Correlation of the Structures of Agricultural Fungicides to Gene

Classification of Heavy-Metal Toxicity by Human DNA Microarray Analysis. ... Development of a DNA Microarray Chip for the Identification of Sludge Bac...
0 downloads 0 Views 229KB Size
Environ. Sci. Technol. 2003, 37, 2788-2793

Correlation of the Structures of Agricultural Fungicides to Gene Expression in Saccharomyces cerevisiae upon Exposure to Toxic Doses E M I K O K I T A G A W A , †,‡ YUKO MOMOSE,† AND H I T O S H I I W A H A S H I * ,† National Institute of Advanced Industrial Science and Technology (AIST), Central-6, Higashi 1-1-1, Tukuba, Ibaraki 305-8566, Japan, and New Energy and Industrial Technology Development Organization (NEDO), Higashi-Ikebukuro 3-1-1, Toshimaku, Tokyo 170-6028, Japan

Correlations between the chemical structures of agricultural fungicides and mRNA expression levels following exposure of Saccharomyces cerevisiae to toxic doses of thiuram, zineb, maneb, TPN, and PCP were examined. Structurally, thiuram, zineb, and maneb are dithiocarbamate fungicides, whereas TPN and PCP are not. To characterize chemical toxicity, genes expression was classified according to the functional groups used by the MIPS database. However, no correlations between the classification scheme and chemical structures were found. Hierarchical clustering of gene expression profiles was performed to characterize the effects of the five chemicals. According to this analysis the similarity of gene expression profiles depended on the similarity of chemical structures. These results suggest that DNA microarray technology has potential for predicting the major chemicals which will cause environmental toxicity and will provide information on new biomonitoring methods.

Introduction The number and type of synthetic chemicals that are being produced worldwide continues to increase significantly. As of October 2001, more than 18 million organic and inorganic substances were registered in the Chemical Abstract Service (CAS). While these industrial chemicals provide numerous benefits, there is no doubt that some have potential to damage the environment and health. Toxicity must be evaluated, and use must be carefully controlled and monitored in order to minimize potential damage. Biological assays, such as the Ames test and animal tests, have been used to estimate toxicity. However, after chemicals have been introduced into the environment, they can be difficult to detect. In fact, only 10% of industrial chemicals can be identified in the environment (1), and this small percentage is considered inadequate for protection. Thus, new methods must be developed to evaluate the toxicity of chemicals and to detect their presence * Corresponding author phone/fax: +81-298-61-6066; e-mail: [email protected]. † National Institute of Advanced Industrial Science and Technology (AIST). ‡ New Energy and Industrial Technology Development Organization (NEDO). 2788

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 12, 2003

in the environment. DNA microarray technology, which makes it possible to obtain expression data of thousands of genes simultaneously, has become an important new technique in toxicology. We are using the yeast Saccharomyces cerevisiae as a model organism for toxicological study because it is a simple, fast-growing eukaryote that has been thoroughly characterized and is also amenable to DNA microarray analysis. Thus, it is possible to analyze expression of its approximately 6000 genes upon exposure to chemical treatment. Previously, we reported the possibility for understanding chemical toxicity and for selecting biological markers using thiuram as a model chemical (2). As a next step, we are focusing on the possibility of using DNA microarray analysis for predicting the toxicity of chemicals in environment. Certain chemical structures correlate with toxicity, and therefore, toxicity may correlate with gene expression (Figure 1). Our interest is in understanding the correlation between chemical structure and gene expression profiles. To examine this possibility, five agricultural chemicals were chosen based on their structures (Table 1) (3). Three of the five are members of the dithiocarbamate family and the others are organochlorines. Thiuram [tetramethylthiuram disulfide, thiram] is a dimethyl dithiocarbamate compound used as a fungicide to prevent damage to pre- and postharvested crops. It is also used to protect seeds, to repel animals, and to process rubber. Zineb [zinc ethylenebis (dithiocarbamate), CAS; 12122-67-7] is used as a fungicide much like thiuram. Maneb [manganese ethylenebis (dithiocarbamate), CAS; 12427-38-2] is one of the fungicides used in the control of many diseases of fruits, vegetables, field crops, and ornamentals with a wider range than other fungicides. TPN [tetrachloroisophthalonitrile (chlorothalonil), CAS; 1897-45-6] is a broad-spectrum organochlorine fungicide. PCP [pentachlorophenol, CAS; 87-86-5] is a chlorinated hydrocarbon and is used primarily as an insecticide and fungicide to protect timber; it may also be used as defoliant, herbicide, and biocide. To discover the correlations between chemical structures and gene expression profiles in yeast following chemical exposure, we used two methods. One was to characterize gene expression according to functional classes. This analysis indicated toxicity but did not reveal correlations with chemical structures. The other method was to use hierarchical clustering of gene expression profiles, which did reveal the desired correlations. These results suggest that DNA microarray technology has the potential for predicting chemical structures that cause major environmental toxicity.

