Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

In the past several years, novel systems (particularly cDNA arrays) have been developed to exploit the wealth of sequence information available for many organisms, including the human. The ability of microarrays to monitor the expression of thousands of genes simultaneously is well established (Schena et al., 1998; Debouck and Goodfellow, 1999; Khan et al., 1999). This tool potentially offers unprecedented power for use in toxicological studies and there has been substantial interest in using DNA arrays as a tool to quickly screen and categorize chemicals (Medlin, 1999; Nuwaaysir et al., 1999; Rockett and Dix, 1999; Lovett, 2000). Recently, the potential value of arrays for screening was made apparent in cancer diagnosis (Golub et al., 1999). To date, however, the applications of the technology to issues in toxicology, while well described in concept, are "few" in practice.

The experiments described in this manuscript build on previous work (Bartosiewicz et al., 2000, 2001) to explore the ability of arrays in categorizing toxicants according to class and to examine the effects of time, dose, and tissue on the pattern of gene expression. Ten chemicals (ENU, DMN, CdCl₂, HgCl₂, clofibrate, DEHP, 3-MC, BaP, CCl₄, and bromobenzene) were selected for this study. The chemicals were selected based on extensive studies of their mechanisms of action as well as, in several cases, their environmental importance. Table 1 lists each of the chemicals, the doses used in this study, and a summary of previously reported effects in the liver and kidney. These chemicals each belong to five separate classes and are paired according to their mechanism of action. ENU and DMN are direct and indirect DNA alkylators, respectively, and in yeast are known to induce a number of DNA repair enzymes, stress proteins, and metallothionein (Bhattacharjee et al., 1998; Jelinsky and Samson, 1999; Krems and Culotta, 1999). CdCl₂ and HgCl₂ are heavy metals that both induce metallothionein as well as numerous other genes (Beyersmann and Hechtenberg, 1997). CCl₄ and bromobenzene are metabolized to reactive electrophiles; cytotoxicity from these is associated with binding of reactive metabolites to protein and/or lipid peroxidation (Recknagel et al., 1989; Lau and Monk, 1993). Like the heavy metals and DNA alkylators, these compounds induce stress proteins and metallothionein as well as a number of antioxidant genes, including -glutamylcysteine synthetase and glutathione reductase (Schiaffonati and Tiberio, 1997; Serfas et al., 1997; Dalton et al., 1999). DEHP and clofibrate bind to the PPAR and cause a pleiotropic response involving the induction of a number of proteins involved in fatty acid -oxidation, including Cyp4a10, fatty acid binding proteins, and acyl-CoA thioesterase (Lindquist et al., 1998; Lapinskas and Corton, 1999). The polyaromatic hydrocarbons BaP and 3-MC act through the aryl hydrocarbon receptor and are known to induce Cyp1a1, 1a2, and 1b1 (Schmidt and Bradfield, 1996; Denison and Heath-Pagliuso, 1998).

Previous work (Bartosiewicz et al., 2000) focused on the validation of the use of microarrays as a means of examining gene expression in an in vivo model. A range of doses of -naphthoflavone, a known inducer of Cyp1a1 and 1a2 genes in murine liver, was used to demonstrate that the sensitivity and dynamic range of microarray analysis was at least equal to, and in many cases better than, Northern blot assays. Assessment of slide-to-slide, spot-to-spot, and animal-to-animal variability indicate that the assays are capable of detecting less than a 2-fold change in gene expression and that the greatest contribution to variability is associated with animal-to-animal differences. A second study using three chemicals (trichloroethylene, BaP, and CdCl₂) of different classes examined the feasibility of using arrays in whole animal studies to classify toxicants based on unique gene expression signatures (Bartosiewicz et al., 2001).

The current work expands these studies to show that even with a relatively limited gene set, the expression patterns associated with chemical exposure result in unique patterns of gene expression that are class specific. Proper characterization of these requires that multiple doses, times, and tissues be assessed.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Earlier applications of DNA arrays in toxicology have been published by our group (Bartosiewicz et al., 2000, 2001). Since the methodology used in the current studies is, except as noted, identical to that used earlier, the methods will be described only briefly.

