WO2007113204A2

WO2007113204A2 - Method for detecting and classifying endocrine disruptor compounds

Info

Publication number: WO2007113204A2
Application number: PCT/EP2007/053021
Authority: WO
Inventors: Wim De Coen; Lotte Moens; Jurgen Del-Favero
Original assignee: Vib Vzw; Universiteit Antwerpen
Priority date: 2006-03-30
Filing date: 2007-03-29
Publication date: 2007-10-11
Also published as: WO2007113204A3

Abstract

The present invention relates to a method for detecting and classifying substances having endocrine disruption activity. More specifically, the invention relates to a limited set of genes from common carp (Cyprinus carpio) that are useful in the detection of environmental contaminants, or in the screening of uncharacterized existing or novel chemicals for possible endocrine disruption activity.

Description

METHOD FOR DETECTING AND CLASSIFYING ENDOCRINE DISRUPTOR COMPOUNDS

The present invention relates to a method for detecting and classifying substances having endocrine disruption activity. More specifically, the invention relates to a limited set of genes from common carp (Cyprinus carpio) that are useful in the detection of environmental contaminants, or in the screening of uncharacterized existing or novel chemicals for possible endocrine disruption activity

The potential of a variety of environmental compounds to modulate the endocrine system, and thereby interfere with the embryonal development, physiology, and reproduction of different wildlife species is of major concern in ecotoxicology (Colborn and Clement, 1992; Kavlock et al., 1996). As some of these chemicals are shown to be present in rainwater, lakes, rivers and seas (Allen et al. 1999; Belfroid et al. 1999; Harries et al. 1996; Tanghe et al., 1999), special attention should be paid to the effects of these endocrine disrupting compounds (EDCs) in aquatic organisms. During the past years, the U.S. Environmental Protection Agency (EPA) convened several international workshops in which the need for screening methods to identify endocrine disruptors was emphasized (Ankley et al., 1997; Kavlock et al., 1996). The development of such screening tools is of immense importance considering the large number of chemicals (-87.000), which are now in commercial use, but for which there are currently no data available about potential endocrine disrupting effects (EDSTAC final report, 1998; Fisher, 2004). Until now, however, no consensus has been reached concerning the appropriate toxicological assays that should be applied to assess the endocrine disruptive effect of chemicals and waste discharges. The development of such test methods is largely hampered by the complexity of the endocrine system - EDCs can act through a variety of different mechanisms and pathways, interacting with each other, and with numerous internal factors, such as signal transduction, hormonal synthesis, release, transport, metabolism, binding, action, or elimination (Kavlock et al., 1996). Moreover, in aquatic organisms most of these possible interfering pathways are still poorly understood; research in this area was until recently mainly focused on estrogenic responses, and assays for the detection of endocrine disruption in aquatic organisms were predominantly based on the detection of vitellogenin (Vtg) in male fish (Sumpter and Jobling, 1995). Vtg is an egg yolk precursor protein that is normally present in high concentrations in female fish undergoing oogenesis, whereas in males very little (if any) Vtg can be detected (Copeland et al., 1986). Although the Vtg gene is normally silent in males, its expression can be induced by exposure to - both natural and synthetic - estrogens (Bromage et al., 1988; Jobling et al, 1995; Sumpter and Jobling, 1995). Vtg expression in male fish has therefore become a very powerful biomarker for estrogen exposure, but its use in the field of endocrine disruption remains limited to the study of estrogenic responses, and fails to deal with the influence of xenobiotics on any other hormonal pathway, such as the androgen, thyroid and corticosteroid axes. Furthermore, it is evident that Vtg induction alone can not explain the wide range of effects found in the environment, as there is only a poor relationship between Vtg production and any adverse effects occurring in the environment (Jones et al. 2000; Giesy et al. 2000; Van Der Kraak et al. 2001 ). The complex nature of the endocrine system, and the diversity of cellular processes involved in hormonal communication, indicate that a variety of unknown, and yet unstudied factors may also play an important role in the disruption of the hormonal control. An excellent way to get a more complete picture of the impact of EDCs, is to map their effects at the molecular level (Denslow et al. 2001 ; Petit et al. 1999; Moens et al., 2003). This can be done by investigating the differential gene expression induced upon exposure to the chemical of interest. There are different ways to study differential gene expression; the so-called 'open' systems (such as Differential Display and Suppression Subtractive Hybridization) require no prior knowledge of the genes contained within the study model. These techniques provide a coherent platform for discovering novel genes, by constructing cDNA libraries consisting of genes that are differentially expressed between two cDNA populations of interest. 'Closed' systems (such as DNA arrays or Real Time PCR) on the other hand, utilize previously identified gene sequences. The DNA array technology for instance, offers the potential to measure the expression of thousands of genes simultaneously, providing a superior screening method which is able to deal with the diversity of processes involved in endocrine disruption. Evidently, the power of DNA arrays greatly depends on the genes comprised in the system; as there is no - or a very limited - prior knowledge of the genes involved in endocrine disruption, the development of a useful DNA array requires that also previously unexplored genes can be included. Several patent applications disclose nucleic acid arrays for toxicity assessment. WO9713877 describes a method for assessing the toxicity of a compound in a test organism by measuring gene expression profiles of selected tissues. WO0026404 describes more specifically a method for detecting a gene affected by an endocrine disruptor by preparing imRNA from cells, tissues or organisms that have been brought in contact with the disruptor, and hybridizing the nucleic acid with a DNA array of genes which may be affected by the endocrine disruptor. JP2005312456 describes a microarray of genes of which the expression is affected by a chemical substance having an activity similar to the activity of estrogen. However, although these systems may be used to detect endocrine disruptor compounds (EDCs), they have the disadvantage that a large set of genes is needed, and none of the systems is capable of classifying the activity of the EDC.

Surprisingly we found that it is possible to elucidate the molecular mechanisms and the effects of endocrine disruptors by combining an open and closed system. A DNA array for the detection of endocrine disruption in common carp (Cyprinus Carpio) has been elaborated. Suppression Subtractive Hybridization PCR (SSH) (Diatchenko et al., 1996) was used to isolate a set of relevant genes, consisting of gender-related and hormone-responsive genes from liver. cDNA macroarrays were developed from the hepatic gender-associated genes, and the expression of a limited set of this genes is capable of detecting the presence of an EDC, and classifying the compound.

