ToxSci Advance Access originally published online on December 15, 2007
Toxicological Sciences 2008 102(1):42-60; doi:10.1093/toxsci/kfm293
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gene Expression Profiles in Rainbow Trout, Onchorynchus mykiss, Exposed to a Simple Chemical Mixture
,2
* Battelle, Marine Research Operations, West Sequim Bay Road, Sequim, Washington 98382
Pacific Northwest National Laboratory, P.O. Box 999, Richland, Washington 99352
1 To whom correspondence should be addressed. Fax: (360) 681-4559. E-mail: sharon.hook{at}pnl.gov.
Received April 20, 2007; accepted October 27, 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
Among proposed uses for microarrays in environmental toxiciology is the identification of key contributors to toxicity within a mixture. However, it remains uncertain whether the transcriptomic profiles resulting from exposure to a mixture have patterns of altered gene expression that contain identifiable contributions from each toxicant component. We exposed isogenic rainbow trout Onchorynchus mykiss, to sublethal levels of ethynylestradiol, 2,2,4,4-tetrabromodiphenyl ether, and chromium VI or to a mixture of all three toxicants Fluorescently labeled complementary DNA (cDNA) were generated and hybridized against a commercially available Salmonid array spotted with 16,000 cDNAs. Data were analyzed using analysis of variance (p < 0.05) with a Benjamani–Hochberg multiple test correction (Genespring [Agilent] software package) to identify up and downregulated genes. Gene clustering patterns that can be used as "expression signatures" were determined using hierarchical cluster analysis. The gene ontology terms associated with significantly altered genes were also used to identify functional groups that were associated with toxicant exposure. Cross-ontological analytics approach was used to assign functional annotations to genes with "unknown" function. Our analysis indicates that transcriptomic profiles resulting from the mixture exposure resemble those of the individual contaminant exposures, but are not a simple additive list. However, patterns of altered genes representative of each component of the mixture are clearly discernible, and the functional classes of genes altered represent the individual components of the mixture. These findings indicate that the use of microarrays to identify transcriptomic profiles may aid in the identification of key stressors within a chemical mixture, ultimately improving environmental assessment.
Key Words: gene expression; microarrays; chemical mixtures; rainbow trout; toxicokinetics.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
Toxicity from chemical mixtures is often difficult to discern because of unpredictable interactions and nonadditive effects. Synergistic interactions of contaminants can cause toxic effects at exposure levels lower than observed in single contaminant exposures (Yang et al., 2004
). Chemical analysis of many chemical mixtures often does not correlate with toxicity because only a fraction of the mixture components associated with observed toxicity may be identified (Donnelly et al., 2004; Eide et al., 2004). Highly contaminated sites are particularly difficult to characterize because of the complexity of the mixture and changes in the chemical composition of the site with deposition and degradation over time (Donnelly et al., 2004). Neurotoxic, immunotoxic, and cytotoxic interactions can be difficult to predict. Despite these challenges, understanding the impact of exposure to multiple chemical contaminants is important for risk assessment as it is more environmentally realistic. Consequently, rapid screening tools for chemical interactions are needed to improve risk assessment for highly contaminated sites (Donnelly et al., 2004).
DNA microarrays are increasingly being used as a tool by environmental toxicologists and offer the potential to help guide regulatory decisions (Ankley et al., 2006). Microarrays are typically printed with cDNA's or oligo gene fragments and allow for the simultaneous measurement of hundreds to thousands of genes. Microarrays have recently been developed for use with a plethora of environmentally relevant species, such as mussels (Mytilus galloprovincialis), Daphnia magna, European Flounder (Platichthys flesus), fathead minnows (Pimephales promelas) and other fishes (Soetaert et al., 2006; Venier et al., 2006; Williams et al., 2006; Wintz et al., 2006). The multivariate nature of microarrays has been proposed as a means of resolving the complexity of contaminant mixtures (Amin et al., 2002). The patterns of genes altered by exposure to a contaminant mixture may include a set of genes characteristic of exposure to a particular compound, and thereby allow identification of that compound as a component of the chemical mixture (e.g., Finne et al., 2007). Also, nonadditive interactions may be discerned by comparing the gene expression patterns from exposure to the mixture with those identified from the individual components (Amin et al., 2002).
In this study, our objective was to describe the transcriptomic profile in the rainbow trout (Onchorynchus mykiss) liver resulting from exposure to a simple mixture and compare this with the gene expression profiles of the individual components of the mixture. Our first aim was to determine if the contributions from each contributing contaminant to the mixture would be discernible in the gene expression pattern generated by exposure to the simple mixture. Another aim was to determine if the changes in gene expression resulting from the mixture were simply an additive or comparative list of genes altered in the individual chemical exposures, or whether the patterns interact in a way that would suggest nonadditive interactions. We chose three environmentally important contaminants with diverse modes of toxic action: ethinyl estradiol (EE2), a xenoestrogen known to decrease the fertility of male trout (Schultz et al., 2003), 2,2,4,4-tetrabromodiphenyl ether (BDE-47), a flame retardant with putative thyroid activity, which also impacts reproduction in fish (Muirhead et al., 2006), and hexavalent chromium (Cr-VI) a metal, which causes multiorgan pathology in salmon (Farag et al., 2006) and oxidative stress in fish (Tagliari et al., 2004). We continuously exposed sexually immature trout to EE2 (100 ng/l) for 7 days and administered the BDE-47 and Cr-VI by direct intravascular injection via an indwelling dorsal aorta cannula. These exposure routes were chosen to carefully control the contaminant dose, minimizing variations due to differences in bioavailability and provide greater consistency in delivered dose across individuals. The latter was intended to better aid in the identification of intrinsic transcriptomic changes associated with exposure to the mixture and individual toxicants. Following the 7-day exposure, the liver was collected and hepatic gene expression was measured via cDNA microarray and a subset of sequences confirmed via q RT PCR (quantitative reverse trancriptase polymerase chain reaction).
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
Fish.
All fish were maintained according to the guidelines established by the Institutional Animal Care and Use Committee of Battelle. All experiments used male isogenic (clonal hybrids) of the OSU x Swanson cross, which reduces phenotypic variation among the trout (Young et al., 1996
). These fish were transferred to the Marine Research Operations, Sequim, WA, laboratory at 530° days in age and continuously reared at Sequim. At the time of the toxicant exposures, the trout were approximately 4500° days in age, which corresponded to 13 months posthatching. The mean body weight ± SD for these fish were 319 ± 39 g (n = 12). These trout were sexually immature based on the size of the testes at termination, which was < 0.1% of body weight. Subsequent kinetic experiments were performed using trout that were approximately 8000° days in age, which corresponded to 22.2 months and were 1020 ± 347 g in body weight. These latter fish were sexually mature and had completed their first spawning 1 month prior to the kinetic studies. Prior to exposures, all trout were maintained in 1400-l circular tanks at a holding density not exceeding 5 kg/m3. Holding tanks were maintained as single pass flow through with a minimum in-flow rate of 4 l/min. Water temperature, dissolved oxygen, and pH were monitored weekly with mean values (± SD) of 12 ± 1°C, dissolved oxygen: 9.6 ± 0.21 mg/l, pH: 8.10 ± 0.03. All trout were maintained under a simulated natural photoperiod regime with graded on and off controls and fed Bio-Oregon soft moist pellets of various sizes based on fish size.
Chemicals.
The study contaminants were > 99% purity and were obtained from the following sources: EE2, Cr-VI, 16,16,17-d3 17β-estradiol were obtained from Sigma. BDE-47 was obtained from Chem Service (West Chester, PA). All other chemicals used were of reagent grade.
Experimental design: gene expression studies.