Animals. All experiments involving animals were approved by the Animal Use and Care Committee at University of California, Davis. Male Swiss-Webster mice (20-25 g) were obtained from Charles River Breeding Laboratories (Hollister, CA) and were housed in animal care facilities at University of California, Davis, which are accredited by the American Association of Laboratory Animal Care. Mice were provided food and water ad libitum and were housed in HEPA-filtered racks for at least 1 week before use.

Animal Treatment and Isolation of mRNA. Chemicals were purchased from Sigma Chemical Co., St. Louis, MO [ethylnitrosourea, diethylhexylphthalate, bromobenzene, 3-methylcholanthrene, benzo(a)pyrene]; Aldrich Chemical Co., Milwaukee, WI (dimethylnitrosamine, clofibrate, carbon tetrachloride, cadmium chloride); or Mallinckrodt Chemical Co., St. Louis, MO (mercury chloride). All chemicals were dissolved in either isotonic saline (CdCl₂, HgCl₂, ethylnitrosourea, dimethylnitrosamine) or corn oil (clofibrate, diethylhexylphthalate, BaP, 3-MC, CCl₄, bromobenzene) and administered intraperitoneally to mice (3-4/group). The volume of vehicle administered was 10 ml/kg. Due to the instability of ENU, this compound was dissolved in saline and used immediately. Doses of each toxicant are indicated in the figures. Controls received either isotonic saline or corn oil only. Mice were killed with an overdose of pentobarbital at times specified in the figure legends; livers and kidneys were homogenized immediately in Triazol (Life Technologies, Bethesda, MD) for RNA isolation. DNase-treated total RNA (500 µg) was used to isolate mRNA using Oligotex dt resin (Qiagen, Valencia, CA) following the manufacturer's instructions.

Preparation of DNA "Chips". Chips containing 260 unique genes involved in phase I and II metabolism, heat shock, DNA repair, inflammation, transcription, and housekeeping were used in the current studies. An updated list of the genes present on the slides is available on the following Web site: http://faculty.vetmed.ucdavis.edu/faculty/arbuckpitt/BuckpittLab/genetemplate.htm. The methods for preparation and analysis of the slides, along with applications to whole animal studies, have been presented in a previous publication (Bartosiewicz et al., 2000) and will be described only briefly here. cDNAs corresponding to the 3' region of the respective genes were purchased from the Image Consortium cDNA mouse libraries through suppliers (Research Genetics, Huntsville, AL, or American Type Culture Collection, Bethesda, MD). Additional cDNAs were obtained from Russell Thomas at the McArdle Laboratory for Cancer Research, Madison, WI. All products were verified as described in detail earlier (Bartosiewicz et al., 2000) and varied in size from 500 to 1200 base pairs. Products used for spotting were generated by PCR and purified using a PCR purification kit (Qiagen). Each purified PCR product (10 µl at concentrations >200 ng/µl) was mixed with 10 µl of 8 M NaSCN in 96-well plates, and the DNA was spotted using the Molecular Dynamics Microarray Spotter Gen III, onto slides treated by vapor deposition of 3-aminopropyltrimethoxysilane. Each gene was spotted in eight separate quadrants of the slide.

