sample Preparation All chemicals were purchased from Sigma Chemical Co. (St. Louis, MO). Fertile White Leghorn chicken eggs (Truslow Farms, Inc., Chestertown, MD) were incubated as previously described (20). Test compounds were administered in solution onto the inner shell membrane as previously described (21). Chromium(VI) was administered as sodium dichromate (Na2Cr207.2H20) at a dose of 50 pmol/kg in 100 pl H20. Arsenic(III) was administered as sodium arsenite (NaAsO2) at a dose of 100 pmol/kg in 100 pl H20.
Dexamethasone was administered as dexamethasone-21-phosphate in 10 1l H20 at a dose of 5 jmol/kg embryo weight. For steady-state mRNA studies, after embryo treatment the livers were removed and immediately frozen at -75°C. H4IIE rat hepatoma and MDA-MB-435 (MDA) cells were grown and treated essentially as previously described (22). UV-C (ultraviolet C) doses (254 nm) were administered to cells in 150-mm cell culture dishes with a UV Stratalinker (Stratagene, Lajolla, CA) or using a calibrated germicidal lamp at the desired dose. Total cellular RNA was subsequently isolated using a guanidine isothiocyanate-cesium chloride gradient technique exactly as previously described (20). For the electrophoretic mobility shift assays (EMSA), cells were washed with icecold phosphate-buffered saline (pH 7.4), containing 1.0 mM EDTA, protease inhibitor mix (4 pg/ml leupeptin; 4 pg/ml soybean trypsin inhibitor; 1 mM benzamidine), 7 pg/MI L-1-p-tosylamino-2phenylethyl chloromethyl ketone, and 0.1 mM dithiothreitol (DTT). Nuclear protein soluble extract was isolated by a procedure of Dignam et al. (23), as modified by Barchowsky (24).

Measurement ofmRNA Levels and EMSAAnalysis
Steady-state mRNA levels were measured by a quantitative solution hybridization assay using gene-specific 5'-[32P]-endlabeled synthetic oligonucleotide probes, exactly as previously described (20). This assay has been shown to provide a sensitive, linear, and quantitative measure of specific steady-state mRNA levels (fmol of mRNA per milligram of total RNA), and can accurately determine differences in mRNA expression of less than 20% (10,20,25,26).

Generation and Analysis of PEPCK-Luciferase Genetic Constructs
All DNA-modifying enzymes and lipofectamine were purchased from GIBCO-BRL.
The pGL3 vectors and the luciferase assay system were purchased from Promega. The PEPCK-luciferase reporter vectors were constructed by cloning a region of the PEPCKpromoter into the multiple cloning site of the pGL3 basic vector. Rat genomic DNA from H4IIE cells was isolated and used as a template for the PCR ampliflcation of the PEPCK promoter. The sequences for the upper (forward) primers were 5'-CTAGCTAGCCAATCACCCCT (NheI, -592 of the rat PEPCKpromoter region) and 5'-CTAGCTAGCCATGT CCCTGCCCCC (NheI, -1162 of the rat PEPCKpromoter region); and the sequence of the lower (reverse) primer was 5'-GGAAGATCTCTACCTTTCTT (Bgl II, +87 of the rat PEPCKpromoter region). All plasmids used were grown in Escherichia coli JM109 strain. The PEPCKpromoter fragments were digested with NheI and BgIII enzymes and ligated into pGL3. Plasmids containing inserts were sequenced with an ABI PRISM DyeDeoxy Terminator Cycle Sequencing kit (Applied Biosystems, Foster City, CA). All sequences were checked against GenBank PEPCKsequences using the Blast sequence analysis program to ensure sequence specificity. Plasmid DNA was purified using Qiagen columns (Qiagen, Valencia, CA) and transfected into H4IIE rat hepatoma cells using the lipofectamine technique (GIBCO-BRL). To obtain stable transfectants, cells were cotransfected with two types of plasmids: pGL3 reporter vectors with different promoters plus pSV2neo that confers resistance to neomycin. Cells were selected in G418 and resistant colonies were subcloned and checked for PEPCK-luciferase insertion. For the luciferase assay, cells were lysed in the luciferase reporter buffer in the wells according manufacturer's instructions (Luciferase assay system, Promega). Luciferase activity was measured using a 96-well plate luminometer (Dynatech Laboratories, Chantilly, VA). Protein concentrations were determined by a standard absorbance DC (detergent compatible) protein assay (Bio-Rad, Hercules, CA).