Trout were separately exposed to EE2, BDE-47, and Cr-VI or as a ternary mixture. All trout were fitted with a dorsal aortic cannula made from 28-G thin wall Teflon tubing using procedures previously described (Schultz and Hayton, 1997). Cannulated trout were individually housed in 370-l circular tanks and allowed a 48-h recovery period prior to exposures. The EE2 exposures were performed as a continuous water exposure lasting 7 days. A nominal water concentration of 100 ng/l EE2 was prepared by adding the EE2 dissolved in methanol. Briefly, the EE2 stock solution was slowly added to the exposure tanks using a peristaltic pump at a flow rate of 0.06 ml/min (equivalent to 0.0015% methanol in the tank water). The exposure tanks were allowed to equilibrate with the EE2 dosing system for 3 days prior to the addition of trout. Both BDE-47 and Cr-VI were administered as two intra-arterial bolus injections, given 3 days apart. The dosing solution of BDE-47 was prepared in dimethyl sulfoxide (DMSO) at a concentration of 10 mg/kg and the Cr-VI was dissolved in 0.9% NaCl (wt/vol) at a concentration of 12.5 µg/kg. The volume of fluid administered in both instances was 0.1 ml/kg corresponding to an injected dose of 1 mg/kg BDE-47 and 12.5 µg/kg Cr-VI. For the mixture experiment, trout were exposed to EE2 as described previously and coadministered BDE-47 and Cr-VI on separate days during the exposure corresponding to the dosing schedule used for the individual chemical exposures. Control trout were either exposed to methanol, DMSO, and saline only or a combination of the three corresponding to the exposures described previously. A summary of the dosing regimens is presented in Figure 1.
|
Experimental design: toxicokinetic studies.
Toxicokinetic experiments were performed to assess the uptake of EE2 and disposition of BDE-47 in trout under exposure conditions similar to that used for gene expression studies. For these experiments, cannulated trout were exposed to EE2 analogously as for the gene expression studies and blood samples were removed at 1, 4.0, 10.0, 20.0 h and then at selected times until 192 h (8 days) and the plasma stored at –80°C for later analysis of EE2. During the water exposure to EE2, trout were also administered a single intra-arterial dose of BDE-47 and Cr-VI on separate days according to the initial dosing schedule in Figure 1. After the BDE-47 dosing, blood samples were removed at 0.167, 0.333, 0.667, 1, 1.5, 3, 6.3, 9.1, 20 h and variously thereafter until 320 h (relative to the time of BDE-47 dosing). Immediately after injection, the cannula was rinsed three times with blood to remove residual traces of the dose to prevent contamination of subsequent blood samples. The EE2 plasma concentration–time profiles were compared with profiles collected from single chemical exposure studies described in Skillman et al. (2006)
. For BDE-47, a separate group of four trout were administered an intra-arterial injection of BDE-47 and sampled in a similar manner to the trout coexposed to EE2 and Cr-VI.
Chemical analysis.
The exposure water, plasma and tissue samples were analyzed for EE2 by gas chromatograph mass spectrometry (GC–MS). Water samples (0.05 l in volume) were fortified with NaCl (5% wt/vol) and extracted with 5 ml of methyl-tert-butyl-ether (MTBE) per sample. Plasma and liver samples (0.2 ml or g) were mixed with an equal of volume of saturated aqueous NaCl solution and extracted with 1 ml of MTBE. Prior to MTBE extraction, liver samples were homogenized with a ground glass/Teflon grinder. The MTBE fractions were removed, evaporated under N2, and subsequently derivatized with N-methyl-N-trimethylsilyl-trifluoroacetamide essentially as described in Schultz et al. (2001) except that 3d-estradiol was added as an internal standard in the extractions. The GC–MS was operated in selected ion monitoring mode with m/z 419 and m/z 425 ions used for 3d-estradiol and EE2 quantification. EE2 recovery from fortified water standards and blood plasma typically exceeded 95%. For BDE-47 analysis, liver samples were homogenized using a ground-glass/teflon grinder with 2 ml of NaCl solution and 20 µl of 10 µg/ml PCB-103 (wt/vol in hexane; purchased from Sigma-Aldrich, St Louis, MO) as the internal standard. Plasma and liver homogenates were then mixed with 1 ml of hexane and vortexed for 30 s and then centrifuged at 3000 x g for 5 min. The hexane layer was transferred to a GC vial. The hexane extracts were analyzed on a Hewlett-Packard 5890 GC equipped with a DB-5 30m, 0.25µM capillary column. The GC was operated in split injection mode with a split ratio of 1:10. Standard curves were prepared using blank matrices spikes with BDE-47. The recoveries for BDE-47 were typically 70–95% depending on matrices. Total Cr was measured in plasma and liver samples after digestion in concentrated nitric acid by inductively coupled plasma-optical emission spectroscopy (Perkin–Elmer model 4300; Perkin-Elmer, Fremont, CA) using matrix matching and standard calibration curves.
Toxicokinetic analysis.
Noncompartmental methods were used to analyze the EE2 and BDE-47 concentration–time profiles. The area-under-the-curve extrapolated to infinity (AUC0
) was estimated using the linear trapezoidal method with the area from the last sampling time to infinity calculated from the slope of the terminal portion of the concentration–time profile (β). Steady-state volume of distribution (Vss) and total body clearance (Clb) were calculated from the intravenous pharmacokinetic profile as described by Yamaoka et al. (1978) and Gibaldi and Perrier (1982). Plasma half-life (t
β) was calculated as tβ = 0.693/β.
RNA extractions.
Fish were euthanized with a lethal overdose of MS 222 (250 mg/l). Plasma was collected for chemical analysis, and liver was immediately removed and subsectioned. One aliquot of liver was frozen for chemical analysis; the remaining pieces were placed in RNAlater (Qiagen, Valencia, CA) and stored following the manufacturer's protocols. RNA was extracted using a standard TRIzol procedure (Invitrogen, Carlsbad, CA) and purified with the TURBO DNAfree kit (Ambion, Austin, TX). Total RNA was quantified via fluorometry using ribogreen reagent (Molecular Probes, Carlsbad, CA) and RNA quality was verified via gel electrophoresis. After processing, SUPRNasin (RNase inhibitor, Ambion) was added to help maintain sample integrity and RNA was stored at –80°C.
Microarray methods.
Salmonid cDNA microarrays were obtained from the GRASP consortium (Dr Ben Koop, University of Victoria, Canada). These arrays have 16,000 cDNA and expressed sequence tags from either Atlantic salmon (Salmo salar) or rainbow trout (O. mykiss). The methods for obtaining the cDNAs for the array, developing the arrays, and validating the arrays themselves are described in detail in Rise et al. (2004a). Sequence homology between the two species is sufficiently high, allowing for the cross species use of the array (Rise et al., 2004a). Array hybridizations were performed in a 3 x 3 replicate design; with three animal replicates and three technical replicates. RNA was transcribed into cDNA and indirectly labeled via an aminoallyl technique (Invitrogen's Superscript cDNA Indirect Labeling kit). Control cDNA was labeled with Cy3 (Amersham), and exposed cDNA was labeled with Cy5. A split control experiment (where control RNA is put into two separate tubes, labeled with Cy3 or Cy5, then recombined for array hybridization) was also performed to examine genes selected as significant due to differences in dye incorporation (Draghici, 2003). Exposed and control samples were paired according to cDNA yield and label incorporation (such that each technical replicate was paired with RNA extracted from a different control animal), combined, and reduced in volume to 32 µl in a vacuum concentrator. Samples were mixed with 20 µg transfer RNA (tRNA) and 20 µg Herring Sperm DNA to prevent nonspecific hybridization, then mixed with 35 µl of modified "Genisphere" hybridization buffer (50% formamide, 40% 20x sodium chloride/sodium citrate (SSC), 9% Denhardt's solution, 1% sodium dodecyl sulfate [SDS]). This mixture was then applied to the arrays and allowed to hybridize overnight (16 h) at 45°C. After hybridization, arrays were washed in SSC/SDS buffers with descending stringency to remove any unhybridized or weakly (nonspecifically) hybridized cDNA's. Arrays were scanned using a Perkin–Elmer ScanArray Express, with laser power and photomultiplier gain varied to equalize fluorescence intensity between channels and to prevent oversaturation of signal intensity.