A first aspect of the invention is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, comprising at least 8 sequences selected from the group consisting of SEQ ID N° 1 - SEQ ID N° 40. An array as used here can be any kind of solid support on which nucleic acid can be fixed, and includes, but is not limited to membranes such as nitrocellulose or nylon, microbeads and rigid solid supports such as glass. Arrays include both macroarrays and microarrays. Preferably, said array is a macroarray. Nucleic acid as used here can be any nucleic acid, including but not limited to RNA, cDNA, or single stranded DNA. EDCs are known to the person skilled in the art; the terms refers to chemical substances that are released in the environment or may be released in the environment (in case of chemicals subjected to a safety assessment) which show hormone-like activities or anti-hormone like activities. Classifying as used here means that the compound can be assigned to a disruptor class, such as, but not limited to androgen-like, thyroid-like or cortisol-like, but it means also that distinction can be made between a compound belonging to an EDC-class, and one not belonging to that class (such as, but not limited to non-androgen-like, non-thyroid-like or non-cortisol-like.) A preferred embodiment is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 4, SEQ ID N° 5, SEQ ID N° 11 , SEQ ID N° 12, SEQ ID N° 13, SEQ ID N° 17, SEQ ID N° 18 and SEQ ID N° 24. Another preferred embodiment is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 2, SEQ ID N° 4, SEQ ID N° 5, SEQ ID N° 8, SEQ ID N° 11 , SEQ ID N° 13, SEQ ID N° 14, SEQ ID N° 16, SEQ ID N° 17, SEQ ID N° 24, SEQ ID N° 27 and SEQ ID N° 29. Still another preferred embodiment is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 3, SEQ ID N° 5, SEQ ID N° 7, SEQ ID N° 8, SEQ ID N° 10, SEQ ID N° 14, SEQ ID N° 17, SEQ ID N° 21 , SEQ ID N° 25, SEQ ID N° 28, SEQ ID N° 29 and SEQ ID N° 31. Still another preferred embodiment is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 7, SEQ ID N° 8, SEQ ID N° 13, SEQ ID N° 15, SEQ ID N° 17, SEQ ID N° 18, SEQ ID N° 20, SEQ ID N° 21 , SEQ ID N° 25, SEQ ID N° 26, SEQ ID N° 29 and SEQ ID N° 30. Yet another preferred embodiment is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 1 , SEQ ID N° 5, SEQ ID N° 6, SEQ ID N° 9, SEQ ID N° 10, SEQ ID N° 13, SEQ ID N° 19, SEQ ID N° 22, SEQ ID N° 23, SEQ ID N° 26, SEQ ID N° 29 and SEQ ID N° 30. Yet another preferred embodiment is a nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 2, SEQ ID N° 1 1 , SEQ ID N° 12, SEQ ID N° 15, SEQ ID N° 18, SEQ ID N° 21 , SEQ ID N° 32, SEQ ID N° 33, SEQ ID N° 34, SEQ ID N° 35, SEQ ID N° 36 and SEQ ID N° 37.

Another aspect of the invention is a method for detecting and/or classifying endocrine disruptor compounds, comprising: (1 ) contacting carp with the test material (2) extracting RNA from the carp that has been contacted with the test substance (3) hybridizing said RNA or its cDNA against a nucleic acid array according to the invention. The test material, as used here, can be a chemical compound needed to be tested on safety and/or endocrine disruptor activity. Alternatively, it is a water sample that is possible contaminated with an EDC. Contacting as used here may be any form of contact. As a non limiting example, the test material can be injected, or it can be diluted in the swimming water of the carp. Preferably, the test material is injected intraperitoneal^. imRNA is extracted from the treated carps shortly after the treatment, preferably between 0 and 100 hours after treatment, even more preferably at 24 hours and/or 96 hours after treatment. RNA may be used as such, or converted into cDNA for hybridization to the array. The nucleic acid might be amplified by PCR before hybridization. It is clear for the person skilled in the art that, in stead of DNA array, the expression of the genes may be checked by an alternative methods, such as by real time PCR or by multiplex PCR. Therefore, another aspect of the invention is a method for detecting and/or classifying endocrine disruptor compounds, comprising: (1 ) contacting carp with the test material (2) extracting RNA from the carp that has been injected (3) detecting the expression of at least 8 genes selected from the group genes comprising SEQ ID N° 1 - SEQ ID N° 40.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1: Hepatic male- and female-associated genes classified into functional categories Figure 2: hormonally controlled genes from liver, classified into functional categories. Both genes that were induced by the hormone exposures and genes that were repressed by these treatments are included in this graph.

Figure 3: Double logarithmic scatter plot of spot intensities after hybridization of identical RNA samples to two separate membranes. The data points cluster along a slope of 1 , as verified by linear regression analysis (R2= 0,8393, Slope = 1.004 ± 0.007912) (GraphPad Prism 4.00, GraphPad Software, San Diego, CA, USA). Ninety-five percent confidence intervals are shown.

Figure 4: Macroarrays demonstrating gene expression profiles for control E2-treated juvenile carp. Figure 5: Graphical representation of differentially expressed genes (as determined by student's Ttest and fold changes) after exposure to E2 for (a) 24h and (b) 96h Figure 6: Relative expression of Vtg 5' and Vtg 3' in cDNA samples from the E2 exposure experiment, as determined by quantitative real time PCR Figure 7: Two-dimensional graph showing the gene expression changes occurring in livers from fish treated with the 14 test compounds. A total of 267 genes with p value < 0.05 were shown to be regulated 2-fold or more by at least one of the compounds. The experimental treatments (y-axis) and genes (x-axis) were clustered using hierarchical clustering. Up- regulation of imRNA transcripts is represented by shades of red, and down-regulation by shades of green. Black means no change.

A, B, C & D: close-up views of some of the major expression similarities between (groups of) compounds.

Figure 8: (A) Distance matrix showing pairwise distances between compounds, based on expression data of a highly discriminatory subset of 12 genes. The distance between two compounds is expressed as the number of genes whose expression differed sufficiently between the compounds. All compounds can be distinguished by at least 3 genes. (B) Hierarchical clustering of the 12 discriminatory genes.

Figure 9: Dendrograms of a set of 14 reference compounds, using 5 different sets of "classifier" genes Figure 10: Expression changes of Vtg 5' and 3' end fragments determined by Real Time quantitative RT-PCR, using RNA samples from the 96h exposure, at the highest concentrations (500 ng/L, 500 μg/L, 2000 μg/L and 100 ng/L for E2, NP, BPA and EE2, respectively). (A) Expression ratios of estrogen-treated versus control fish, normalized to 18S rRNA, (B) Relative abundance of Vtg 5' end to Vtg 3' end mRNA, in control and exposed fish. Real time PCR experiments were repeated three times and their averages and standard deviations (bracketed) are shown.