Statistica Analysis
Data were analyzed for statistical significance by ANOVA, Student's t-test, and nonparametric analysis, where appropriate, using the Instat and Prism software programs (GraphPad Software, San Diego, CA).

Results
Initial experiments were conducted to examine whether there were preferential effects of chromium and arsenic on expression of a model-inducible gene, PEPCK, in 14-day chick embryo livers. The chick embryo represents a simple, inexpensive, and easily manipulated in vivo model and has been an excellent system for examining effects of carcinogens on gene expression Environmental Health Perspectives * Vol 106, Supplement 4 * August 1998 based on our previous studies (10)(11)(12)(13)(34)(35)(36). Chick embryo development has been extensively characterized, providing a large background of information about its embryology, physiology, and biochemistry. This model has been used to investigate the genotoxicity of over 50 chemicals using a wide variety of genetic end points, including chromosomal damage, sister chromatid exchange, inhibition of DNA synthesis, unscheduled DNA synthesis, carcinogen-DNA binding, and DNA alkaline elution (10,14,21,(37)(38)(39). The chick embryo also possesses an active, highly inducible hepatic mixed-function oxidase enzyme system capable of metabolizing indirectacting mutagen-carcinogens to active forms (20,21,26,37,38,40). The chick embryo can also be used to specifically examine the developmental effects of agents on an embryonic system and mimics the human fetus in many important aspects with regard to its toxicology (21,37,38,40,41). Thus, this system represents an excellent nonmammalian whole animal toxicology model.
The highest doses of chromium(VI) and arsenic(III) that caused no overt toxicity or lethality in dose-response experiments were determined to be 50 imol/kg and 100 pmol/kg, respectively (10,12,42). At these doses there is little or no effect on overall DNA, RNA, or protein synthesis, and the embryos can be hatched normally with no obvious toxic effects (10,12,42). but had no effect on expression of ,-actin ( Figure 1C). Chromium and arsenic also significantly altered the response of the PEPCKgene to glucocorticoid induction ( Figure 1A,B). Chromium increased the induction response of PEPCKto glucocorticoids initially (1-2 hr) but at later times almost completely suppressed the response of PEPCKto its normal induction signal ( Figure IA). Arsenic had similar effects, although the magnitude of the response was greater and occurred over a more protracted time than the response to chromium ( Figure 1B).
These effects were then examined in the H4IIE rat hepatoma cell line to investigate the molecular basis for this phenomenon. This cell line expresses PEPCKin a basal and hormone-inducible manner similar to that of the liver in vivo, and has been used extensively to examine PEPCKregulation [reviewed by Granner et al. (43) and Lucas and Granner (44)]. Initial experiments established the toxicity dose-response to chromium and arsenic using a colonyforming assay as a measure of cell survival ( Figure 2). These experiments established the maximal noncytotoxic doses as well as the minimal doses which produced

E°-
Time after chromium(VI), hr complete cell death. These were used in subsequent experiments to examine responses to low (i.e., noncytotoxic) versus high (i.e., cytotoxic) doses of heavy metals. Using these low doses, the effects of chromium and arsenic on PEPCK expression were examined in the H4IIE cells, and compared to the effects of mitomycin C (MMC), an organic DNA cross-linking agent with similar genotoxic properties as chromium(VI). Figure 3 shows that chromium and arsenic significantly suppressed basal expression of PEPCK in a manner similar to MMC. We had previously shown that chromium, arsenic, and MMC also suppressed expression in these cells of the inducible gene mdrl, which codes for P-glycoprotein, but has no effect on several constitutive genes (22,42). Thus, this appeared to be a good model for examining in detail the mechanism by which these metals selectively affect inducible gene expression in cells.