Microarray data analysis.
Data were extracted using ScanArray Express software (Perkin–Elmer). The median fluorescence intensity with background subtracted was imported into a MIAME compliant database (Brazma et al., 2001). GeneSpring (Agilent, Santa Clara, CA) microarray analysis software was used for further analysis. Data were LOWESS normalized (Draghici, 2003), and spots that did not meet a minimum signal intensity were removed. The resultant signal information was analyzed using one-way analysis of variance (ANOVA) (p = 0.05), assuming normality but not equal variances with a Benjamani–Hochberg correction for multiple comparisons (GeneSpring). GeneSpring's cross gene error model, which determines the likelihood of observing a specific fold change to the likelihood of observing a fold change of 1, was active during this test. A list of differentially expressed genes was prepared, comprised of those genes that demonstrated a statistically significant change in expression for each toxicant treatment (Draghici, 2003). A t-test (p < 0.05), with a Benjamani–Hochberg correction for multiple comparisons, was used to determine which of the chemical exposures caused significant alteration of the expression of a given gene. Gene Ontology (GO) terms were taken from the information available on the GRASP website http://web.uvic.ca/cbr/grasp/. These terms were assigned by members of the GRASP consortium as described on the website (von Schalburg et al., 2005a).
Gene expression data were further analyzed via Hierarchical Cluster Analysis and principle components analysis (PCA). Gene trees were created to group similar genes and allow for better visualization of the data (Butte, 2002). Genes that were found to be significantly different in expression in at least one treatment were used to create a gene tree using the GeneSpring package (Agilent). Different treatments were clustered using GeneSpring's Condition Tree function. Trees were created using a Pearson's correlation (Claverie, 1999) as a similarity measure, and branches that were more than 95% similar were merged. PCA was also performed on conditions (contaminant exposure treatments in this case) within Genespring, using the list of genes found to be significantly altered via ANOVA (p < 0.05) with a Benjamani–Hochberg multiple test correction. All genes in the PCA analysis were unweighted.
A BLAST database was constructed of all sequences in the Gene Ontology database (version 20070701, "seqdblite"). cDNA sequences representing clones on the microarray were BLASTed against these sequences, with an e-value threshold of 1e–15. A total of 7585 cDNA sequences had at least one hit against the GO sequences. For each contig, the GO terms associated with the highest-scoring blast hit were recorded. For each GO term, starting from the root of the GO hierarchy, the number of cDNA sequences associated with that term or any subterm was recorded.
XOA methodology and evaluation.
XOA, short for cross-ontological analytics, is an algorithm for computing similarities across GO codes, gene products, or both. Details on the XOA algorithm and evaluation can be found in Posse et al. (2006) and Sanfilippo et al. (2007). A detailed description of the user interface for this system is provided in Riensche et al. (2007). Several approaches can be used to integrate the textual evidence in XOA. Because the approach for this study is GO driven, GO terms were used as the common mode of comparison to augment the previous gene annotation made by the GO and BLAST analysis, and to correlate the biological responses between the perpetrator (the toxicants) and the effector (the expressed genes) in terms of processes, molecular functions, and signaling pathways.
The automated extraction of GO terms from GoPubMed was designed and implemented, using NCBI and GoPubMed web-services within the XOA tool. For extracting the GO terms of the toxicants, all possible synonyms (ethynylestradiol, ethinylestradiol, synthetic estrogen, EE2, Xenoestrogen, etc., were the list of synonyms for the toxicant, EE2) were used as search terms to extract all the abstracts referencing that toxicant. For the genes, the "gene description" from BLAST results was used, as the search term in GoPubMed. Two caveats with keyword-based search analysis for genes are (1) the genes with description terms as "unknown" or "hypothetical" cannot be included, and (2) splice variants of genes cannot be distinguished. The National Center for Biotechnology Information (NCBI) web-service implemented in XOA tool sends the resulting PubMed abstract IDs for each toxicant (or for each gene) to GoPubMed to retrieve a set of GO terms that had an accuracy percentage of 100%. These GO terms were then subjected to XOA analysis.
Quantitative real-time PCR.
RNA for quantitative real-time PCR (qRT PCR) was collected, purified, quantified, and stored as described in "RNA extraction." All qRT PCR analyses were performed using Applied Biosystems 7300 Real time system and the one step RT PCR master mix reagents (Applied Biosystems, Foster City, CA). Standards for each of the specific genes to be validated via qRT PCR were made either from the cDNA clones used to print the array (a kind gift from Dr Ben Koop, University of Victoria) or from previously isolated rainbow trout genes (J. A. Small, unpublished). Plasmids of genes to be used as standards were transcribed in vitro (Riboprobe system, Promega, Foster City, CA) and quantified via fluorometry (Ribogreen quantitation kit, Molecular Probes). Primers used for q RT PCR are given in Table 1. Transcription levels in treated and untreated fish were compared to a dilution series of the above standards. All measurements (samples and standards) were made in triplicate, and measurements were taken from three replicates for each treatment. All samples and standards were compared with a no reverse transcriptase control (to eliminate the possibility that signal resulted from DNA contamination), and each plate contained no template controls to serve as blanks. Data were normalized to expression levels of beta actin. Significance was determined via a Student's t-test (p < 0.05).
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
Toxicokinetics and Dosimetry
A summary of the administered doses and selected internal dose metrics is presented in Table 2. The measured and nominal values for the EE2 exposure water concentration were quite similar and averaged 102 ng/l during the exposures based on sampling done immediately before the fish were added and on day of termination. The total concentration of Cr was relatively low in plasma of the Cr-VI dosed trout and was similar to values observed in control fish (Table 2). However, the levels in the liver were elevated approximately threefold. This pattern of disposition is consistent with previously published studies examining sublethal chromium water exposure in salmonids (Farag et al., 2006
; Patton et al., in press). Previous studies have identified the kidney as a site of Chromium accumulation (Farag et al., 2006). There was no EE2 or BDE-47 detected in any control trout samples. Comparisons of plasma and liver concentrations of the toxicants after single or mixture administration indicated no significant differences (Table 2). However, plasma levels of BDE-47 appeared to be higher, and liver concentrations lower, in fish exposed to the mixture as opposed to those trout only exposed to BDE-47. We decided to investigate this further by assessing the toxicokinetics of BDE-47 in a separate cohort of trout administered a single intra-arterial dose of BDE-47 with and without coexposure to EE2 and Cr-VI.
|
In Figure 2A, the BDE-47 plasma concentration–time profiles are shown for the BDE-47 only and coexposure with EE2 and Cr-VI treatment groups. The plasma profiles of BDE-47 appear to decline in a multiexponential manner, with three exponential phases discernable. The terminal elimination phase for both exposure groups appeared to begin between 24 and 48 h. However, plasma concentrations of BDE-47 in the coexposure group were consistently higher during the secondary and terminal elimination phases (beginning approximately 6–10 h after dosing). This effect is also reflected in the toxicokinetic parameters between the treatment groups, with the plasma AUC0
increased for BDE-47 when coexposed with the other contaminants (Table 3). Both the total body clearance (Clb) and steady-state volume of distribution (Vss) were decreased for BDE-47 in the mixture treatments (Table 3). Despite these changes in toxicokinetic parameters, there was no difference in elimination half-life and only slight differences in the liver concentration of BDE-47 in the trout exposed for the gene expression studies (Tables 2 and 3). Collectively, the toxicokinetic analysis indicates that a greater fraction of the administered BDE-47 dose is retained in the plasma compartment and less extravascular distribution is occurring. With regard to EE2, visual inspection of the profiles indicated that there was no difference in uptake or plasma levels between the respective treatment groups.