Materials and Methods Strains, Media, and Chemicals. Saccharomyces cerevisiae S288C (R SUC2 mal mel gal2 CUP1) was used as the indicator strain and was grown in YPD (2% polypeptone, 1% yeast extract, 2% glucose) at 25 °C. Thiuram, zineb, maneb, and TPN were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). PCP was purchased from Nacalai Tesque Inc., (Kyoto, Japan). Yeast growth was measured at 25 °C in a Biomek 2000 Laboratory Automation Workstation (Beckman Coulter, Inc., Fullerton, CA) with a spectrophotometer. A yeast preculture was inoculated into YPD supplemented with chemicals at different concentrations in 96-well plates. Absorbance at 650 nm was measured automatically every 2 h to obtain yeast growth curves for each treatment. Cell survival was determined by plating cells on YPD before and after each 2 h chemical treatment and subsequently counting colony forming units (CFU). 10.1021/es026156b CCC: $25.00

 2003 American Chemical Society Published on Web 05/16/2003

FIGURE 1. Relationships among chemical structure, toxicity, and gene expression profiles. Yeast Treatments. Yeast cultures in YPD were diluted and grown overnight to an optical density (OD660) of 1.0. Chemicals at various concentrations were added, and cells were allowed to grow for an additional 2 h. Cells were harvested by centrifugation and stored at -80 °C until use. Microarray Experiments. Microarray experiments were carried out as previously described (2). Total RNA were extracted by the hot-phenol method (4), and mRNA were purified with an Oligotex-dT30 mRNA purification kit (TaKaRa, Otsu, Shiga, Japan). Poly (A) +RNA was purified from about 400 µg of total RNA. Labeling and hybridization were also performed as described (2). The mRNA extracted from treated cell and nontreated cell were labeled with Cy5dUTP and Cy3-dUTP, respectively. Yeast DNA microarrays were purchased from DNA Chip Research, Inc. (Yokohama, Kanagawa, Japan). Scanning and Data Analysis. A Scan Array 4000 laser scanner (GSI Lunomics, Billeria, MA) was used to acquire hybridization signals. Array images were analyzed with Gene Pix 4000 (Inter Medical). The background intensity around each spot was subtracted from each, and spots having less intensity than the value following mentioned as cutoff value were disregarded. The cutoff value was the median value plus two standard deviations (SD) for all background signals

on each array. Expression data were evaluated as the ratio of the value from treatment to that from the untreated control. Global normalization was carried out using the median of log2 expression ratio for whole probes on one array. Normalized expression value was evaluated by the coefficient of variation (CV), which was calculated as the ratio of the SD to the mean of triplicate expression values. Data for which the CV was greater than one were considered invalid and were not used. Classification of Genes. The MIPS functional catalog (http://www.mips.biochem.mpg.de/proj/yeast/; 2002/2/5) (5, 6) was used for classification analysis. Nineteen groups were used: 1. cell cycle and DNA processing (628 genes), 2. cell fate (427 genes), 3. cell rescue, defense, and virulence (278 genes), 4. cellular communication/signal transduction mechanism (59 genes), 5. cellular transport and transport mechanisms (495 genes), 6. classification not yet clear-cut (115 genes), 7. control of cellular organization (209 genes), 8. energy (252genes), 9. metabolism (1066 genes), 10. protein activity regulation (13 genes), 11. protein fate (folding, modification, destination) (595 genes), 12. protein synthesis (359 genes), 13. protein with binding function or cofactor requirement (structural or catalytic) (4 genes), 14. regulation of/interaction with cellular environment (199 genes), 15. subcellular localization (2258 genes), 16. transcription (771 genes), 17. transport facilitation (313 genes), 18. transposable elements, viral, and plasmid proteins (116 genes), 19. unclassified proteins (2399 genes). “Metabolism” groups were further classified into seven subclasses: 1. amino acid metabolism (204 genes), 2. C-compound and carbohydrate metabolism (415 genes), 3. lipid, fatty acid, and isoprenoid metabolism (213 genes), 4. metabolism of vitamins, cofactors,