We then examined the effects of chromium and arsenic on the nudear levels ofvarious transcription factors that might be associated with such changes in gene expression. It has been postulated that toxins can alter gene expression by activation of cell-signaling pathways leading to increases in nuclear transcription factor levels and/or activities leading to altered mRNA expression. Cells were treated with low and high doses of metals and nudear extracts were isolated. Transcription factor binding was determined using EMSA analysis. An example of data generated in these studies is shown in Figure 4, which demonstrates a significant effect of a 1-hr noncytotoxic chromium treatment on nuclear binding levels of the transcription factor Spl . Using this approach, we then examined in detail the effects of chromium and arsenic on nuclear levels of AP-1, NF-KB, and Spl transcription factors, since these factors have been implicated in the regulation of both PEPCK and mdrl genes, which are affected by these agents (3,43,(45)(46)(47)(48)(49). As summarized in Table 1, there were specific effects of each agent on transcription factor binding that were agent-, dose-, factor-and cell linedependent. Both chromium and arsenic generally caused significant increases in AP-1 binding in both cell lines at both doses, with the exception that the high dose of chromium suppressed AP-1 binding in the MDA cells. In contrast, the effects on Spl binding were quite variable. Chromium significantly increased Spl binding in the MDA cells, whereas it had no effect on SpI binding levels in the H4IIE cells. Conversely, arsenic had no effect on Spl binding in the MDA cells, whereas it decreased Spl in the H4IIE cells at the high dose. Metal effects on NF-KB binding were very cell-type specific. Arsenic had little or no effect on NF-icB binding in the MDA cells, whereas chromium increased binding, and neither agent had any effect on NF-KB in the H4IIE cells. Taken together, these results suggest that effects of chromium and arsenic on nudear levels of specific transcription factors may be a component of tissue-specific alterations in gene expression caused by these metals. These effects may contribute to the overall biologic effects of these metals in humans and may play a role in long-term effects oflow-dose exposures such as occur in metal-induced carcinogenesis.
Previously, it had been demonstrated that high-dose UV or y irradiation as well as treatment with certain chemical carcinogens can stimulate a characteristic response pattern in cells that has been called the UV response (50)(51)(52). Since activation of certain transcription factors, particularly AP-1 and NF-iKB, has been shown to be a component of the UV response pathway were treated with 2 pM chromium(VI) for the indicated times, and total nuclear protein extracts were isolated as described in "Materials and Methods." Electrophoretic mobility shift analysis was performed using a radiolabeled duplex oligonucleotide fragment corresponding to the Spl recognition sequence from the human MDR1 promoter region and levels of Spl binding (arrow) were analyzed as described in "Materials and Methods." Blank, no nuclear extract added; 0 hr control, extract from cells treated with solvent alone.     (50,51,53,54), we investigated in more detail whether this could be contributing to effects of metals on gene expression in our experiments. Table 2 summarizes the results of studies examining the effects of chromium and arsenic on hallmarks of the UV response pathway, i.e., mRNA expression of c-fos and c-jun (components of the AP-1 transcription factor), and gadd45 and gadedl53 (growth arrest-and DNA damageinducible genes), and increases in nuclear levels of AP-1 and NF-iKB. This was compared to effects of UV, tumor necrosis factor a (TNF-a), and y-irradiation on these parameters and on expression of PEPCKand mdrl mRNA. Although UV, tumor necrosis factor [3 (TNF[3), and y-irradiation induced most or all of the hallmarks of the UV response in these cells, chromium, arsenic, and MMC did not alter the majority of these parameters. The exception was an increase in nuclear AP-1 levels by both metals, but this response was not accompanied by the predicted increases in c-fos or cjun mRNA expression. In contrast, chromium, arsenic, and MMC all altered mRNA expression of PEPCK and mdrl, whereas UV, TNFa, andy-irradiation did not. These results demonstrate conclusively that the effects of noncytotoxic doses of chromium and arsenic (and MMC) on gene expression do not involve activation of the UV response pathway per se. We then examined whether specific genetic elements within the promoters of target inducible genes could be responsible for mediating the effects of chromium and arsenic on their expression. Constructs consisting of portions of the rat PEPCKgene promoter were fused with the luciferase reporter gene and transfected into H4IIE cells to generate stably transfected cell lines with integrated copies of the transgene in their genome. The PEPCK-luciferase gene was basally expressed