|
|
Gene Expression
Of the 16,042 cDNA spots on the chip, 13,346 (83%) were detectable in all four chemical exposures and were used for subsequent analysis. Exposure to each of the contaminants caused significant changes in gene expression, as shown in Figure 3, whereas a "split control" experiment found no genes with significant changes in expression (data not shown). Following exposure to BDE-47, 102 genes (0.76%) had significantly altered expression (ANOVA, p < 0.05), (Figure 3A, Supplementary Material Table S1). The gene products were altered in expression roughly 1.5- to 5-fold, 59 genes are downregulated and 43 are upregulated. Some of the individual genes with altered expression include glyceraldehyde-3-phosphatse and a ubiquitin conjugating enzyme, both upregulated fivefold, and vitellogenin, serum albumin 2, and a 14-kDa apolipoprotein, all downregulated roughly threefold.
|
Exposure to Cr-VI caused changes in the expression of 167 genes (1.25%), of which 37 were upregulated and 130 were downregulated (Fig. 3B, Supplementary Material Table S2). The dynamic range of changes in expression was greater for Cr-VI than for BDE, with genes being as much as 14-fold upregulated (differentially regulated trout protein, a protein with unknown function, putatively assigned to "reproductive processes" via XOA analysis) and 48-fold downregulated (WD repeat domain 39). Some of the other genes with altered expression following exposure to Cr-VI include precerebellin like protein (upregulated roughly threefold), complement component C3 (upregulated roughly twofold), and a DNA repair (RecA homolog Rad51) gene, upregulated 1.5-fold.
Exposure to EE2 altered the expression of 131 genes (0.98%) the majority of which were downregulated (Fig. 3C, Supplementary Material Table S3). Forty genes were upregulated, including traditional biomarkers of EE2 exposure, vitellogenin (upregulated sixfold), and a vitelline envelope protein (upregulated 10-fold). The gene with the greatest induction (15-fold increase) following EE2 induction is "differentially regulated trout protein" which has unknown function. The 90 downregulated genes include serum albumin 2 (downregulated 11-fold), lipid binding proteins (downregulated 10-fold), and alphaglobin (downregulated threefold).
The transcriptomic response to exposure to the contaminant mixture (shown in Fig. 3D, Supplementary Material Table S4) was roughly comparable to the response to any of the individual contaminants, as measured by the number of genes changed (161–1.2%) and the magnitude of altered expression (ranging from fivefold upregulated to 20-fold downregulated). Of the genes with altered expression, some genes, such as "differentially regulated trout protein" and Serum albumin 2, are altered in response to multiple contaminants. Differentially regulated trout protein is the most upregulated gene following both EE2 and Cr exposure, and is also the most upregulated gene following exposure to the mixture. Serum albumin 2 is downregulated following exposure to each of the individual components of the mixture (downregulated twofold following exposure to either BDE or Cr, downregulated roughly 15-fold following exposure to EE2), and is roughly eightfold downregulated following exposure to the mixture. Differential expression of genes between the mixture and single chemical exposures were observed for vitellogenin, which is altered roughly sixfold as measured by the microarray following exposure to EE2 and roughly twofold as measured by the microarray following exposure to the mixture. A DNA repair homolog was also differentially expressed between the single chemical exposure and the mixture. This gene is upregulated roughly fourfold following exposure to the mixture and 1.5-fold following exposure to Cr. Some genes, such as a glutamate receptor (upregulated 3.5-fold) and heat shock protein 71 (downregulated sevenfold), are only altered by exposure to the mixture and are not significantly altered by any of the individual contaminant exposures (Fig. 3).
The degree of overlap in gene expression between those genes altered by exposure to each component of the chemical mixture (individual exposures) and the chemical mixture itself are shown Table 4. Fifteen genes had consistently altered expression following each of the chemical treatments. As discussed previously, some genes are uniquely expressed in only one of the chemical treatments. There is at least some overlap in gene expression profile between the exposures to individual components of the mixture, but this is typically the smallest fraction of genes in each category. However, the greatest degree of overlap is between individual components of the mixture and exposure to the chemical mixture. These relationships show that the mixture profile is made up of contributions from the individual components, and suggest that the gene expression profiles are somewhat additive. However, because there are 47 genes uniquely expressed in the mixture, there may be unique modes of toxicity caused by exposure to multiple compounds. Genes that were uniquely downregulated in response to exposure to the mixture included cytochrome oxidase II, whereas those that were upregulated include a nicotinamide adenine dinucleotide (reduced) ubiquinone oxidoreductase subunit and glutathione S-transferase (GST) kappa. This appears to be an important observation as these genes are either directly associated with mitochondrial activities such as oxidative phosphorylation or in the case of GST-kappa, protect against oxidative injury in mitochondria and peroxisomes (Morel et al., 2004).
|
The function of genes with altered expression, as determined by their corresponding GO terms, is shown in Figure 4. For each treatment, the molecular function and the biological process in which the genes are involved are plotted (as a percent contribution to the total number of genes altered by each chemical treatment). How the GO categories of genes in the toxicant responsive gene expression profile compares with the representation of that functional category on the microarray as a whole is shown in Table 5. The biological function group "transport" was always among the largest groups. This group was disproportionately affected and it only comprises 3% of all genes on the chip. Nonetheless, exposure to each toxicant and to the toxicant mixture caused changes in both individual genes and in functional groups of genes. Following exposure to BDE, genes with structural or calcium binding functions were upregulated (Fig. 4A) relative to their distribution on the array, as were genes involved in cell adhesion, muscle contraction and protein biosynthesis (Fig. 4B). Genes in GO functional categories that bind to DNA or had hydrolase activity were upregulated following exposure Cr in proportion to their abundance on the array (Fig. 4C), as were genes involved in transcription (Fig. 4D). Lipid binding genes were proportionally downregulated after exposure to Cr (Fig. 4C). Genes in GO categories lipid, protein, or oxygen binding functions were downregulated following exposure to EE2 (Fig. 4E), as were those genes involved in the inflammatory response and blood coagulation (Fig. 4F). Genes involved in cytoskeleton organization were upregulated relative to their abundance on the microarray (Fig. 4E).
|
|
When the GO profiles generated by exposure to the contaminant mixture are examined, some changes in functional categories that resemble the individual contaminants are evident, because other changes are unique to the mixture itself. For instance, exposure to the mixture downregulates genes with functions (as determined by their associated GO terms) in either lipid, protein, or oxygen binding (Figs. 4E and 4G), as does exposure to EE2. The upregulation of genes with hydrolase activity and transcription functional categories following exposure to the mixture resembles the pattern of change in expression following exposure to Cr (Figs. 4C and 4G). Both exposure to BDE and the contaminant mixture cause downregulation in the genes involved in protein biosynthesis (Figs. 4B and 4H). The downregulation in copper binding genes appears unique, for instance, as does the downregulation in genes involved in mitochondrial electron transport and metabolism (Figs. 4G and 4H). There are also some signals present in the individual exposure that are absent from the mixture exposure. The upregulation of structural and calcium binding genes observed following exposure to BDE is not apparent in the mixture exposure. Genes that are downregulated uniquely following exposure to the mixture have functions in protein biosynthesis and adenosine triphosphate (ATP) synthesis, whereas upregulated are involved in protein cycling, and oxidative stress.
In all cases, the functional group with the greatest number of genes is the unknown group, often comprising more than 60% of altered genes. The proportion of genes with unknown function is overrepresented in the gene expression profiles relative to the number of unknown genes on the array (52%).