Preparation of cDNA Probes Labeled with Cy3 or Cy5 and Hybridization. Cy3 and Cy5 labeled probes were prepared using 1 µg of mRNA from control or treated animals, respectively. Synthesis of the labeled first strand was conducted using oligo dt plus random hexamer primers and Superscript II; template RNA was removed by incubation with RNase. Single-stranded cDNA probes were purified using a PCR purification kit (Qiagen). Probe mixtures were evaporated in a vacuum centrifuge. Poly A (Amersham Life Sciences, Arlington Heights, IL) and mouse Cot DNA (Life Technologies) were added to the hybridization solution at a final concentration of 16 and 40 ng/µl, respectively. Probes were resuspended in hybridization solution containing 50% formamide, 5× SSC, and 0.1% SDS. Cy3 (control) and Cy5 (treated) probes were mixed, the probe mixture was then placed on a previously prepared array, and the slides were hybridized overnight at 42°C in a humid chamber. Following hybridization, slides were placed in a wash solution (2× SSC, 0.1% SDS) for 5 min at 37°C with gentle shaking. Slides were then washed once in 0.1× SSC, 0.1% SDS at room temperature for 5 min, and twice in 0.1× SSC at room temperature for a total time of 5 min. The slides were then rinsed briefly in water and dried with a gentle stream of nitrogen.

Analysis of Fluorescence Spots. Slides were scanned on the Molecular Dynamics MicroArray Scanner. Both Cy3 and Cy5 channels were scanned at a photomultiplier tube setting of 750 V. Analysis of the data sets was performed using ArrayVision Software (Imaging Research, St. Catherines, ON, Canada). The fluorescence intensity of each spot was calculated using local median background subtraction. The relative fluorescent units (RFUs) were then normalized to the median signal of probe (Cy3 and Cy5) for that slide. The change in gene expression for each spot was calculated as a ratio of RFU_treated/RFU_control. Median spot values for all eight spots were used to calculate the data. The data on those genes up-regulated or down-regulated by chemical treatment are presented as the percentage of the Cy-3 controls. Only genes whose expression was above 2-fold induction or below 2-fold repression compared with controls were considered for statistical analysis. Significant differences between control and treatment doses were determined using a two-tailed t test. In some cases, a Mann-Whitney Rank Sum test was performed if the t test criteria for normally distributed population was not met. P values <0.05 were considered statistically significant.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Altered Gene Expression by DEHP and Clofibrate in Liver and Kidney. Three genes were significantly induced over control in liver by intraperitoneal injection of both clofibrate (500 mg/kg) and DEHP (1100 mg/kg) (Fig. 1 left). Two of the genes (acyl-CoA thioesterase and Cyp4a10) are involved in -oxidation, whereas the third gene, insulin growth factor-binding protein modulates the activity of insulin-like growth factor (Schuller et al., 1994). The expression of all three genes was increased 2.5- to 3-fold in the liver in response to DEHP. The induction of these three genes was greater (at least 20-fold at optimal time points) in the clofibrate-treated animals. In addition, clofibrate also significantly induced Cyp2b9, a fatty acid binding protein (-oxidation), and metallothionein II by 3.5- to 7-fold. In the kidney, only Cyp4a10 was significantly induced by both peroxisome proliferators (Fig. 1B). Changes in expression levels in kidney were less than in liver. In addition to Cyp4a10, clofibrate also induced acyl-CoA thioesterase, igf-binding protein-1, and the metallothioneins, whereas DEHP repressed aldehyde dehydrogenase and induced fatty acid binding protein in kidney. Maximal induction in both tissues with both peroxisome proliferators occurred 4 to 8 h after treatment.

View larger version (163K): [in this window] [in a new window]   Fig. 1.   Up-/down-regulation of genes in liver (left) and kidney (right) at 4, 8, 24, or 48 h after i.p. administration of DEHP, clofibrate, 3-MC, BaP, DMN, or ENU to mice. Values in boxes are the means for three to four animals and are presented as percentage of control. Values that are colored represent genes that are significantly (P < 0.05) up-/down-regulated at one or several time points and color denotes the degree of induction/repression. Gray shading is used to denote those genes where expression was significantly up-/down-regulated at at least one time point. *Indicates genes uniquely altered by that specific compound or class of compounds.