FIGURE 2. Effects of chemicals on yeast growth. A, B, C, D, and E are thiuram, zineb, maneb, TPN, and PCP, respectively. Open triangles indicate growth in the absence of chemicals, and open circles indicate the following concentrations; thiuram: 75 µM, zineb: 2 ppm, maneb: 2 ppm, TPN: 10 µM, and PCP: 50 µM. Black circles indicate concentrations 3-fold higher than the concentration shown by open circles, and black boxes indicate concentrations 3-fold less. VOL. 37, NO. 12, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2789

TABLE 1. Test Chemicals

and prosthetic groups (86 genes), 5. nitrogen and sulfur metabolism (67 genes), 6. nucleotide metabolism (148 genes), 7. phosphate metabolism (33 genes). The classified genes were further separated by their expression ratio converted to log2, and the number of classified genes was represented in a histogram. Clustering. Hierarchical clustering is a statistical method to correlate mathematical values and can be used here to make groups (trees) among gene expression profiles (7-9). The trees were made using GeneSpring software (Silicon Genetics, Redwood City, CA). Settings for the calculation were as follows: similarity was measured by standard correlation; the separation ratio value was 1.0; and the minimum distance value was 0.001. Genes used for this analysis are described in the text.

Results and Discussion Conditions for Chemical Treatments. To determine conditions for the chemical treatments, Saccharomyces cerevisiae was grown with various concentrations of five different chemicals. As shown in Figure 2, thiuram, zineb, maneb, TPN, and PCP showed significant growth inhibition. Generally, no significant changes were observed in the pattern of gene expression under conditions that slightly inhibited growth, and insufficient signals were detected under lethal conditions as well (data not shown). Concentrations of 75 µM (2), 2 ppm, 2 ppm, 10 µM, and 50 µM of thiuram, zineb, maneb, TPN, and PCP, respectively, were not lethal but had a significant effect on cellular growth. Under these conditions, a decreased CFU was not observed compared to that before the addition of each chemical (data not shown). 2790

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 12, 2003

Functional Characterization of Genes Altered by the Chemical Treatments. Valid expression data were grouped by functional categories and were divided into seven classes within each functional category. From Figure 3, the peaks of“cell rescue, defense and virulence”, “energy”, “protein fate”, and “unclassified proteins” were clearly shifted by each of the five treatments, many genes in these categories were shown high expression ratio. The peaks of “metabolism” were also shifted upward by each of the five treatments. As in Figure 3, genes classified in “metabolism” were subdivided based on subcategories of MIPS (Figure 4 A-E). From the standpoint of characterizing chemical effects on gene expression, the functional classification given in Figures 3 and 4 can help clarify chemical toxicity. Similarly, it may be possible to evaluate the toxic affects of unknown chemicals or mixtures of chemicals. However, the pattern of repression and induction caused by the chemically similar compounds thiuram, zineb, and maneb was not easily distinguishable from that caused by the dissimilar compounds TPN and PCP. In other words, it was difficult to correlate chemical structures with gene expression by this way. Clustering Analysis by Genome Wide Gene Expression Profiles. Although the functional classification was useful for characterizing chemical toxicities, it was not helpful in understanding correlations between chemical structure and gene expression profiles. An attempt was made to analyze the profiles using a clustering method. Hierarchical clustering is a classification tool that allows visualizing relationships among chemical stress conditions by organizing them on the basis of gene expression profiles. The results are obtained