and fully hormone responsive in these cell lines in a manner similar to that of the native PEPCKgene ( Figure 5A,B). In addition, the constructs containing either 1.2 or 0.6 kb of normal rat PEPCKpromoter were fully responsive to chromium, arsenic, and MMC, suggesting that this region of the PEPCK promoter is sufficient to confer carcinogen sensitivity to another gene and therefore contains a target(s) for their effects. It further suggests that integration into nuclear chromatin structure is important for eliciting these effects. Interestingly, the two wild-type constructs differed qualitatively from each other in their basal response. The 1.2-kb construct exhibited a significant decrease, whereas the 0.6-kb construct exhibited a significant increase in PEPCK-luciferase expression in response to each of the three agents ( Figure 5A). This suggests that an element(s) between -0.6 and -1.2 kb of the PEPCK promoter can contribute qualitatively to the basal response of this promoter to these agents. It had previously been shown that the proximal 0.6 kb of the rat PEPCK promoter was sufficient to confer normal hormone responsiveness to heterologous constructs (43). Both the 1.2-and 0.6-kb construct cell lines exhibited a normal dexamethasone response ( Figure 5B), and both constructs also exhibited a complete inhibition of their dexamethasone response following treatment with each of the three agents ( Figure 5B).
Constructs were then transfected with a 1.2-kb portion of the PEPCKpromoter in which either the two tandem glucocorticoid response elements or the two adjacent accessory factor elements had been mutated using a site-directed mutagenesis technique that leaves the surrounding sequences intact. Mutation of these elements by deletion had previously been shown to significantly decrease or abolish ND, not determined. "Cells were treated and analyzed as described in "Materials and Methods" and in Figure 4 and Table 1 hormone-inducible response of this promoter (43,44). As expected, although these constructs displayed a normal basal expression ( Figure 5A), they had a significantly reduced responsiveness to induction by dexamethasone ( Figure 5B). These constructs did not show any effect of treatment with chromium, arsenic, or MMC followed by dexamethasone ( Figure 5B). Interestingly, their basal expression was also observed to be completely refractory to Alternatively, cells were treated with chromium, arsenic, or MMC, and beginning 4 hr later, cells were treated with 0.1 pM dexamethasone for 20 hr to measure effects on hormone-inducible PEPCK-luciferase expression (B). Data are expressed as a percent of the basal control values. Each bar represents the mean ± SD of values from three individual wells.
Environmental Health Perspectives * Vol 106, Supplement 4 * August 1998 chromium, arsenic, or MMC treatments ( Figure 5A). These results suggest that the responsiveness of the PEPCKgene to glucocorticoid induction is an important component of its ability to respond to chromium and arsenic treatments at both the basal and hormone-inducible level.

Discussion
The principal purpose of these studies was to develop systems to enable us to determine at the molecular level the basis for the preferential effects of the carcinogenic heavy metals chromium(VI) and arsenic(III) on inducible gene expression. We had previously examined the preferential effects of 16 different chemical carcinogens, induding the genotoxic metals chromium(VI) and cisplatin, and other organic genotoxic agents that induce different types of DNA damage, on the expression of several different model inducible genes. Those studies demonstrated that treatment of 14-day chick embryos with a single administration of chromium(VI), at a dose that produced no overt toxicity but which caused significant levels of DNA damage (14), had profound effects on expression of several inducible genes but had no effect on expression of several constitutively expressed genes (10). Both the basal and inducible expression of the inducible ALA synthase, CYP2H1 and PEPCKgenes were markedly affected by the chromium treatment, whereas the albumin, transferrin, and 3-actin genes were refractory to this treatment. The effects on expression of the inducible genes were seen at both the steady-state mRNA and transcriptional levels, and the time courses for these effects closely matched the time course for chromium-induced DNA damage and repair (10,12,14). Interestingly, certain effects most closely correlated with chromium-DNA monoadduct formation, whereas other effects were more closely associated with chromium-DNA crosslink formation (10,12,14). This same general phenomenon, i.e., preferential effects on inducible genes and correlation with DNA damage and repair, has also been observed in this system with a large number of other genotoxic and carcinogenic agents. These include the cross-linking agents cisplatin (11), and MMC ( (13); a chemical that induces strand breaks, bleomycin (11); and several synthetic acridine-based mono-and bis-intercalating agents (42).