EXAMPLES Materials and methods to the examples

Construction of gender-associated cDNA libraries using Suppression Subtractive Hybridization In a first SSH experiment, male liver RNA was subtracted against female liver RNA and vice versa. Sexually mature carp with a weight of approximately 30 g were obtained from a commercial fish supplier. Fish were acclimatized for 4 weeks in aerated aquaria filled with softened tap water, at a temperature of 20°C. They were exposed to ambient light and fed once a day, with Phoenix Pond sticks (Enterprise house, Pinxton, Notts, England). After decapitation, liver tissue was isolated and stored in liquid nitrogen. RNA was prepared using to the Totally RNA Isolation kit and RNA pools were made of 15 animals per gender. The RNA pools were treated with 1 U RNase-free DNase and 1 U RNase inhibitor per 30 μl sample (Fermentas, St. Leon-Rot, Germany), followed by a phenol/chloroform extraction. The concentration was determined by spectrophotometry, and RNA integrity was assessed by denaturing formamide- 101 agarose gel-electrophoresis. Suppression Subtractive Hybridization was performed as described by Moens et al. (2003). The subtracted libraries were cloned into a pGEM-T Easy vector (Promega Corporation, Madison, Wl, USA), and sequencing was performed by the VIB Genetics Service Facility (Flanders Interuniversity Institute for Biotechnology, Wilrijk, Belgium). The sequence data were handled by SSHSuite, a software package developed at the Bioinformatics unit of the molecular genetics research group, Flanders Interuniversity Institute for Biotechology, University of Antwerp (Belgium) (Weckx et al. 2004). The sequences were first trimmed for vector and adaptor sequences, and assembled into contigs. These contiguous sequences were compared to the National Center for Biotechnology 110 Information nr, nt, est, swissprot and month databases using BLASTn and BLASTx (Altschul et al. 1997). Hits with E < 1 e-5 for BLASTx and E < 1 e-20 for BLASTn were considered significant.

Construction of hormone-responsive cDNA libraries using Suppression Subtractive Hybridization

In a second set of SSH's, liver RNA from hormone-exposed fish was subtracted against control RNA, and vice versa.

Therefore, four separate exposure experiments were conducted, in which juvenile carp were intraperitoneal^ (ip) injected with the following model hormones: 17-beta-estradiol (E2), 1 1- ketotestosterone (11 KT), triiodothyronine + L-thyroxine (T3+T4) and hydrocortisone. Exposures were conducted according to the OECD guidelines for testing chemicals on biotic systems (nr. 203, acute toxicity test for fish). Juvenile fish were obtained from the University of Wageningen (The Netherlands). The fish were acclimated for 2 weeks in aerated tap water before hormone injection took place. Fish were exposed to a 14:10 h lightdark photoperiod and fed commercial feed (Antwerp Aquaria, Aartselaar, Belgium) at a ratio of 2% of the maximum body weight (estimated at 2,5 g). Groups of fish received a single ip dose of E2 (5 μg/g) (Sigma-Aldrich Corporation, Bornem, Belgium), 1 1-KT (5 μg/g) (Steraloids Inc, Newport R. I., USA), T3+T4 (5 μg/g each) (Sigma-Aldrich Corporation) or hydrocortisone (10 μg/g) (Sigma-Aldrich Corporation). E2, 1 1-KT and hydrocortisone were dissolved in corn oil, thyroid hormones in a physiological saline solution (0.9% NaCI) that was made basic with NaOH. Control fish were injected with vehicle control. 24h and 96h after injection, 15 hormone-treated and 15 control fish were decapitated and dissected. RNA was isolated from liver (as described above), and 2 SSH's were performed according to the following experimental set-up: a pool of liver RNA from fish treated with 1 1-ketotestosterone and hydrocortisone was prepared, and a pool containing RNA from fish exposed to 17-beta-estradiol and thyroid hormone. Furthermore, liver RNA isolated from control fish from the corresponding exposure experiments was pooled. Subtractions were performed between RNA from hormone-exposed fish versus control fish, as described above. Cloning, sequencing and homology searches were performed as described above.

Development of cDNA macroarrays 192 gender-associated cDNA inserts from liver corresponding to different gene fragments were amplified using the pGEM-T vector primers T7 (δ'-TAATACGACTCACTATAGGG-S') and SP6 (5'- TATTTAGGTGACACTATAG-S') (Eurogentec, Seraing, Belgium). cDNA of the Green Fluorescent Protein (GFP) was amplified with specific primers (5'- ATGGTGAGCAAGGGCGAGGAGC-S' and 5'- CTTGTACAGCTCGTCCATGCCG-3'). Amplified cDNA fragments were concentration-adjusted and spotted in triplicate on a nylon membrane filter (Hybond-N+, Amersham Biosciences, Roosendaal, The Netherlands) with a gridding robot (BIOMEK 2000, Beckman Coulter, Fullerton, CA, USA), using an eight-pin replication tool and a software package developed by the Group of Molecular genetics, department Plant genetics, Flanders Interuniversity Institute for Biotechnology, University of Ghent.

Each spot contained approximately 25 ng of cDNA. The membranes were denatured in 0,2 M NaOH and the DNA was cross-linked by UV radiation (250 mJoules) (Stratalinker, Stratagene, La JoIIa, CA, USA).

Radioactive target labeling, macroarray hybridizations & gene expression analysis

Macroarray hybridizations were performed for the exposure experiments with 17-beta-estradiol (5 μg/g) and hydrocortisone (10 μg/g) (described above), ds cDNA was synthesized starting from 1 μg DNase-treated (Fermentas) total RNA, using the SMART PCR cDNA Synthesis kit (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). 100ng cDNA was then labeled with [α-33P] dATP (2'-deoxyadenosine 5'-triphosphate) (Amersham Biosciences) via random priming. The specific activity of each target was determined using a beta-counter. The blots were prehybridized at 68°C during one hour. After prehybridization, the ds-labeled targets were denatured (5 min. at 95°C), and added to the filters, taking care that for each target the same amount of cpm's (counts per minute) was used. The blots were then hybridized overnight at 68°C. After hybridization, the filters were washed 4 x 20 min. both with low and high stringency wash solutions. More details on target labeling, prehybridization and hybridization procedures can be found in Moens et al. (2003). After hybridization, the membranes were exposed to a storage phosphor screen (PerkinElmer Life And Analytical Sciences, Boston, MA, USA) for different time periods (4h, 24h & 72h). Scanning of the screens with the Packard BioScience Cyclone Storage Phosphor Scanner (PerkinElmer Life And 170 Analytical Sciences) produced digitalized images that were analyzed using the Aϊda analysis software (Raytest, Straubenhardt, Germany). For each spot a local background was subtracted, and the values were normalized to a trimmed mean, i.e. spots are referred to the mean value of all spots, except for bad quality dots and dots comprising the upper and lower 5% intensity values. This normalization method was preferred above the standard housekeeping-gene approach, as these household-genes are often unexpectedly regulated, leading to misinterpretation of the expression values (Stoyanova et al., 2004; Kroll et al., 2002; Butte et al., 2001 ; Selvey et al., 2001 ). Gene array data were analyzed using linear regression (GraphPad Prism 4.00, GraphPad Software, San Diego, CA, USA) and student's T-test for independent samples (Statistica, StatSoft, Tulsa, OK, USA). Spots with fold differences higher than two, combined with a p-value < 0.05 (as calculated by the student's T-test for independent samples) were considered differentially expressed.