An overview of the results of the XOA analysis where functional categories have been assigned to genes with unknown functions is provided in Table 6. XOA provides functional annotation for 401 of the 561 differentially expressed genes, whereas the GO database provides GO annotations for only 212 of them. As explained in the Methods section, 160 genes were not used for XOA analysis as are described as "hypothetical" or "putative" or "unknown." A gene can have one or more molecular functions, be used in one or more biological processes and may be associated with one or more cellular components. Genes were typically assigned multiple functions, as shown in Tables S5–S9. For subsequent analysis, genes were assigned a single function by simple weighting of each occurrence.
|
Hierarchical clustering analysis was used to generate the gene tree depicted in Figure 5A. This analysis suggests that the gene expression profile resulting from EE2 exposure most closely resembles the profile generated by exposure to the mixture, and BDE exposure contribute to the mixture profile least. The mixture exposure is the most dynamic of all contaminant treatments, with a higher portion of genes showing alteration in expression. The distribution of genes with similar function (as defined by either GO ontology or by XOA analysis) was not always in tight clusters. As shown in Figure 5B, many of the genes assigned by XOA analysis to have a role in apoptosis appear in distinct regions of the gene tree, yet other apoptosis related genes are seemingly isolated. Similar trends are seen for genes with involved in reproductive processes (Fig. 5C), in inflammatory response (Fig. 5D), in transporter activity (Fig. 5E), and genes with unknown function (Fig. 5F). These pathways were chosen to be illustrative of overall trends in gene distribution. One interesting trend does emerge from the data though; genes involved in energy pathways are clustered together (Fig. 5G), perhaps because they are consistently downregulated following exposure to the mixture.
|
When the data are clustered by PCA on conditions (Fig. 6), each of the treatment groups is plotted in three-dimensional space, which indicates the largest determinant of gene expression pattern is chemical treatment. The profiles resulting from exposure to BDE and EE2 are grouped most closely, suggesting that these two contaminants cause similar changes in gene expression. The gene expression profile resulting from the exposure to the mixture is intermediate between the other points, suggesting that each component of the mixture contributes to the gene expression profile resulting from exposure to the three compounds in combination.
|
The gene expression measurements obtained via q RT PCR (Fig. 7) show general agreement with the data obtained via microarray analysis. The q RT PCR data were normalized to the expression of beta actin, which was unaffected by any of the treatments (t-test, p> 0.05, data not shown). The data generally agree as to direction and significance of the change in direction, but not to the scale of alteration. For most of the genes measured, the magnitude of change in expression measured via q RT PCR (an absolute measure of changes in gene expression) was greater than via microarray (a semiquantitative measure of gene expression). For instance, the array reported vitellogenin upregulated almost 6 fold and the vitelline envelope protein upregulated roughly 10-fold following exposure to EE2, because the q RT PCR measurement reported these genes as being upregulated greater than 10,000 and 100-fold, respectively (Figs. 7H and 7I). Other discrepancies between the q RT PCR and the microarray measurements include the estrogen receptor (ER), which is not changed following exposure to any of the chemical treatments as measured via the microarray yet is significantly (t-test, p < 0.05) upregulated following exposure to both EE2 and the mixture (Fig. 7D), ATPase II, which is upregulated following exposure to EE2 as measured via q RT PCR but not by microarray (Fig. 7B), and precerebellin, which is measured as downregulated in the mixture via microarray and upregulated via q RT PCR (Fig. 7F).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
This study was designed to compare the hepatic gene expression profiles in rainbow trout after exposure to a contaminant mixture to those profiles after exposure to a single contaminant. A male isogenic strain of rainbow trout was used to minimize differences due to individual variability, which can confound the results of array experiments (Oleksiak et al., 2002
). Exposure routes were chosen to ensure consistent delivery of dose to the liver among individuals within a single treatment and between the individual contaminant exposures and the mixture. Our prior experience with EE2 indicated simple and rapid absorption behavior during water exposures with rapid partitioning between plasma and the liver (Skillman et al., 2006). Technical difficulties associated with water exposure of poorly soluble contaminants such as BDE-47 or slowly absorbed contaminants such as Cr-VI were circumvented by the use of intravascular injection. Duration of the exposure was chosen such that either tissue contaminant levels had come to equilibrium relative to exposure (EE2) or allowed sufficient time for extravascular distribution to occur (BDE-47; Fig. 2A). Although Cr-VI toxicokinetics was not characterized, past studies by others indicate rapid distribution and elimination (Van Der Putte et al., 1980). The exposure levels for each contaminant were selected to be sufficient to allow measurement in the liver and plasma, yet avoid any overt toxicity. We also initially assumed that the toxicokinetic behavior of the individual contaminants would be the same when administered as a mixture. This does appear reasonable for EE2 and Cr-VI, however BDE-47 internal distribution appears to be shifted away from extravascular tissues during the mixture exposures, which would explain the higher plasma concentrations of BDE-47 measured after the single chemical exposures (Table 2). An explanation for this observation may be a consequence of vitellogenin induction that occurs during EE2 exposures (Skillman et al., 2006). Past studies in fish suggest that vitellogenin can bind significant quantities of lipophilic xenobiotics (Monteverdi and Di Giulio, 2000). Thus, we speculate that the increased plasma concentration of BDE-47 observed during the mixture exposure was a result of greater plasma protein binding due to the induction of vitellogenin caused by the EE2 exposure.
Because no genes were significantly altered in our "split control" experiment, we believe that the genes we identified to be altered following exposure to one or more chemical contaminants are truly altered in expression, and are not a result of any array artifact such as a labeling bias (Draghici, 2003). The array data generated a gene expression profile that was unique to each contaminant and to the mixture (Fig. 3). These genes could be linked to molecular function and biological process (Fig. 4), and may provide some insight into the mode of toxic action for both the individual exposures and the mixture.
The qRT PCR data and array data agree as to the direction and significance of any alterations in gene expression, but not to the magnitude of change. Other studies with the GRASP arrays have also found agreement between the array data and PCR validation (Hook et al., 2006a,b; Rise et al., 2004b; von Schalburg et al., 2005b). As discussed in a previous paper (Hook et al., 2006a,b), the differences between the two measures may arise from the semiquantitative nature of microarrays (Bartosiewicz et al., 2001) or from saturation of spots on the array (Lopez et al., 2004).
In our initial work with this array, we were surprised to find that the cDNA spot encoding the ER was not upregulated in the livers of male trout following exposure to EE2. Using our previously developed q RT PCR assay, we were able to measure upregulation of the ER
in RNA extracted from the same tissue (Hook et al., 2006a,b, 2007). We used a previously developed qPCR assay as opposed to one designed for the sequence on the array to ensure that we were measuring the correct sequence for the ER. The reason for this difference may be that the two assays were measuring two different isoforms of the trout ER (Hook et al., 2006a,b, 2007; Nagler et al., 2007). The sequence of the ER spotted an the array (Gene ID CA062026
[GenBank]
) is most similar to the GRE encoding the ER obtained from Oncorhynchus mykiss (NCBI BLAST, Altschul et al., 1990, accession number z16419, e value =1 x 10–157), described by Le Roux et al. (1993). The sequence targeted by the q PCR assay is most similar to the ER obtained from Oncorhynchus masou (accession number AAS92970) (NCBI BLASTX, e value = 3 x 10–75). It is now established that there are two subtypes of estrogen receptors in trout, ER and ERβ, both of which occur as two isoforms (ER1, ER2, ERβ1, ERβ2; Nagler et al., 2007). The ER has been shown to be inducible following estrogen exposure, whereas ERβ is basally expressed (Sabo-Atwood et al., 2004). Despite the annotation, the sequence on the array does not code for either subtype, but may instead code for an upstream portion of the gene regulatory element.
One of the clear results from this study is that distinct genomic profiles resulting from exposure to individual contaminants were evident in the profile resulting from the simple mixture. This finding is consistent with work from contaminated field sites. For instance, Williams et al. (2003) compared hepatic gene expression using a custom array made up of known biomarker genes in winter flounder collected from relatively unpolluted and polluted sites. The fish from the polluted estuary were chronically exposed to many environmental contaminants, yet a gene expression signature representative of exposure to polyaromatic hydrocarbons was evident (Williams et al., 2003). Recent studies using primary hepatocytes from trout to study chemical mixtures have also identified some genes representative of exposure to individual contaminants in the gene expression profiles resulting from exposure to a chemical mixture (Finne et al., 2007). However, that study emphasizes that many of the genes present in the profiles resulting from exposure to individual contaminants are absent from the mixture.