FIGURE 3. Histograms of genes expression levels by functional class. Yeast genes were classified by function according to MIPS and quantified on the basis of expression level (log2). The quantification class was shown on the top right; each symbol indicates the number of genes the expression level converted to log2 was less than -3, from -3 to -2, from -2 to -1, from -1 to 1, from 1 to 2, from 2 to 3, and over than 3. A, B, C, D, and E are thiuram, zineb, maneb, TPN, and PCP, respectively. The functional classification is shown on the right; 1. cell cycle and DNA processing, 2. cell fate, 3. cell rescue, defense, and virulence, 4. cellular communication/signal transduction mechanism, 5. cellular transport and transport mechanisms, 6. classification not yet clear-cut, 7. control of cellular organization, 8. energy, 9. metabolism, 10. protein activity regulation, 11. protein fate (folding, modification, destination), 12. protein synthesis, 13. protein with binding function or cofactor requirement (structural or catalytic), 14. regulation of/interaction with cellular environment, 15. subcellular localization, 16. transcription, 17. transport facilitation, 18. transposable elements, viral and plasmid proteins, 19. unclassified proteins. as tree dendrograms (7-9). In these trees, chemicals having similar effects on gene expression profiles are placed close to one another. Clustering analysis of the current experiment and data from a previous experiment using Cd (10, 11) are shown in Figure 5. All available gene expression data were incorporated into the analysis. In this figure, the zineb treatment and the maneb treatment branches are sideby-side, suggesting that these two chemicals have similar effects in yeast. The PCP treatment branch is most distant from the others. Although both TPN and PCP are organochlorines, TPN and PCP were not found to be similar to each other. TPN was identified as being more similar to dithiocarbamate than to PCP. This may be because the CN-group

of TPN had similar effects to dithiocarbamates than C1. Clustering with Selected Genes. To be useful, methods developed to predict the toxic effects of environmental chemicals must be simple, rapid, and economical. Analysis of an entire genome may be excessive for this purpose, as an expression of unaffected genes may obscure possible correlations. As a potential solution, we used a subset of genes that showed an expression ratios of over 2-fold. Eight hundred and thirty four, 262, 101, 1063, and 570 genes from the experiments involving thiuram, zineb, maneb, TPN, and PCP treatments, respectively, were selected (Table 2). Genes with expression ratios of over 2-fold by each of the five chemical treatments were recognized as “common” VOL. 37, NO. 12, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2791

FIGURE 4. Histograms of genes expression levels in “metabolism” class. A, B, C, D, and E are thiuram, zineb, maneb, TPN, and PCP, respectively. The subclassification shown on the right in “metabolism” class: 1. amino acid metabolism, 2. C-compound and carbohydrate metabolism, 3. lipid, fatty acid, and isoprenoid metabolism, 4. metabolism of vitamins, cofactors, and prosthetic groups, 5. nitrogen and sulfur metabolism, 6. nucleotide metabolism, 7. phosphate metabolism.

TABLE 3. The List of Genes Selected as “Commonly” ORF

FIGURE 5. Cluster analysis of whole expression profiles following treatment of yeast with six different chemicals. A dendrogram was drawn to indicate the relationship among the effects of each chemical treatment. The numbers shown in each branch are (1-correlation coefficient).

TABLE 2. Number of Selected Genes in Each Experiment thiuram zineb maneb total valid data over than 2-fold (each chemical specific) less than 0.5-fold (each chemical specific)

5740 5530 834 (196) 545 (110)

TPN

PCP

5751 5752 5752 5753 5498 5704 5714 5691 262 101 1063 570 (47) (19) (405) (253) 144 30 975 159 (67) (9) (504) (60)

(intersection, product set). Thirty genes were found in this product set and were used to make a tree (Table 3). As shown in Figure 6 A, the effects of thiuram and zineb were recognized as being more similar than those caused by the other chemical treatments. Maneb was placed at a branch adjacent to the thiuram and zineb branches. The effects of PCP and Cd were found to be similar. The result was not consistent with Figure 5 that was used with whole profiles. In Table 3, most of the genes were involved in detoxification for oxidation stress, it 2792

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 12, 2003

YBL064C YBR008C YBR256C YCR102C YCR107W YDL243C YDR070C YFL014W YFL056C YFL057C

gene

ORF

YGR043C YGR256W YHR029C YHR179W AAD3 YJR010W AAD4 YJR155W YKL001C HSP12 YKL142W AAD6 YKR071C YLL060C FLR1 RIB5

gene

ORF

YLR303W YLR460C YML131W OYE2 YMR173W MET3 YNL331C AAD10 YOL150C MET14 YOL151W MRP8 YOL165C YPL171C GTT2 YPL280W

gene MET17

GND2

DDR48 AAD14 GRE2 AAD15 OYE3

was implied that evaluation using only such genes might cause erroneous judgment. On the other hand, genes that were expressed over 2-fold by only one chemical were recognized as “specific” (difference set). The number of specifically expressed genes was 196, 47, 19, 405, and 253 for thiuram, zineb, maneb, TPN, and PCP treatments, respectively (A total of 920 genes were identified as specific (http://kasumi.nibh.jp/∼iwahashi/.) The result of clustering these specifically expressed genes is shown in Figure 6B. The relationship of the five chemicals shown in Figure 6B is different than that shown in Figures 5 and 6A. The difference is that the zineb branch is now closer to those of the maneb. One might have expected that the three dithiocarbamates (thiuram, zineb, and maneb) would have been grouped together; in particular, zineb and maneb would have been expected to be the most closely related. As expected, these three chemicals were classified closely, and zineb and maneb were recognized as being more similar to one another. The result is because the clustering calculation was based on differences; similarities were excluded. Genes with expression ratios of over 2-fold by at least one chemical were designated as “fluctuating” (union set, sum