Real Time Quantitative PCR

To determine the expression level of the Vtg 5' end versus the 3' end in control and E2 exposed fish, a Real time PCR was carried out. In this experiment, the same RNA samples as used for the macroarray hybridizations - control (24h), E2 exposed (24h), control (96h) & E2 exposed (96h) - were purified using the RNEasy Mini kit (Qiagen, Venlo, The Netherlands). The concentration of the purified RNA was determined by spectrophotometry. For each sample, 1 μg of RNA was reverse transcribed in the presence of oligo dT primers using the Superscript First Strand cDNA Synthesis System for RT PCR (Invitrogen SA, Merelbeke, Belgium). PCR primers were designed using the LightCycler probe design software (Roche Diagnostics, Vilvoorde, Belgium). Primer sequences are listed in table 3. Real time PCR reactions (20 μl) were performed using the LightCycler FastStart DNA Masterplus SYBR Green I kit (Roche Diagnostics), as recommended by the manufacturer, and using a final primer concentration of 0.5 μM each. Amplification was detected using LightCycler System. Cycling parameters were 95 °C for 5 min, followed by 40 cycles of 10 sec at 95°C, 10 sec at 58 °C, and 12 sec at 72 °C. Melting curves were performed for all reactions to monitor the quality of amplicons and reactions. This was done by starting at 95°C, followed by 15 sec at 55 °C, and gradually heating with a temperature transition rate of 0.10 °C s-1 until a temp of 95°C was reached. The expression of the Vtg 5' end relative to the 3' end was reported as: 2^"ΔCt (Livak and Schmittgen, 2001 ), where C₁ represents the threshold cycle (mean from duplicate reactions), and ΔCT = CT vtg 5 -CT vtg 3 ■ Chemicals, exposure, and preparation of liver RNA

Fish were treated with 14 OECD-recommended reference chemicals, with different mechanisms of action (Table 1 ). All chemicals were obtained from Sigma Aldrich (Bornem, Belgium), except MT and 1 1 KT, which were purchased from Steraloids Inc. (Newport, Rhode Island, USA).

Juvenile fish were obtained from the University of Wageningen (The Netherlands). Fish were acclimated for 3 weeks in aerated OECD water prior to treatment. The fish were exposed to a 14:10 h lightdark photoperiod and fed commercial feed (Antwerp Aquaria, Aartselaar, Belgium) at a ratio of 2% of the maximum body weight (estimated at 1 ,2 g). Two dosing methods were applied; in a first series of experiments, fish received an aqueous exposure to three concentrations [A1 , A2, A3] of each substance, in parallel with a control [AO], using a semistatic experimental set-up (renewal at 48h). After 24h and 96h of exposure, liver tissue was dissected from 10 fish per exposure condition. Additionally, in a second series of experiments fish were intraperitoneal^ (ip) injected with a single dose of each compound, and dissected 24h post-injection. Control fish received an ip injection of the vehicle control, without any chemical. Concentrations for both delivery methods were based on literature data, and are summarized in table 1. RNA was prepared using to the Totally RNA Isolation kit (Ambion, Austin, Texas, USA), followed by a DNAse treatment (Fermentas, St. Leon-Rot, Germany) and a phenol/chloroform extraction. The concentration was determined by spectrophotometry, and RNA integrity was assessed by denaturing formamide-agarose gel-electrophoresis.

Analysis of expression changes of Vtg 5' and 3' end fragments RNA samples were purified using the RNEasy Mini kit (Qiagen, Venlo, The Netherlands), and the concentration of the purified RNA was determined by spectrophotometry. 1 μg of RNA was reverse transcribed in the presence of random hexamer primers using the Superscript First Strand cDNA Synthesis System for RT PCR (Invitrogen SA, Merelbeke, Belgium). PCR primers were designed using the LightCycler probe design software (Roche Diagnostics, Vilvoorde, Belgium). Primer sequences were as follows: Vtg 5' end, 5'- GGTGTGGTTGGAAAGG-3' and δ'-CGGTAATGTGGTTGGC-S'; Vtg 3' end, 5'- GTTCAGTCAGCTTTGCC-3' and δ'-GTGGTTCTGATTGGTGC-S'; 18S rRNA, 5'- ACAGGTCTGTGATGCC-3' and δ'-CCAATCGGTAGTAGCG-S'. Real time PCR reactions were performed in triplicate, using the LightCycler FastStart DNA Master^plJS SYBR Green I kit (Roche Diagnostics), according to the manufacturer's instructions, and using a final primer concentration of 0.5 μM each. Amplification was detected using LightCycler System. Cycling parameters were 95 °C for 5 min, followed by 40 cycles of 10 sec at 95°C, 10 sec at 58 °C, and 12 sec at 72 °C. Melting curves were performed for all reactions to monitor the quality of amplicons and reactions. Reactions were started at 95°C, followed by 15 sec at 55 °C, and gradually heated with a temperature transition rate of 0.10 °C s^"1 until a temp of 95°C was reached. Gene expression ratios were calculated as reported by Pfaffl (2001 ).

Fluorescent target labeling and microarray hybridizations

For the aqueous exposure experiments, hybridization series were performed using a loop- design, by which pairs of labeled cDNA from the three exposure concentrations of each compound were hybridized in the following way: A0-A1 , A2-A1 , A2-A3, A0-A3. Using this strategy, a replicate hybridization is built in for every exposure condition, creating more reliable data sets. Hybridizations for the ip injection experiments were carried out in duplo. cDNA microarrays were prepared by mechanical spotting of 960 endocrine-related gene fragments. Each cDNA clone was spotted four times on the slides, allowing replicate analyses. In addition, a set of artificial control genes, consisting of 10 calibration controls, 8 ratio controls and 2 negative controls, was spotted in 36 replicates (Lucidea Universal ScoreCard, Amersham Biosciences, Roosendaal, The Netherlands) on each array, to assist the evaluation of the quality of the array data. For fluorescent target labeling and hybridization protocols duplicate labeled cDNA targets were prepared by converting 7 μg DNase-treated total RNA into aminoallyl-dUTP labeled cDNA using the Superscript Il Reverse transcriptase kit (Invitrogen, Paisly, UK). Lucidea reference and test imRNA spikes, corresponding to the calibration and ratio controls spotted on the arrays, were added to the RNA samples from reference and test populations respectively. RNA was hydrolyzed and unincorporated nucleotides were removed (QiaQuick PCR purification kit, Qiagen, Crawley, UK). The aminoallyl-labeled cDNA samples were then covalently coupled to Cy3- (for AO and A2 cDNA populations) or Cy5- (for A1 and A3 cDNA populations) esters (Amersham Biosciences). Reaction mixtures were purified once more, and the labeling efficiency was determined by spectrophotometry. The threshold for optimal dye incorporation was 150 pmol and a frequency of incorporation of 20-50 was considered appropriate for hybridizations. Following analysis of incorporation, the fluorescently labeled samples were dried to completion in a vacuum centrifuge and resolved in hybridization solution. Microarray slides were prehybridized, and the target cDNA solution was denatured and added to the slides. Hybridization took place overnight (16-18hrs) at 42°C. After hybridization, slides were washed and dried with N₂. Finally, slides were scanned using the Genepix Personal 4100A scanner (Axon Intruments, Union City, CA, USA).