The changes in gene expression following exposure to individual contaminants are consistent with past findings previously published in the literature. We found upregulation in the three traditional biomarkers of xenoestrogen exposure, vitellogenin, vitelline envelope proteins (Arukwe et al., 2001), and the ER, as has been reported in other array studies (Larkin et al., 2002, 2003a,b). In our previous studies, we also found that immune function genes and oxygen binding proteins were downregulated in response to EE2 (Hook et al., 2007). Other researchers working with EE2 have also observed downregulation of immune function genes, and downregulation of these genes has been noted as a generalized contaminant stress response (Koskinen et al., 2004; Williams et al., 2003). 1-Cys peroxiredoxin, among the genes we found upregulated in response to Cr-VI, has been previously reported to be upregulated in response to Cr-VI in winter flounder (Chapman et al., 2004).
With regard to reproducibility between studies, the gene expression data presented in this study are not entirely consistent with the data we have presented in previous work (Hook et al., 2006a,b). This may be due in part to age differences of the fish (trout in this study were 8 months older) although both groups of trout were sexually immature. We also used different dose treatment levels of EE2 and have demonstrated that gene expression profiles vary with EE2 dose (Hook et al., 2006b). For Cr-VI and BDE-47, we used different doses, exposure routes (ip versus intervascular injection), and euthanized the animals at a different time following exposure, and we hypothesize that these differences caused the differences in gene expression profile. This suggests that age of the animal, dose, route of exposure, and duration of exposure should all be taken into account when gene expression studies are compared.
The hepatic gene expression profiles we obtained suggest that any toxicity we observed would reflect exposure to all three contaminants. No single contaminant dominated the transcriptomic profile of the mixture exposure (Table 4), as would be expected with a comparative response (Folt et al., 1999) and has been seen in other studies (Finne et al., 2007). Instead, all three individual exposures contributed some genes to the final pattern. However, some genes were expressed only following exposure to the contaminant mixture, and others were expressed only following exposure to one of the individual contaminants and were not found in the mixture's transcriptomic profile. If the response to the mixture were completely nonadditive or interactive, a dominance of the mixture gene expression profile by genes altered uniquely by the combination of individual components, as opposed to those with altered expression following exposure to the individual components, would be expected. In a field exposure, the additive nature of the gene expression profile will allow for the identification of contaminants contributing to any observed toxic response, because the component of the gene expression pattern unique to the chemical mixture would allow for the identification of any nonadditive mechanisms of toxicity. Interestingly, gene expression changes unique to the mixture suggest mitochondrial dysfunction was occurring. Disturbances in the normal expression of genes associated with oxidative phosphorylation and protection against lipid peroxidation lead us to speculate that exposure to the mixture was over stimulating production of reactive oxygen species and inducing mitochondrial oxidative stress. Two components of the mixture that are most likely to promote this response are EE2 and Cr-VI. Chromium is well established to cause oxidative stress in fish and mammals (Tagliari et al., 2004). Recent studies in rodents indicate estrogen exposure impairs hepatic mitochondrial function (Moreira et al., 2007), which may in part be due to receptor mediated changes in expression of genes associated with mitochondrial respiration (O'Lone et al., 2007). Thus, we propose that exposure to Cr-VI and EE2 in trout act synergistically to cause mitochondrial dysfunction, which in turn is likely to tax the fish's metabolic reserves more greatly than exposure to any chemical alone, such that a more pronounced generalized stress response is observed.
One of the chief findings from the analysis of the gene ontology annotations was that all of the exposures resulted in a large number of genes with significant changes in regulation that had unknown function. Typically, there was a greater proportion of unknown genes in the toxicant responsive expression profiles (up to 75%) than was present on the microarray as a whole (52%). Of the genes with known function, many fit into composite categories such as transport. Although "transport" is a broad grouping of many different biological functions, it is interesting to note that all four toxicants disproportionately downregulated genes in this GO category. Nonetheless, the GO annotations may provide potentially useful information about the mechanism of toxicity. The functional categories of genes altered by exposure under each scenario are unique, and differ from the proportions of genes found on the array as a whole, suggesting distinct modes of toxic action. Furthermore, the downregulation in genes encoding copper binding proteins and involved mitochondrial electron transport genes may suggest mechanisms of interactive effects in conferring toxicity among the three contaminants. Although our current efforts to use GO ontology to analyze the modes of action of these contaminants are currently hampered by the high number of unknown genes, we are confident that developing techniques such as the XOA analysis, described in Sanfilippo et al. (2007), will be useful in adding clarification to current annotations once this approach is fully developed.
A limitation of using GO to characterize the transcriptomic response modes these contaminants is the high number of unknown genes. To improve the GO analysis, we sought to include textual information that is specific to the contaminants and mode of action. Although the XOA analysis shows great promise for the interpretation of gene function, the technique is not without its limitations. Because it is a text-based analysis tool, genes with no annotation (such as the many "unknown" genes on the GRASP array) cannot be further annotated using this technique. Also, as shown in the Supplementary Tables S5–S8, multiple, often conflicting, functions are typically assigned to each gene. Determining which function is likely to be relevant in a given situation and tissue is often not trivial, and is often somewhat subjective. Because this technique relies on published abstracts in PubMed, it must be applied with caution in nonmammalian models. For instance, the EE2 responsive genes were putatively assigned to "menstruation," "pregnancy," and "lactation" (see Table S5), processes that do not occur in fish. Also, functions were consistently assigned to apoptotic, reproductive and inflammatory pathways. It is possible that these assignments were made because these pathways are well studied by the biomedical community and consequently are well represented in PubMed. None the less, the technique allows researchers to form testable hypothesis about putative gene function, which is invaluable in determining toxicological mode of action and can inform traditional molecular biology and toxicology experiments.
It would be expected that genes with similar functions would cluster together (e.g., Draghici, 2003). We did not consistently observe this, either with genes whose functional identity was observed via XOA analysis (Figs. 5B and 5C) or with genes whose identity was assigned by GO. One possible reason for our observation may be the high number of genes with unknown function. As depicted in Figure 5F, these genes are distributed in almost every hierarchical cluster. It could be speculated that if the functions of some of these genes were resolved, the groupings of some functional categories may seem more coherent. The one coherent cluster that was formed, in the energy pathway genes, does have unknown genes (those in Fig. 5G without GO following the Gene ID). Another potential contributing factor for this lack of distinct clusters is that the functional groupings are very coarse for both the XOA analysis and the gene ontology terms. It is theoretically possible that two genes that are both identified as being involved in "reproductive processes" or "lipid binding" have distinctly different functions, and as a consequence, different gene expression profiles.
The PCA results seem to provide more information about the relationships between treatments. Hypothetically, if the gene expression profile resulting from exposure to all three contaminants as a mixture were to contain a high number of genes with altered expression not observed to be altered following exposure to any individual component of the mixture, the mixture exposure would plot by itself in space. In this study, the mixture gene expression profile is located between all other contaminant exposures, suggesting that each treatment contributes to the mixture, showing that the gene expression profile resulting from exposure to the chemical mixture has contributions from all three of the individual components. One cautionary note on interpreting PCA data is the contribution of neutrally expressed genes. Because all genes are given equal weight, two contaminants may be grouped together because the same sets of genes are not altered, despite having different modes of toxic action. We believe this is the explanation for the clustering of EE2 and BDE-47, which do not cause the same changes in gene expression. PCA is likely to be most predictive when used with small lists of highly significantly altered genes, as we did in this study.