The results of many of these studies have recently been summarized (11). The preferential effects of each of these agents strongly correlated with DNA damage, which supports the general hypothesis that alterations in gene expression can be used as a marker for DNA damage in vivo. Although the inducible genes were all responsive to DNA-damaging agents, the time course, direction, and magnitude of each response was both agent specific and gene specific. However, similarities in specific effects among groups of agents were based on the type of DNA damage induced, rather than the chemical structure, metabolism, lipophilicity, etc., of the agents themselves, which lends further support to the hypothesis that induction of DNA damage was the basis for the preferential effects on inducible gene expression. Others have also examined effects on gene expression by some of these agents as well as other carcinogens (55-58) that cause a variety of DNA lesions, including simple alkylations, bulky lesions, and cross-links. All of the carcinogens tested to date have been demonstrated to selectively alter inducible gene expression. We and others (55-57) have examined a number of different inducible genes; all responded to carcinogen treatment. In contrast, none of the constitutively expressed genes tested (p-actin, transferrin, albumin, and ax-tubulin) were responsive. Effects on inducible gene expression have been observed in both the rat and chick embryo in vivo (10,13,34,55,59) and in primary chick embryo, adult rat, and rat embryo hepatocytes and rat hepatoma cell lines in culture (56)(57)(58). All of these systems show similar responses. Thus, carcinogen-induced alterations in inducible gene expression are a general phenomenon, and responsiveness appears to be independent of the specific system, carcinogen, gene, or induction pathway studied.
In contrast to these organic and inorganic carcinogens that are believed to principally target DNA as their mechanism of action, the mechanism of action of nongenotoxic carcinogens such as the heavy metals arsenic and cadmium is not well understood. However, it is known that these agents interact with specific subsets of proteins, principally through sulfhydryl interactions, and it has been postulated that they may act through epigenetic mechanisms to act as carcinogens.
We had hypothesized that arsenic may also specifically and preferentially alter inducible gene expression much like chromium, but through a different mechanism than that of chromium or the other genotoxic agents. The goal of these studies was to examine this hypothesis in detail. Our results indicate that chromium and arsenic both preferentially alter inducible gene expression, and in a similar manner. Each agent had significant effects on basal and hormoneinducible expression of a model inducible gene, PEPCK, in both the chick embryo in vivo system and in rat hepatoma H4IIE cells in culture. Each agent also altered expression of PEPCK-luciferase genetic constructs in which their effects on basal and inducible expression were strongly correlated with the responsiveness of the transgene to glucocorticoids, suggesting that this pathway is important for the response and may be the primary target. These results indicate that this genetic approach will be useful for identifying specific regulatory pathways that mediate these responses.