Data Acquisition, preprocessing and detection of differential expression

Scanned images were analyzed using Genepix pro 4.1 software (Axon Intruments), for spot identification, and quantification of raw fore- and background intensities of the spots.

Raw data were background corrected and normalized using the Variance Stabilization and Normalization method (VSN) in R (Huber et al., 2002). This method incorporates data normalization, a model for the dependence of the variance of the mean intensity, and a variance stabilizing data transformation. Subsequently, a linear model was used to compute the contrasts A0-A1 , A0-A2, and A0-A3 from the loop A0-A1 , A2-A1 , A2-A3, A0-A3, for each of the aqueous exposure experiments.

To determine differential gene expression, an empirical Bayes-test (a moderated t-test) was run, at a p-value of 0.05. Additionally, differentially expressed genes need to have an absolute M-value > 1 (2-fold change), and an A-value > 8 (with M = Iog2 (Cy5/Cy3) and A = Iog2 (Cy3^*Cy5)^1/2). For a gene to be "zero" (unaffected by the exposure), its A should still be > 8, and its absolute M-value < 0.3 (fold change 1.23). Using this combination of a cut-off fold change value and a statistical significance test as a criterion for differential gene expression, both the biological relevance and a more statistically sound approach were taken into account.

Clustering analysis and discriminatory gene selection

Prior to hierarchical cluster analysis, data sets were processed in the following way: for every compound, the expression value corresponding to the most effective treatment condition (i.e. the ip injection experiment or one of the different water exposure conditions) was selected for each individual gene. This means that it is possible that for one gene the most effective expression value was taken from e.g. the low concentration exposure, after 24h, whereas for another gene the high exposure level at 96h was more effective. By using this strategy, every single differentially expressed gene resulting from the different exposure conditions is included in the analyses, so that potential concentration- and/or time-related differences in effects can be taken into account.

Pearson's uncentered hierarchical clustering analyses were performed using the Acuity 4.0 software (Axon Instruments).

To build a diagnostic gene set, we aimed to find a subset of genes - out of the collection of differentially expressed fragments - that separated the compounds in an optimal way. For each gene, a quaternary value was computed: over-expressed (M strongly positive), under- expressed (M strongly negative), not-expressed (M around 0), or unclear (M intermediate). The expression of two genes was termed sufficiently different if:

1 ) both gene A and gene B have consistent behavior, i.e., the genes cannot have been measured over-expressed in one condition (exposure time, concentration) and under- expressed in another, AND 2) one of the following:

- gene A is overexpressed in some conditions, and gene B underexpressed in some conditions, or vice versa. - gene A is over- or under-expressed in some conditions, and gene B is not-expressed in some conditions, or vice versa. The distance between two compounds was then calculated as the number of genes whose expression differed sufficiently between the compounds. Using a generational genetic algorithm, subsets of 8 to 12 genes were selected such that the total number of pairwise small compound distances <3 was minimal. The stopping criterion was 100 generations without improvement. Population size was 20, mutation was applied on a per-individual basis with probability 0.5, cross-over was also applied with probability 0.5. Fifty independent runs were replicated. The resulting gene sets each corresponded to a distance matrix summing up all the distances for the different compound combinations.

Example 1: Construction of cDNA libraries using SSH

Three Suppression Subtractive Hybridizations were performed. The first SSH led to the isolation of gender-associated sequences from liver. Here, RNA isolated from male liver was subtracted from RNA from female liver and vice versa, generating two cDNA libraries per tissue, enriched for female- and male-associated gene fragments respectively.

Due to a relatively high redundancy in sequences, PCRs and sequencing reactions were performed only for a part of the clones isolated during the SSH procedure: 288 female- associated and 192 male- associated cDNA fragments were amplified and sequenced. The 288 female-associated genes could be reduced to a collection of 123 unique clones, of which 59 corresponded to known genes in the NCBI databases, 15 were homologue to unknown clones, and 49 showed no significant homology to the sequences in the databases. Of the 192 sequenced male-associated gene fragments, 107 were found to be unique. This set contained 22 genes corresponding to known genes in the databases, 23 to unknown clones, and 62 with no homology to any of the sequences in NCBI. After sequence homology searching, the full set of both male- and female- associated sequences (unknown and overlapping sequences included) was divided into functional categories, which are summarized in figure 1. In a second set of SSH's, hormone-related gene fragments from liver were isolated. To accomplish this, fish were exposed to 4 model hormones: 17-beta-estradiol (E2), hydrocortisone, 1 1- ketotestosterone (1 1 KT), and a mixture of triiodothyronine (T3) and L- thyroxine (T4). Since SSH is a relatively expensive and laborious method, it was decided to pool the RNA from the different exposure experiments in the following way: (1 ) RNA from fish exposed to E2 was pooled with RNA from fish treated with T3+T4, and (2) RNA from 11 KT- treated fish was mixed with RNA from fish exposed to Cortisol. Each time, the corresponding control RNA was pooled as well. SSH was then applied to isolate differentially expressed genes between control pools and (1 ) E2/T3+T4 exposed and (2) 1 1 KT/cortisol exposed RNA populations. SSH's were performed in both directions, resulting in four cDNA libraries - per SSH a cDNA bank consisting of genes that were upregulated by the hormone exposures, and a library with genes that were downregulated by these treatments. For each cDNA library, 96 clones were sequenced. The full set of sequences isolated from the different hormone- associated cDNA libraries (unknown and overlapping sequences included) was, after sequence similarity searching, divided into functional categories, which are summarized in figure 2. Each series of the graph contains both up- and downregulated genes.

The full collection of sequenced gene fragments resulting from the 3 SSH's was assembled into contigs to exclude overlapping of sequences. This resulted in 397 different gene fragments, of which 160 showed a significant homology to sequences in the NCBI databases, 65 were homologue to unknown clones, and 172 showed no homology at all to any known sequences.