Induction and Repression of Genes in Liver and Kidney by 3-MC and BaP. 3-MC and BaP were administered intraperitoneally at doses of 80 and 100 mg/kg, respectively. Both chemicals significantly induced GST Mu, and Cyp1a1 and 1a2 in the liver (Fig. 1 left). 3-MC also induced several acute phase proteins, whereas BaP induced igf-binding protein-6, -glutamylcysteine synthetase, and repressed Mx-1, an interferon-induced protein (Staeheli et al., 1986). 3-MC and BaP both significantly induced Cyp1a1 and 1a2 in the kidney (Fig. 1 right) although the induction of these two genes was less than observed in the liver. 3-MC induced the same set of genes in the kidney as in the liver with the exception of igf-binding protein-6 and metallothionein II. Maximal induction in both tissues for each chemical generally occurred between 4 and 24 h.

Changes in Gene Expression in the Liver and Kidney by DMN and ENU. Seven genes were significantly induced or repressed in liver with the use of both DNA alkylators, including acute phase proteins, metallothioneins, p21, and monokine-induced by -interferon (Fig. 1 left). In addition, ENU significantly induced Bax alpha and GST Mu, whereas DMN repressed both Cyp2e1 and mouse transglutaminase. In the kidney, ENU was the only chemical of the pair to significantly alter gene expression (Fig. 1 right). The genes that were induced or repressed in response to ENU included proteins involved in acute phase, phase I and II metabolism, heat shock, apoptosis, and DNA damage response.

Induction and Repression of Genes in Liver by CdCl₂ and HgCl₂. At the intraperitoneal dose administered (5 mg/kg CdCl₂ and 8.3 mg/kg HgCl₂) both heavy metals induced 12 genes in common (Fig. 2 left). Genes that were induced or repressed in the liver included heat shock proteins, acute phase proteins, metallothioneins, and antioxidant genes. CdCl₂ also induced or repressed an additional 11 genes that are involved in the immediate early response, inflammation, stress, and DNA damage response, whereas HgCl₂ uniquely induced or repressed four genes (Cyp2a, GST Pi, p21, and GST 5.7). Fifteen genes in the kidney exhibited altered expression levels in response to both heavy metals (Fig. 2 right). The response was generally greater in the kidney compared with the liver and fewer genes that were significantly altered in the kidney were unique to one specific metal. The 15 genes altered code for proteins involved in DNA damage response, phase I and II metabolism, oxidative stress, acute phase, heat shock, and metal binding. Seven genes were uniquely altered by either CdCl₂ or HgCl₂ and included transcription factors and stress proteins. Most significant gene induction occurred from 4 to 8 h, excluding GST Mu and GST Pi whose repression was significant at 48 h.

View larger version (194K): [in this window] [in a new window]   Fig. 2.   Up-/down-regulation of genes in liver (left) and kidney (right) at 4, 8, 24, or 48 h after i.p. administration of CdCl2 or HgCl2 to mice. Values in boxes are the means for three to four animals and are presented as percentage of control. Values that are colored represent genes that are significantly (P < 0.05) up-/down-regulated at one or several time points and color denotes the degree of induction/repression. Gray shading is used to denote those genes where expression was significantly up-/down-regulated at at least one time point. *Indicates genes uniquely altered by that specific compound or class of compounds.

Induction and Repression of Genes in Liver and Kidney by CCl₄ and Bromobenzene. CCl₄ and bromobenzene were administered intraperitoneally at a dose of 1 and 2.5 g/kg, respectively. All animals treated with bromobenzene died before the 48-h time point. At these doses, 14 genes were induced or repressed in the liver by both chemicals (Fig. 3 left). These included genes involved in the immediate early response, DNA damage response, oxidative stress, heat shock, and the acute phase. Six significantly altered genes were unique to CCl₄ and 10 were unique to bromobenzene. Among these 16 genes several of the genes were involved in phase II metabolism, oxidative stress, and transcription. In the kidney (Fig. 3 right), CCl₄ and bromobenzene induced or repressed eight genes in common, which are linked to the heat shock response, DNA damage response, and phase II metabolism. Maximal induction of genes in kidney and liver occurred at 8 h for CCl₄-treated animals and at 24 h after bromobenzene treatment.