Our data show that there are discernible changes in gene expression attributable to each component of a simple chemical mixture in the transcriptomic profile resulting from exposure to a contaminant mixture. There are also some genes that are expressed in the mixture treatment alone, suggesting the possibility for nonadditive toxicity. These findings show that microarrays may be a valuable tool for environmental monitoring, because they would allow regulators to identify the dominant toxicants in a mixed chemical exposure, and that microarrays may also be valuable to examine the toxicological mechanisms that cause interaction effects between contaminants.
|
SUPPLEMENTARY DATA |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
Supplementary data are available online at http://toxsci.oxfordjournals.org/.
|
FUNDING |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
U.S. Department of Energy under contract (DE-AC06-76RLO 1830), with Battelle PNNL, National Institute of Environmental Health Sciences grant (5R01ES012446-03) and National Science Foundation grant (0540693).
|
NOTES |
|---|
2 Present address: Genomic Medicine Institute, Cleveland Clinic (NE50), 9500 Euclid Avenue, Cleveland, Ohio 44195.
|
ACKNOWLEDGMENTS |
|---|
Arrays used in this project were obtained from the GRASP project http://web.uvic.ca/cbr/grasp/. We would like to thank B. Koop and G. Cooper for technical assistance, Rick Riensche and Antonio Sanfilippo for the XOA analysis.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONSUPPLEMENTARY DATAFUNDINGREFERENCES |
|---|
Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J. Mol. Biol. (1990) 215:403–410.[CrossRef][Web of Science][Medline]
Amin RP, Hamadeh HK, Bushel PR, Bennett L, Afshari CA, Paules RS. Genomic interrogation of mechanism(s) underlying cellular responses to toxicants. Toxicology (2002) 181–182:555–563.[CrossRef]
Ankley GT, Daston GP, Degitz SJ, Denslow ND, Hoke RA, Kennedy SW, Miracle AL, Perkins EJ, Snape J, Tillitt DE, et al. Toxicogenomics in regulatory ecotoxicology. Environ. Sci. Technol. (2006) 102(1):42–60.
Arukwe A, Kullman SW, Hinton DE. Differential biomarker and protein expressions in nonylphenol and estradiol-17β treated rainbow trout (Onchorynchus mykiss). Comp. Biochem. Phys. C (2001) 129:1–10.[CrossRef]
Bartosiewicz M, Penn S, Buckpitt A. Applications of gene arrays in environmental toxicology, fingerprints of gene regulation associated with cadmium chloride, benzo(a)pyrene and trichloroethylene. Environ. Health Perspect. (2001) 109:71–74.[Web of Science][Medline]
Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, Causton HC, et al. Minimum information about a microarray experiment (MIAME)—Toward standards for microarray data. Nat. Rev. Genet. (2001) 29:365–371.[CrossRef]
Butte A. The use and analysis of microarray data. Nat. Rev. Genet. (2002) 1:951–960.[CrossRef]
Chapman LM, Roling JA, Bingham LK, Herald MR, Baldwin WS. Construction of a subtractive library from hexavalent chromium treated winter flounder (Pseduopleuronectes americanus) reveals alterations in non-selenium glutathione peroxidases. Aquat. Toxicol. (2004) 67:181–194.[CrossRef][Web of Science][Medline]
Claverie J-M. Computational methods for the identification of differential and co-ordinated gene expression. Hum. Mol. Genet. (1999) 8:1821–1832.
Donnelly KC, Lingenfelter R, Cizmas L, Falahatpisheh MH, Qian Y, Tang Y, Garcia S, Ramos K, Tiffany-Castiglioni E, Mumtaz MM. Toxicity assessment of complex mixtures remains a goal. Environ. Toxicol. Pharmacol. (2004) 18:135–141.[CrossRef]
Draghici S. Data Analysis Tools for DNA Microarrays (2003) London: Chapman & Hall.
Eide I, Neverdal G, Thorvaldsen B, Arneberg R, Grung B, Kvalheim OM. Toxicological evaluation of complex mixtures: Fingerprinting and multivariate analysis. Environ. Toxicol. Pharmacol. (2004) 18:127–133.[CrossRef]
Farag AM, May T, Marty GD, Easton M, Harper DD, Little EE, Cleveland L. The effect of chronic chromium exposure on the health of Chinook salmon (Oncorhynchus tshawytscha). Aquat. Toxicol. (2006) 76:246–257.[CrossRef][Web of Science][Medline]
Finne EF, Cooper GA, Koop BF, Hylland K, Tollefsen KE. Toxicogenomic responses in rainbow trout (Onchorynchus mykiss) hepatocytes exposed to model chemicals and a synthetic mixture. Aquat. Toxicol. (2007) 81:292–303.
Folt CL, Chen CY, Moore MV, Burnaford J. Synergism and antagonism among multiple stressors. Limnol. Ocean (1999) 44:864–877.
Gibaldi M, Perrier D. Pharmacokinetics (1982) Marcel Dekker, Inc. New York. p.487.
Hook SE, Skillman AD, Small JA, Schultz IR. Gene expression patterns in rainbow trout, Onchorynchus mykiss, exposed to a suite of model toxicants. Aquat. Toxicol. (2006a) 77:372–385.[CrossRef][Web of Science][Medline]
Hook SE, Skillman AD, Small JA, Schultz IR. Dose response relationships in gene expression profiles in rainbow trout, Onchorynchus mykiss exposed to ethinylestradiol. Mar. Environ. Res. (2006b) 62:S151–154.[CrossRef][Web of Science][Medline]
Hook SE, Skillman AD, Small JA, Schultz IR. Temporal changes in gene expression in rainbow trout exposed to ethynyl estradiol. Comp. Biochem. Phys. (2007) 145:73–85.
Koskinen H, Pehkonen P, Vehniainen E, Krasnov A, Rexroad C, Afanasyev S, Molsa H, Oikari A. Response of rainbow trout transcriptome to model chemical contaminants. Biochem. Biophys. Res. Commun. (2004) 320:745–753.[CrossRef][Web of Science][Medline]
Larkin P, Folmar LC, Hemmer MC, Poston AJ, Denslow ND. Expression profiling of estrogenic compounds using a sheepshead minnow cDNA Macroarray. Environ. Health Perspect. (2003b) 111:839–846.[Web of Science]
Larkin P, Folmar LC, Hemmer MJ, Poston AJ, Lee HS, Denslow ND. Array technology as a tool to monitor exposure to fish xenoestrogens. Mar. Environ. Res. (2002) 54:395–399.[CrossRef][Web of Science][Medline]
Larkin P, Sabo-Attwood T, Kelso J, Denslow ND. Analysis of gene expression profiles in largemouth bass exposed to 17-B-estradiol and to anthroprogenic contaminants that behave as estrogens. Ecotoxicology (2003a) 12:463–468.[CrossRef][Web of Science][Medline]
Le Roux MG, Theze N, Wolff J, Le Pennec JP. Organization of a rainbow trout estrogen receptor gene. Biochim. Biophys. Acta (1993) 1172:226–230.[Medline]
Lopez F, Rougemont J, Loriod B, Bourgeois A, Loi L, Bertucci F, Hingamp P, Houlgatte R, Granjeaud S. Feature extraction and signal processing for nylon DNA microarrays. BMC Genomics (2004) 5:38.[CrossRef][Medline]
Monteverdi GH, Di Giulio RT. In vitro and in vivo association of 2,3,7,8-tetrachlorodibenzo-p-dioxin and benzo[a]pyrene with the yolk-precursor protein vitellogenin. Environ. Toxicol. Chem. (2000) 19:2502–2511.[CrossRef][Web of Science]
Moreira PI, Custodio JBA, Nunes E, Moreno A, Seica R, Oliveira CR, Santos MS. Estradiol affects liver mitochondrial function in ovariectomized and tamoxifen-treated ovariectomized female rats. Toxicol. Appl. Pharmacol. (2007) 221:102–110.[CrossRef][Web of Science][Medline]
Morel F, Rauch C, Petit E, Piton A, Theret N, Coles B, Guillouzo A. Gene and protein characterization of the human glutathione S-transferase kappa and evidence for a peroxisomal localization. J. Biol. Chem. (2004) 279:16246–16253.