Example 2: Macroarray experiments (performed with a set of liver genes)

To demonstrate the usefulness of the applied experimental set-up, the first set of isolated gene fragments - the gender-associated cDNA library from liver - was used to develop a cDNA macroarray. Therefore, 192 different gene fragments from this gene bank were spotted on a nylon membrane. These macroarrays were then used in hybridization experiments, to determine the gene expression pattern induced by estradiol and Cortisol exposures. Yet, to be able to compare signal intensity values between different macroarrays, we first had to determine the interarray variability. To accomplish this, two identical liver RNA samples were hybridized onto two separate membranes. After background subtraction and normalization of the signal intensities, the average value of the triplicate spots was calculated. Figure 3 shows a scatter plot correlating the log transformed average intensity values from the triplicate spots from the two membranes. Data points cluster along a slope of one (R2= 0,8393, Slope =1 ,004 ± 0.007912), with a stronger correlation towards the higher intensities, as could be expected (Richmond et al. 1999).

Subsequently, gene expression profiles induced by E2 and Cortisol exposures were determined. For both substances, hepatic RNA from 15 exposed and 15 control fish dissected at two different timepoints (24h and 96h) was converted to cDNA and radiolabeled. Next to the experimental targets, a radiolabeled target corresponding to the GFP gene was synthesized, and divided among the experimental probes to serve as an external hybridization control. The images obtained after hybridization and scanning are given in figure 4.

After background subtraction and normalization, gene array data were analyzed using student's T-test for independent samples. Spots with fold differences higher than two and a p- value < 0.05 (from triplicate measurements) were considered differentially expressed. Both the 24h and 96h exposure to E2 resulted in a set of differentially expressed genes, that are graphically represented in figure 5. Exposure to hydrocortisone on the other hand resulted in less affected genes. Example 3: Real Time Quantitative PCR

Since macroarray experiments revealed that the two termini of the Vtg gene responded in a quite different way to estrogen exposure, a real time, quantitative PCR was performed to determine the relative expression levels of the 5' versus the 3' end, within the same cDNA population, for both exposed and unexposed fish. To be able to compare the expression of one gene (or gene terminus) relative to another gene (terminus) - even though within the same cDNA population - it is important that the reactions amplify with similar efficiencies. The efficiency of each primer set correlates with the slope of the CT versus log cDNA (or reverse transcribed RNA) concentration plots, obtained by running serial dilutions of cDNA. The slopes of the CT versus log cDNA quantity were -3.29 for Vtg 5', and -3.83 for Vtg 3'. Alternatively, the efficiencies of two different amplicons may be considered as equal if the slope of the ΔC_T versus log cDNA concentration is approximately equal to 0 (Livak and Schmittgen, 2001 ). The slope of the ΔC_T (C_T vtg 5 - C_T vtg 3) versus log cDNA quantity was -0.01 , demonstrating that the efficiency of amplification of the two termini is similar, and that the expression of these transcripts can be directly compared with one another. A direct comparison of the expression of Vtg 5' relative to Vtg 3' imRNA was made for each of the cDNA populations from the E2 exposure experiment - control (24h), E2 exposed (24h), control (96h) & E2 exposed (96h) (figure 6). The results show that the expression of Vtg 5' is higher than the Vtg 3' expression level, in all of the cDNA samples tested - also in unexposed, control fish. However, the ratio between Vtg 5' and Vtg 3' differs between the different samples, being lowest in the control cDNA (24h exposure) (where Vtg 5' is ~3-fold more abundant than Vtg 3') and highest in the cDNA from 96h E2 exposed fish (where Vtg 5' mRNA was ~19-fold higher compared to Vtg 3' mRNA).

Example 4: Evaluation of gene expression profiles for 14 reference endocrine disrupting chemicals

In order to determine the toxicant-induced gene expression changes, liver RNA was collected

24h and 96h following aqueous exposure to the compounds, or 24h post-injection. Competitive hybridizations of fluorescently labeled cDNA targets derived from control versus treated livers were used to measure relative abundance of the probes on a custom carp microarray, containing 960 endocrine-related gene fragments. For each exposure condition, duplicate hybridizations were performed. We conducted statistical analysis of the microarray data and determined differential expression using an empirical Bayes-test (p<0.05), combined with a cut-off fold change of 2. The quality of individual microarrays was evaluated using the Lucidea Scorecard references. All microarrays included in the analysis fulfilled our prerequisite quality parameters. All treatments caused transcriptional changes with respect to their corresponding time-matched controls. It was noted that for a few compounds only a small number of gene expression responses were obtained for some of the treatment conditions, which was probably due to a sub-optimal concentration range, exposure time or exposure route. However, in those cases, data collected from the two alternative dosing methods complemented each other very well, such that for every compound an informative gene expression data set was obtained. In total, the two exposure series resulted in 267 different genes that were significantly altered by at least one of the compounds. In table 2 the number of regulated genes for each compound and some examples of relevant gene names and their corresponding biological function are listed. These include the induction of vitellogenin by all of the estrogens and methyltestosterone (MT), the inhibition of several transferrin transcripts by Bisphenol A (BPA), ethinylestradiol (EE2) and flutamide (FLUT), and the stimulation of cytochrome c oxidase genes by nonylphenol (NP) and Cortisol (CORT).

Example 5: Clustering analysis and discriminatory gene selection

In the present study we aimed to investigate if our developed cDNA microarray may be used to reveal chemical-specific signature patterns. To that end, we first subjected the complete gene expression dataset to a two-dimensional hierarchical clustering (Eisen et al., 1998). For every compound, the expression value corresponding to the most effective treatment condition was selected for each gene. Both transcripts and compounds were then clustered on the basis of similarity, and the resulting relationships between compounds, based on gene expression profiles, and vice versa, are highlighted in a dendrogram (figure 7). A visual inspection of the compound dendrogram indicates that our microarray has the potential to distinguish RNA samples derived from fish exposed to the different chemicals. There are both some striking expression similarities and dissimilarities between the compounds. The estrogenic compounds (17beta-estradiol (E2), 4-nonlyphenol (NP), Bisphenol A (BPA) and 17alpha-ethinylestradiol (EE2)) for example are well grouped together, and 17alpha-methyltestosterone (MT), an androgen which is presumed to be metabolically converted into estrogen (Hornung et al., 2004), is also found in the "estrogen sub-node". Yet, in spite of some marked expression similarities between these compounds, it is clear that each individual member of the estrogen class has a unique gene expression pattern, distinct from the other compounds. Next to the estrogens, the Cortisol modulators (hydrocortisone (CORT) and copper chloride (CuCI₂)) are also clustered together, although the correlation is not very strong. Further, dibutyl phthalate (DBP), flutamide (FLUT) and thyroid hormone (T3-T4) were found to be quite closely related, and also tamoxifen (TAM) and propylthiouracil (PTU) share some similarities. A close-up view of some of the major expression analogies between the different (groups of) compounds is given is figure 1 A, B, C, and D.