View larger version (188K): [in this window] [in a new window]   Fig. 3.   Up-/down-regulation of genes in liver (left) and kidney (right) at 4, 8, 24, or 48 h after i.p. administration of CCl4 or bromobenzene to mice. Values in boxes are the means for three to four animals and are presented as percentage of control. Values that are colored represent genes that are significantly (P < 0.05) up-/down-regulated at one or several time points and color denotes the degree of induction/repression. Gray shading is used to denote those genes where expression was significantly up-/down-regulated at at least one time point. *Indicates genes uniquely altered by that specific compound or class of compounds.

Dose-Response Study of CdCl₂, DMN, and CCl₄. To illustrate dose-response relationships between toxicant exposure and gene expression, mice were treated with CdCl₂, DMN, or CCl₄ and gene expression in liver was examined at 8 h. The expression of selected genes is shown graphically in Fig. 4.

View larger version (21K): [in this window] [in a new window]   Fig. 4.   A, dose-response expression of p21, hsp e71, hsp86, and -glutamylcysteine synthetase in livers from mice treated with varying doses of CCl4. Data are the mean ± S.D. for three to four animals at each dose. *Significant differences from control P < 0.05. B, dose-response expression of p21, monokine induced by -interferon and Bax alpha in livers from mice treated with varying doses of DMN. Data are the mean ± S.D. for three to four animals at each dose. *Significant differences from control P < 0.05.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The sheer number of new synthetic entities being introduced annually combined with a need to understand the potential deleterious effects of most of these chemicals dictates that new methods be sought for rapid assessment and characterization of the toxicity of these agents.

Our work has focused on understanding the potential limitations of moderately sized expression arrays to determine whether microarrays offer a useful alternate approach in toxicity testing and environmental monitoring using whole animals. The current study has addressed the issue of whether distinct patterns of gene expression are observed for separate classes of chemicals to establish whether this would enable the classification of unknown compounds. Moreover, we have explored dose-effect relationships to ascertain whether gene expression analyses are a sufficiently sensitive endpoint to offer potential as an environmental monitoring tool.

In this study, comparison of gene induction between the 10 chemicals examined shows distinct groupings of chemical classes. Both peroxisome proliferators, DEHP and clofibrate, induced a number of genes involved in fatty acid -oxidation that are linked to the activation of the PPAR. On our DNA array, three genes induced by treatment with DEHP and clofibrate were unique for that class of compounds (acyl-CoA thioesterase, Cyp4a10, and igf-binding protein). Likewise, for every other chemical class represented in this study, there is at least one gene whose expression is uniquely up-regulated by that class of chemical but not by treatment with the other four classes of compounds (Figs. 1-3). These results illustrate that microarray analysis, even with a highly focused set of genes, is capable of discriminating between separate classes of chemicals.

In selecting the cDNAs for our arrays, many were chosen because they were known to be up-/down-regulated in response to the chemicals under study, and the results of the current work are entirely consistent with numerous studies that focused on one or more genes. The peroxisome proliferators and DNA alkylators increased the expression of several genes, which had not been previously reported to be affected by these compounds. Furthermore, the DNA alkylators did not induce a number of genes that had been anticipated, including O⁶-methylguanine-DNA methyltransferase.

Administration of DEHP and clofibrate increased the expression levels of Cyp4a10 and acyl-CoA thioesterase, genes whose expression is linked to PPAR binding. Both peroxisome proliferators also up-regulated igf-binding protein-1, a protein involved in modulating the response of insulin-like growth factor (Schuller et al., 1994). Previous reports indicate connections between the PPAR, insulin, and insulin-like growth factor in adipocyte differentiation (Zhang et al., 1996; Boney et al., 2000). The up-regulation of igf II receptors with exposure to peroxisome proliferators is also described (Lake et al., 2000), and although the induction of igf-binding protein-1 has not been noted earlier, its expression is consistent with these results.