recognized as similar. On the other hand, clustering specific genes was based on differences between chemical effects, so branches tended to form in the dendrograms. In the case of the fluctuating genes, the calculation is based on both similarities and differences of chemical effects. Thus, it was concluded that, of the above clustering methods, based on fluctuating genes may be better. In the present study, gene expression profiles in yeast resulting from exposure to five chemicals were obtained. The profiles were characterized based on gene function and hierarchical clustering. The classification by function may be used to understand chemical toxicity, and the hierarchical clustering may have application in predicting chemical structures that cause environmental toxicity. For the purposes of predicting chemical structures using gene expression profiles, whole genome data should not be used, but rather that of a smaller number of genes, selected as fluctuating or specific may suffice. Selected individual genes may be assayed by new biomonitoring methods such as real-time RT-PCR, promoter assays, or immunological methods.

Acknowledgments We thank Dr. Koga (Institute for Biological Resources and Functions Biologically Active Substances Research Group AIST) for providing useful program, ChipCleanser.

Literature Cited

FIGURE 6. Cluster analysis of the effects of six different chemicals on gene expression patterns. Analyzed genes were selected as common (A), specific (B), and fluctuating (C). The numbers shown in each branch are (1-correlation coefficient). set). In this selection, 1684 genes were found (Figure 6C) (http://kasumi.nibh.jp/∼iwahashi/). Analysis of the five chemicals and that of Cd led to the same tree as was produced using clustering analysis of the specific profiles (Figure 6B). On the other hand, 545, 144, 30, 975, and 159 genes showed an expression ratio of less than 0.5-fold (Table 2). Although minor differences were found, almost the same results were obtained in the case of using genes showing expression ratios over 2-fold (data not shown). Clustering common genes was dependent on the similarity of chemical effects, so chemical effects tended to be

(1) Suzuki, M.; Utsumi, H. Bioassay; Koudannsya Press: Tokyo, 1998; pp 3-4. (2) Kitagawa, E.; Takahashi, J.; Momose, Y.; Iwahashi, H. Environ. Sci. Technol. 2002, 36, 3908-3915. (3) TOXNET. /http://toxnet.nlm.nih.gov/ (accessed Oct. 2001). (4) Kohrer, K.; Domdey, H. Methods Enzymol. 1990, 194, 398-401. (5) Mewes, H. W.; Hani, J.; Pfeiffer, F.; Frishman, D. Nucleic Acids. Res. 1998, 26, 33-37. (6) Mewes, H. W.; Heumann, K.; Kaps, A.; Mayer, K.; Pfeiffer, F.; Stocker, S.; Frishman, D. Nucleic Acids Res. 1999, 27, 44-48. (7) Eisen, M. B.; Spellman, P. T.; Brown, P. O.; Botstein, D. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 14863-14868. (8) Ohmine, K.; Ota, J.; Ueda, M.; Ueno, S.; Yoshida, K.; Yamashita, Y.; Kirito, K.; Imagawa, S.; Nakamura, Y.; Saito, K.; Akutsu, M.; Mitani, K.; Kano, Y.; Komatsu, N.; Ozawa, K.; Mano, H. Oncogene 2001, 20, 8249-8257. (9) Murata, Y.; Momose, Y.; Hasegawa, M.; Iwahashi, H.; Komatsu, Y. Chemi-Bio. Inform. J. 2002, 2, 18-31. (10) Momose, Y.; Kitagawa, E.; Iwahashi, H. Chemi-Bio. Inform. J. 2001, 1, 41-50. (11) Momose, Y.; Iwahashi, H. Environ. Toxicol. Chem. 2001, 20, 2533-2360.

Received for review September 13, 2002. Revised manuscript received March 25, 2003. Accepted April 2, 2003. ES026156B

VOL. 37, NO. 12, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2793