In a second phase, we investigated whether a small set of informative genes could be identified, whose expression pattern is sufficient to discriminate between compounds. Therefore, distances (i.e. the number of genes whose expression differs sufficiently) between the compounds were calculated, and subjected to a genetic algorithm, which enabled the selection of subsets of genes that separated the compounds in an optimal way. Figure 8 shows the distance matrix and transcriptional profile of the gene set with the highest discriminating power between the 14 compounds. The distance matrix, summarizing all pair wise distances between the 14 compounds, as calculated in material and methods, demonstrates that all chemicals can be distinguished from any other chemical by a combination of at least 3 genes. Smallest distances were found between T3-T4 and DBP, and TAM and PTU, respectively. When compounds were hierarchically clustered based on the expression levels of this 'discriminatory gene set' (figure 8B), the estrogens were clearly clustered together again. Furthermore, CORT and CuCI₂ were also grouped, and a node was formed by DBP, T3-T4 and FLUT. Finally, PTU and TAM were also clustered together. It is noted that all of the described compound associations based on the expression pattern of this small subset of genes, correspond to those observed after clustering the whole array data set (figure 7). This indicates that our discriminatory gene set comprises the most informative genes of the array, making it a valuable tool for describing key gene expression effects resulting from chemical exposure.

Using the same approach, we aimed to find a subset of genes, out of the set of gene fragments that are differentially expressed, that separates the compounds after water exposure in an optimal way. Distances between two compounds were calculated as the number of genes whose expression differed sufficiently between the compounds. Using a genetic algorithm, subsets of 8 to 12 genes ("classifiers") were selected such that the total number of pairwise small compound distances <3 was minimal. The resulting classifiers each correspond to a distance matrix/dendrogram summing up all the distances for the different compound combinations. Several informative classifiers could be selected, each leading to a different distance matrix/dendrogram. A number of these dendrograms were able to make a distinction between all of the different compounds, even within the same toxicant class. Some, but not all, (classes of) compounds clustered together as was expected according to their presumed mechanism of action. Figure 9 shows the dendrogram obtained with 5 different sets of classifier genes. Classifier 2 has the strongest discriminative power, as all the compounds can be distinguished by the expression pattern of at least 2 genes out of the set.

Our results show that using the developed microarray or even an array based on a subset of genes has the potential to discriminate between different classes of endocrine disruptors. It allows to discriminate on the mode of action of "unknown" chemicals: by testing the unknown compound under water+injection exposure, or by water contact or injection alone, and by evaluating the gene expression profile we can evaluate whether the compound has specific endocrine disrupting properties. Its further use as a classification/prediction tool can be further investigated, by evaluating additional compounds, also by testing effluents samples or extracts.

Example 6: expression changes of Vtg 5' and 3' end fragments induced by endocrine disrupting chemicals

In order to validate our microarray data, we examined the expression changes induced by 17 beta-estradiol (E2), 4-nonylphenol (NP), bisphenol A (BPA) and ethinylestradiol (EE2) in two presumed alternative splice variants of the vitellogenin gene in carp (Vtg 5' end and Vtg 3' end) by Real Time PCR (figure 10). RNA samples were collected as described in example 4. The sequence located at the 5' terminus represents a gene fragment which is present in both the alternative splice products, and the 3' end fragment represents the longest splice variant. In figure 10, the proportion of Vtg 5' expression relative to Vtg 3' imRNA in control and estrogen- treated fish is shown. In line with our previous work, we observed a difference in induction level between the two Vtg splice variants: for each of the estrogenic compounds, the Vtg 5' end was more strongly induced compared to the Vtg 3' end, which confirms our hypothesis the Vtg 5' and 3' end respond differently to female hormones. Furthermore, it was shown that there is a basal difference in expression level of the two splice variants - with a higher basal expression level for the shorter splice variant - which increased after estrogen exposure (figure 10). These findings suggest that the regulation of Vtg pre-mRNA splicing in carp is under estrogenic control.

REFERENCES

- Allen Y, Scott AP, Matthiessen P, Haworth S, Thain JE, Feist S, 1999. Survey of estrogenic activity in United Kingdom estuarine and coastal waters and its effects on gonadal development of the flounder Platichthys flesus. Environmental Toxicology and Chemistry 18: 1791 -1800.

Altschul SF, Madden TL, Schaffer AA, Zhang JH, Zhang Z, Miller W, Lipman DJ, 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Research 25: 3389-3402.

- Ankley, G. T., Johnson, R. D., Toth, G., Folman, L. C, Detenbeck, N. E., and Bradbury, S. P. (1997). Development of a research strategy for assessing the ecological risk of endocrine disruptors [Review]. Toxicolology 1 : 71-106.

- Belfroid AC, Van der Horst A, Vethaak AD, Schafer AJ, Rijs GB, Wegener J, Cofino WP, 1999. Analysis and occurrence of estrogenic hormones and their glucuronides in surface water and waste water in The Netherlands. Sci Total Environ. 12:101-108. - Bromage nR, Cumaranatunga PRC, 1988. Egg production in the rainbow trout. In:

Recent advances in aquaculture, vol 3 (Roberts RJ, Muir JF, eds). Croom Helm, London, pp. 63-138.

Butte AJ, Dzau VJ, Glueck SB, 2001. Further defining housekeeping, or "maintenance," genes Focus on "A compendium of gene expression in normal human tissues". Physiol Genomics. 21 :95-96.

Colborn, T., Clement, C, Eds., 1992. Chemically-Induced Alterations in Sexual and Functional Development: The Wildlife/Human Connection. Princeton Scientific Publishing Co., Inc. Princeton, NJ.

- Copeland PA, Sumpter JP, Walker TK, Croft M, 1986. Vitellogenin levels in male and female trout (Salmo gairdneri Richardson) at various stages of the reproductive cycle.

Comp Biochem Physiol 83B: 487-493.

Denslow ND, Bowman CJ, Ferguson RJ, Lee HS, Hemmer MJ, Folmar LC, 2001. Induction of gene expression in sheepshead minnows (Cyprinodon variegatus) treated with 17 beta-estradiol, diethylstilbestrol, or ethinylestradiol: The use of mRNA fingerprints as an indicator of gene regulation General and Comparative Endocrinology

121 : 250-260.

Diatchenko L, Lau YFC, Campbell AP, Chenchik A, Moqadam F, Huang B, Lukyanov K, Gurskaya N, Sverdlov ED, Siebert PD, 1996. Suppression subtractive hybridization: A mathod for generating differentially regulated or tissue-specific cDNA probes and libraries. Proc. Nat. Acad.Sci. USA 93:6025-6030. EDSTAC. Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) Final Report, August 1998. Federal Register Notice: 63, (248) 71541-71568, URL: http://www.epa.9ov/sdpoly/oscpendo/edspoveryiew/Tinalrpt.htm. Eisen MB, Spellman PT, Brown PO, Botstein D, 1998. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 8;95(25): 14863-8.

Fisher JS, 2004. Are all EDC effects mediated via steroid hormone receptors? Toxicology. 205:33-41.