The DNA alkylators DMN and ENU were expected to alter the expression levels of some of the DNA repair enzymes represented on the array, as well as some of the stress proteins, acute phase proteins, and the metallothioneins. Although the DNA alkylators affected the expression of the metallothioneins at the earliest time point and up-regulated p21, an indicator of DNA damage (Serfas et al., 1997), no genes specifically involved with DNA repair were altered. Alternatively, the DNA alkylators induced Bax alpha (Figs. 1 left and 4B), a protein involved in shuttling the cell toward apoptotic death (Pritchard and Butler, 1989; Oltvai et al., 1993; Horn et al., 2000). The DNA alkylators also affected the expression of several genes involved in the acute phase of inflammation and induced and repressed monokine induced by -interferon (Figs. 1 left and 4B). This protein is proposed to participate in immune and inflammatory responses and we surmise, based on the up-regulation at lower concentrations of DMN, that it may play a role in the acute phase of inflammation (Farber, 1990). Additionally, down-regulation of Cyp2e1 occurs with DMN, CCl₄, and bromobenzene. DMN, CCL₄, and bromobenzene are activated by Cyp2e1 (Lauriault et al., 1992) and, in turn, reactive metabolites of at least CCl₄ are known to destroy Cyp2e1 as well as neighboring microsomal membranes (Osawa et al., 1995). It is not clear, however, why Cyp2e1 was so strongly repressed rather than induced on exposure to these three chemicals.

In addition to the ability of even a moderately sized array to classify compounds by unique expression patterns, several other issues were addressed in these studies: 1) how do time, dose, and tissue selection affect the chemical "signature" obtained for a class of compounds; and 2) are the methods sufficiently robust to provide reliable assessments?

Both time and dose are important factors in determining not only which genes are up-regulated but also to what extent their expression is increased. In this study, clofibrate induced a number of genes 4 h after administration, but it was not until the 8-h time point that those same genes showed significant induction by DEHP. Similarly, in the kidney, almost all of the genes up-regulated in response to bromobenzene treatment were induced by CdCl₂ and HgCl₂. However, the time of induction varied greatly among the three chemicals. The majority of high-level gene induction occurred at 4 to 8 h for the heavy metals and at 24 h in the bromobenzene-exposed mice. Additionally, dose-response studies of three chemicals (BaP, DMN, and CCl₄) illustrate the need to determine a dose that is sufficiently high to induce and/or repress the majority of genes affected by treatment with that specific chemical (Fig. 4, A and B).

Increases in the expression levels of numerous genes were greater in kidney than in the liver of mice treated with CdCl₂ or HgCl₂, whereas up-regulation of all genes affected by DEHP and clofibrate was more pronounced in liver than kidney. These differences suggest that DNA array experiments involving multiple tissues may be helpful in identifying the target tissues of unknown toxicants and may potentially add a level of discrimination for compounds with similar expression profiles.

An area of concern, which has received attention, is whether the technique is sufficiently consistent to be used effectively as a screening method. By spotting individual genes multiple times and using several animals per time point we were able to account for variability associated with dust, uneven slide coatings, and animal variations. Data collected with our array system was therefore highly consistent, and we were able to observe statistically significant gene induction.

Although the consistency of the data obtained in the past few years has improved dramatically, the work presented here suggests that relatively toxic doses of chemical must be administered to obtain significant alterations in gene expression. There are likely several issues associated with this, including animal-to-animal variability of the outbred mouse strain used in these studies. In addition, strain, species, sex, nutritional status, and a host of other differences may strongly influence the gene expression patterns observed in response to toxicant exposure. Finally, all of the measurements made in the current studies were performed with RNA isolated from whole tissues, yet many of the chemicals used in the studies show highly focal toxicity. The fact that some of these chemicals influence only a small percentage of the total cell population in heterogeneous tissues such as the kidney likely decreases the sensitivity (signal to noise) of the measurements. Nevertheless, the current work demonstrates the potential for the use of large-scale gene expression data in assessing toxicity of a large number of chemical entities. We conclude, however, that studies of gene expression patterns as they are currently configured, are unlikely to be of much use in environmental monitoring unless exposure levels are high.