It is interesting to note that the effects of chromium and arsenic on the integrated constructs containing 1.2 kb of promoter region were very similar to those on the native rat PEPCKgene in the H4IIE cells, i.e., a suppression in both basal and hormone-inducible expression, suggesting that this region of the promoter is sufficient to replicate the metal effects on the normal gene. The effects of chromium and arsenic on the 0.6 kb construct, in contrast, were more similar to those seen for PEPCKin the chick embryo liver, i.e., an increase in basal expression and a decrease in hormone responsiveness, suggesting that the distal 0.6 kb provided additional elements that qualitatively change the response of the rat PEPCKgene to these agents. A complete loss of basal responsiveness of the rat PEPCKgene promoter to chromium and arsenic was observed in mutated constructs in which there was also a loss of hormone responsiveness. This is very similar to what had previously been observed in the chick embryo system, in which the responsiveness of the native chicken PEPCKgene was intimately associated with its response to dexamethasone (12). In that study, we showed that the PEPCKgene lost its chromium sensitivity concomitant with the normal developmental loss of liver PEPCKhormone responsiveness that occurs between 14 and 17 days of embryonic development (12,60), even though the liver PEPCK gene was still basally expressed and still Environmental Health Perspectives * Vol 106, Supplement 4 * August 1998 responsive to induction by cAMP at 17 days (12). In addition, we have recently observed over that same developmental time period that the kidney PEPCKgene gained hormone responsiveness, and concomitantly gained sensitivity to chromium treatment (42). However, in a similar developmental study, we observed that the chick embryo liver PEPCK gene was refractory to the effects of MMC at 8 days of development even though it was fully glucocorticoid inducible (61). Thus, hormone responsiveness appears to be necessary but insufficient by itself to elicit a carcinogen response in this gene. In that study, the PEPCKgene became MMC sensitive by day 10 of embryonic development, concomitant with a major change in the chromatin structure of the PEPCK gene promoter region (61). These results suggest that both glucocorticoid responsiveness and chromatin structure are critical for the selective responsiveness of the PEPCKgene to carcinogen treatment.
Chromium and arsenic also altered nuclear transcription factor binding, although the patterns were different for the two agents and were also dose-and cell-type specific. Previous studies by Fornace and coworkers (52,53,62) and others (50,51,54) reported that treatment of cells with high doses of UV irradiation or other cytotoxic agents causes a characteristic UV response in cells. The response to UV irradiation itself was best characterized and involved a rapid increase in expression of mRNAs for c-fos and c-jun (components of AP-1), growth arrest and DNA damage-inducible (gada) genes, especially gadd45 and gaddl53, as well as other genes (53,54,63). There was also a rapid increase in binding ofAP-1 and NF-KB to their target DNA regulatory sequences (50,51). Other agents that were reported to mimic aspects of this response included X-irradiation, oxidative stress (e.g., by hydrogen peroxide), and certain DNAdamaging agents, including MMS, cisplatin, and MMC, although only at high, i.e., overtly cytotoxic, doses. We investigated whether our treatments stimulated aspects of the UV response as a number of inducible genes have been shown to be regulated by AP-1 and/or NF-icB, including PEPCK(64) and mdrl (65) that have been shown to be sensitive to chromium and arsenic in our studies. However, our results clearly indicate that noncytotoxic doses of chromium and arsenic are not acting through the mammalian UV response pathway to exert their preferential effects on inducible gene expression.
We observed a significant increase in both AP-1 and Spl binding by low-dose chromium treatments, whereas arsenic increased AP-1 but had little or no effect on Spl at low doses. Chromium significantly increased NF-icB binding and arsenic induced modest increases in NF-iKB at low doses in the MDA cells, although neither agent had a significant effect on NF-icB binding at low or high doses in the H4IIE cell line. However, chromium and arsenic had no effect on other hallmarks of the UV response pathway. In addition, positive controls such as UV or TNFa that stimulate the UV response did not alter PEPCK expression. Thus, effects of chromium and arsenic on the UV response may involve apoptotic or other mechanisms that occur only at very high doses and represent late events, but which are separate from the pathways leading to alterations in gene expression that occur at lower doses. The best chemical DNA-damaging agent to induce the UV response to date is the simple direct-acting alkylating agent MMS (66,67). However, in addition to DNA damage, MMS also induces large amounts of protein adducts including protein crosslinks (68), suggesting a possible alternative mechanism for stimulation of the UV pathway (69,70). It is possible that chemical stimulation of the UV pathway requires a certain threshold of nonspecific protein adducts that can only be obtained at very high doses of these agents but that are not critical to the carcinogenicity of these agents at lower doses. Similarly, arsenic has been shown