Giesy JP, Pierens SL, Snyder EM, Miles-Richardson S, Kramer VJ, Snyder SA, Nichols KM, Villeneuve DA, 2000. Effects of 4-nonylphenol on fecundity and biomarkers of estrogenicity in fathead minnows (Pimephales promelas). Environmental Toxicology and Chemistry 19: 1368-1377.

- Harries JE, Sheahan DA, Jobling S, Matthiessen P, Neall P, Routledge EJ, Rycroft R, Sumpter JP, Tylor T, 1996. A survey of estrogenic activity in United Kingdom inland waters. Environmental Toxicology and Chemistry 15: 1993-2002. - Hornung Michael W., Jensen Kathleen M., Korte Joseph J., Michael D., Durhan

Elizabeth J., Denny Jeffrey S., Henry TaIa R. and Ankley Gerald T., 2004. Mechanistic basis for estrogenic effects in fathead minnow (Pimephales promelas) following exposure to the androgen 17α-methyltestosterone: conversion of 17α- methyltestosterone to 17α-methylestradiol. Aquat Toxicol. 66(1 ): 15-23 - Huber W, Von Heydebreck A, Sultmann H, Poustka A, Vingron M, 2002. Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics 18:S96-S104.

- Jobling S, Reynolds T, White R, Parker MG, Sumpter JP, 1995. A variety of environmentally persistent chemicals, including some phthalate plasticizers, are weakly estrogenic. Environ. Health Perspect. 103: 582-587.

- Jones PD, De Coen WM, Tremblay L, Giesy JP, 2000. Vitellogenin as a biomarker for environmental estrogens. Water Science and Technology 42: 1-14.

- Kavlock, R. J., Daston, G. P., DeRosa, C, Fenner-Crisp, P., Gray, L. E., Kaattari, S., Lucier, G., Luster, M., Mac, M. J., Maczka, C, et al. (1996). Research needs for the risk assessment of health and environmental effects of endocrine disruptors: A report of the

U.S. EPA-sponsored workshop. Environ. Health Perspect. 104(Suppl. 4), 715-740. Kroll TC, WoIfI S. 2002. Ranking: a closer look on globalisation methods for normalisation of gene expression arrays. Nucleic Acids Res. 30:e50. Livak KJ, Schmittgen TD, 2001. Analysis of relative gene expression data using real- time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25:402-408. Moens LN, Van der Ven K, Caturla M and De Coen WM, 2003. Toxicogenomic applications in environmental toxicology. In: In Vitro Methods in Aquatic Ecotoxicology (Carmel Mothersill and Brian Austin, eds). Berlin: Springer Verlag, 185-220.

- Petit F, Le Goff P, Cravedi JP, Kah O, Valotaire Y, Pakdel F, 1999. Trout oestrogen receptor sensitivity to xenobiotics as tested by different bioassays. Aquaculture 177:

353-365.

Pfaffl, M.W. 2001. A new mathematical model for quantification in real time RT-PCR. Nucleic Acids Res. 29: e45.

Richmond CS, Glasner JD, Mau R, Jin H, Blattner FR 1999. Genome-wide expression profiling in Escherichia coliK-12. NucleicAcidsRes.27: 3821-3835.

- Selvey S, Thompson EW, Matthaei K, Lea RA, Irving MG, Griffiths LR 2001. Beta-actin- -an unsuitable internal control for RT-PCR. MoI Cell Probes. 15: 307-31 1. Stoyanova R, Querec TD, Brown TR, Patriotis C. 2004. Normalization of single-channel DNA array data by principal component analysis. Bioinformatics details - Sumpter JP and Jobling S 1995. Vitellogeneis as a biomarker for estrogenic contamination of the aquatic environment. Environmental Health perspectives 103: 173-178.

Tanghe T, Devriese G, Verstraete W, 1999. Nonylphenol and estrogenic activity in aquatic environmental samples. Journal of Environmental Quality 28: 702-709. - Van Der Kraak G, Hewitt M, Lister A, McMaster ME, Munkittrick KR, 2001. Endocrine toxicants and reproductive success in fish. Human and Ecological Risk Assessment 7: 1017-1025.

- Weckx S, De Rijk P, Van Broeckhoven C, Del-Favero J, 2004. SSHSuite, an integrated software package for analysis of large scale suppression subtractive hybridisation data. Biotechniques. 36: 1043-1045.

Claims

1. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, comprising at least 8 sequences selected from the group consisting of SEQ ID N° 1 - SEQ ID NMO.

2. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 4, SEQ ID N° 5, SEQ ID INI" 1 1 , SEQ ID N° 12, SEQ ID N° 13, SEQ ID N° 17, SEQ ID N° 18 and SEQ ID N° 24.

3. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 2, SEQ ID N° 4, SEQ ID N° 5, SEQ ID N° 8, SEQ ID N° 1 1 , SEQ ID N° 13, SEQ ID N° 14, SEQ ID N° 16, SEQ ID N° 17,

SEQ ID N° 24, SEQ ID N° 27 and SEQ ID N° 29.

4. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 3, SEQ ID N° 5, SEQ ID N° 7, SEQ ID N° 8, SEQ ID N° 10, SEQ ID N° 14, SEQ ID N° 17, SEQ ID N° 21 , SEQ ID N° 25, SEQ ID N° 28, SEQ ID N° 29 and SEQ ID N° 31.

5. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 7, SEQ ID N° 8, SEQ ID N° 13, SEQ ID N° 15, SEQ ID N° 17, SEQ ID N° 18, SEQ ID N° 20, SEQ ID N° 21 , SEQ ID N° 25, SEQ ID N° 26, SEQ ID N° 29 and SEQ ID N° 30.

6. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 1 , SEQ ID N° 5, SEQ ID N° 6, SEQ ID N° 9, SEQ ID N° 10, SEQ ID N° 13, SEQ ID N° 19, SEQ ID N° 22, SEQ ID N° 23, SEQ ID N° 26, SEQ ID N° 29 and SEQ ID N° 30.

7. A nucleic acid array for detecting and/or classifying endocrine disruptor compounds, according to claim 1 , comprising at least SEQ ID N° 2, SEQ ID N° 1 1 , SEQ ID N° 12,

SEQ ID N° 15, SEQ ID N° 18, SEQ ID N° 21 , SEQ ID N° 32, SEQ ID N° 33, SEQ ID N° 34, SEQ ID N° 35, SEQ ID N° 36 and SEQ ID N° 37.

8. A method for detecting and/or classifying endocrine disruptor compounds, comprising: (1 ) contacting carp with the test material (2) extracting RNA from the carp that has been injected (3) hybridizing said RNA or its cDNA against a nucleic acid array according to any of the claims 1-7.

9. A method for detecting and/or classifying endocrine disruptor compounds, comprising: (1 ) contacting carp with the test material (2) extracting RNA from the carp that has been injected (3) detecting the expression of at least 8 genes selected from the group genes comprising SEQ ID N° 1 - SEQ ID N° 40.