Muirhead EK, Skillman AD, Hook S, Schultz IR. Oral exposure of PBDE-47 in fish: Toxicokinetics and reproductive effects in Japanese medaka (Oryzias latipes) and fathead minnows (Pimephales promelas). Environ. Sci. Technol. (2006) 40:523–528.[Medline]
Nagler JJ, Cavileer T, Sullivan J, Cyr DG, Rexroad C. The complete nuclear estrogen receptor family in the rainbow trout: Discovery of the novel ER alpha 2 and both ER beta isoforms. Gene (2007) 392:164–173.[CrossRef][Web of Science][Medline]
Oleksiak MF, Churchill GA, Crawford DL. Variation in gene expression within and among natural populations. Nat. Genet. (2002) 32:261–266.[CrossRef][Web of Science][Medline]
O'Lone R, Knorr K, Jaffe IZ, Schaffer ME, Martini PGV, Karas RH, Bienkowska J, Mendelsohn ME, Hansen U. Estrogen receptors alpha and beta mediate distinct pathways of vascular gene expression, including genes involved in mitochondrial electron transport and generation of reactive oxygen species. Mol. Endocrinol. (2007) 21:1281–1296.
Pakdel F, Metivier R, Flouriot G, Valotaire Y. Two different estrogen receptor (ER) isoforms with different estrogen dependencies are generated from the trout ER gene. Endocrinology (2000) 141:571–580.
Patton G, Dauble D, McKinstry C. Evaluation of early life stage fall chinook salmon exposed to hexavalent chromium from a contaminated groundwater source. Environ. Monit. Assess (2007) 133:285–295.[CrossRef][Web of Science][Medline]
Posse C, Sanfilippo A, Gopalan B, Riensche R, Beagley N, Baddeley B. Cross-Ontological Analytics: Combining associative and hierarchical relations in the Gene Ontologies to assess gene product similarity. In: Proceedings of International Workshop on Bioinformatics Research and Applications (2006) Reading, U.K.
Riensche RM, Baddeley BL, Sanfilippo A, Posse C, Gopalan B. (2007) XOA: Web-Enabled Cross-Ontological Analytics, IEEE International Workshop on Service Oriented Technologies for Biological Databases and Tools, July 2007. Utah, USA.
Rise ML, Jones SRM, Brown GD, von Schalburg KR, Davidson WS, Koop BF. Microarray analyses identify molecular biomarkers of Atlantic salmon macrophage and hematopoietic kidney response to Piscirickettsia salmonis infection. Physiol. Genomics (2004b) 20:31–35.
Rise ML, von Schalburg KR, Brown GD, Mawer MA, Devlin RH, Kuipers N, Busby M, Beetz-Sargent M, Alberto R, Gibbs AR, et al. Development and application of a Salmonind EST database and cDNA microarray: Data mining and interspecific hybridization characteristics. Genome Res. (2004a) 14:478–490.
Sabo-Atwood T, Kroll KJ, Denslow ND. Differential expression of largemouth bass (Micropterus salmoides) estrogen receptor isotypes by alpha, beta and gamma by estradiol. Mol. Cell. Endocrinol. (2004) 218:107–118.[CrossRef][Web of Science][Medline]
Sanfilippo A, Posse C, Gopalan B, Tratz S, Gregory M. Combining hierarchical and associative gene ontology relations with textual evidence in estimating gene and gene product similarity. IEEE Trans. NanoBiosci (2007) 6(1).
Schultz IR, Hayton WL. Influence of body fat on trifluralin toxicokinetics in rainbow trout (Oncorhynchus mykiss). Environ. Toxicol. Chem. (1997) 16:997–1001.[CrossRef][Web of Science]
Schultz IR, Orner G, Merdenik JL, Skillman A. Dose-response relationships and pharmacokinetics of vitellogenin in rainbow trout after intravascular administration of 17alpha-ethynylestradiol. Aquat. Toxicol. (2001) 51:305–318.[CrossRef][Web of Science][Medline]
Schultz IR, Skillman A, Nicolas JM, Cyr DG, Nagler JJ. Short-term exposure to 17 alpha-ethynylestradiol decreases the fertility of sexually maturing male rainbow trout (Oncorhynchus mykiss). Environ. Toxicol. Chem. (2003) 22:1272–1280.[CrossRef][Web of Science][Medline]
Skillman AD, Nagler JJ, Hook SE, Small JA, Schultz IR. Dynamics of ethynylestradiol exposure in rainbow trout (Oncorhynchus mykiss): Absorption, tissue distribution and hepatic gene expression patterns. Environ. Toxicol. Chem. (2006) 25:2297–3005.
Soetaert A, Moens LN, Van de Ven K, Van Leemput K, Naudts B, Blust R, De Coen WM. Molecular impact of propiconazole on Daphnia magna using a reproduction-related cDNA array. Comp. Biochem. Phys. C (2006) 142:66–76.
Tagliari KC, Vargas VMF, Zimiani K, Cecchini R. Oxidative stress damage in the liver of fish and rats receiving an intraperitoneal injection of hexavalent chromium as evaluated by chemiluminescence. Environ. Toxicol. Pharmacol. (2004) 17(3):149–157.[CrossRef]
Van Der Putte I, Lubbers J, Kolar Z. Effect of pH on uptake, tissue distribution and retention of hexavalent chromium in rainbow trout (Salmo gairdneri). Aquat. Toxicol. (1980) 1(1):3–18.[Medline]
Venier P, De Pitta C, Pallavicini A, Marsano F, Varotto L, Romuali C, Dondero F, Viarengo A, Lanfranchi G. Development of mussel mRNA profiling: Can gene expression trends reveal coastal water pollution? Mutat. Res. (2006) 602:121–134.[Web of Science][Medline]
von Schalburg KR, Rise ML, Brown GD, Davidson WS, Koop BF. A comprehensive survey of the genes involved in maturation and development of the rainbow trout ovary. Biol. Reprod. (2005a) 72:687–99.
von Schalburg KR, Rise ML, Cooper GA, Brown GD, Gibbs AR, Nelson CC, Davidson WS, Koop BF. Fish and chips: Various methodologies demonstrate utility of a 16,006-gene salmonid microarray. BMC Genomics (2005b) Sep 15:6–126.
Williams TD, Gensberg K, Minchin SD, Chipman JK. A DNA expression array to detect toxic stress response in European flounder (Platichtys flesus). Aquat. Toxicol. (2003) 65:141–157.[CrossRef][Web of Science][Medline]
Williams TD, Diab AM, George SG, Godfrey RE, Sabine V, Conesa A, Minchin SD, Watts PC, Chipman JK. Development of the GENIPOL European Flounder (Platichthys flesus) microarray and determination of temporal transcriptional responses to cadmium at low dose. Environ. Sci. Technol. (2006) 40:6479–6488.[Medline]
Wintz H, Yoo LJ, Longuinov A, Wu YY, Steevens JA, Holland RD, Beger RD, Perkins EJ, Hughes O, Vulpe CD. Gene expression profiles in fathead minnow exposed to 2,4-DNT: Correlation with toxicity in mammals. Toxicol. Sci. (2006) 94:71–82.
Yamaoka K, Nakagawa T, Uno T. Statistical moments in pharmacokinetics. J Pharmacokinet Biopharm (1978) 6:587–558.[CrossRef]
Yang RSH, El-Masri HA, Thomas RS, Dobrev ID, Dennison JE Jr, Bae D-S, Campain JA, Liao KH, Reisfeld B, Andersen ME, et al. Chemical mixture toxicology: From descriptive to mechanistic, and going on to in silicio toxicology. Environ. Toxicol. Pharmacol. (2004) 18:65–81.[CrossRef]
Young WP, Wheeler PA, Fields RD, Thorgaard GH. DNA fingerprinting confirms isogenicity of androgenetically-derived rainbow trout lines. J. Hered. (1996) 87:77–81.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||