to be heat-shock mimetic, although this also occurs at much higher doses than those which led to alterations in PEPCKgene expression in this study. However, lower doses did not significantly alter heat-shock protein expression in preliminary studies (42). Thus, it is also unlikely that this pathway is the principal mediator of these effects at the lower doses. Figure 6 summarizes our current model for the actions of chromium and arsenic on PEPCKgene expression. Inducible gene expression ultimately involves the convergence of cell signaling pathways with DNA-chromatin at the level of the gene promoter. Thus, it is possible that two different agents such as chromium and arsenic, which may act at different points of the same pathway, might lead to similar effects on expression of a targeted inducible gene. Chromium is proposed to act principally through its action as a DNA-damaging agent. We hypothesize that chromiuminduced DNA lesions lead to alterations in and is metabolized through a variety of small molecule and enzyme-reductive pathways to Cr(V), Cr(IV), and Cr(lIl) species that can interact directly with DNA-chromatin to form monoadducts, DNA interstrand, intrastrand and DNA-protein cross-links, and other forms of DNA damage. These intermediates may also activate or repress specific transcription factors either directly or indirectly through signaling pathways. Arsenic(lil) can also enter the cell and activate transcription factors directly or indirectly. Arsenic does not directly interact with DNA but may also indirectly alter chromatin structure through interaction with chromosomal proteins. These chromium and arsenic effects on transcription factors and DNA-chromatin ultimately converge at the promoters of targeted inducible genes such as PEPCK, causing alterations in specific DNA-protein interactions that ultimately lead to changes in transcription and gene product expression. specific DNA-protein interactions within the chromatin of an inducible gene promoter, leading to changes in transcription of that gene. Inducible genes such as PEPCKmay be preferentially susceptible to these effects because of their large DNasehypersensitive regions which represent areas of nonnucleosomal decondensed chromatin and regions of complex DNA-transcription factor interactions. In addition, chromium also has significant effects on nuclear levels of specific transcription factors, and these are likely to contribute to the overall effects of chromium on inducible gene expression. We further propose that arsenic acts principally through direct or indirect effects on specific transcription factors and other signaling pathways rather than on DNA per se. Arsenic may also be able to affect chromatin structure within an inducible gene promoter through mechanisms involving alterations in histone and/or nonhistone chromosomal proteins. Collectively, these effects may ultimately lead to alterations in specific DNA-protein interactions within Environmental Health Perspectives * Vol 106, Supplement 4 * August 1998 5i inducible gene promoters in a manner similar to that of chromium. We have demonstrated that chromium and arsenic have similar effects on expression of the hormone-regulated PEPCKgene, both in the chick embryo in vivo and in the H4IIE cell line in culture. These effects are closely correlated with glucocorticoid regulation of this gene through the glucocorticoid receptor pathway and its DNA recognition elements within the PEPCKpromoter region. Examining the details of how chromium and arsenic specifically alter this regulatory pathway will be the focus of future studies.
In summary, we have examined the mechanistic basis for the strong preferential effects of the heavy metals and human carcinogens chromium(VI) and arsenic(III) on inducible gene expression. Our studies indicate that each of these agents has spe-cific effects on transcription factor binding and expression of certain genetic constructs that may provide insight into the cell signaling and gene regulatory pathways that mediate their carcinogenic effects in vivo. A combination of both cis effects involving specific regions of the PEPCK promoter and their chromatin structure, and trans effects involving effects on specific transcription factors appear to work in concert to contribute to the overall effects of these heavy metals on inducible gene expression. Determining the mechanisms by which these carcinogenic metals may selectively alter gene expression would have important implications for understanding the molecular basis for the impact of these agents on the carcinogenic process and overall human cancer incidence in exposed populations. Although each of these agents may have specific effects and act through independent mechanisms, they may also have profound but different effects when present in different combinations in the environment. Understanding these interactions at the molecular level is critical for an accurate assessment of the overall health effects of these substances on the human population. Elucidating these molecular events may also eventually provide sensitive biomarkers for evaluating human exposures, for example, by defining a specific transcription factor, signaling pathway, or sentinel gene that might be indicative of prior arsenic or chromium exposure, much as metallothionein expression can serve as an indicator of cadmium exposure. This will first require a more detailed understanding of the basis for these effects.