Ecotoxicogenomics: Emerging Technologies for Emerging Contaminants1

Authors

  • Helen C. Poynton,

    1. Respectively, ORISE Postdoctoral Fellow, Associate Professor, Department of Nutritional Sciences and Toxicology, UC Berkeley, Berkeley, California 94720 [Poynton now at USEPA, 26 W. Martin Luther King Dr., Cincinnati, Ohio 45268].
    Search for more papers by this author
  • Chris D. Vulpe

    1. Respectively, ORISE Postdoctoral Fellow, Associate Professor, Department of Nutritional Sciences and Toxicology, UC Berkeley, Berkeley, California 94720 [Poynton now at USEPA, 26 W. Martin Luther King Dr., Cincinnati, Ohio 45268].
    Search for more papers by this author

  • 1

    Paper No. JAWRA-07-0185-P of the Journal of the American Water Resources Association (JAWRA). Discussions are open until August 1, 2009.

(E-Mail/Vulpe: vulpe@berkeley.edu)

Abstract

Abstract:  In recent years, new classes of aquatic pollutants have received attention from environmentalists, scientists, and regulators due to their introduction into the environment, unforeseen effects associated with the pollutants, or enhanced analytical techniques presently capable of detecting them. Many of these emerging contaminants are not well studied, making predictions regarding their toxicity to aquatic organisms or environmental fate difficult. Genomic technologies, including DNA microarrays, have been employed in many areas of biology to study disease states and the interaction of chemicals and nutrients with organisms. Here, we present the potential utility of DNA microarrays to address the challenges of emerging contaminants. DNA microarrays produce gene expression profiles creating an illustration of how a pollutant is acting within an exposed organism. Homology searches and online tools such as Gene Ontology can aid in the inferring a Mode of Toxicity, which may guide toxicity testing and risk characterization. Signature gene expression profiles offer the potential to uncover novel biomarkers of exposure and predict the presence of these contaminants in aquatic organisms. The No Observed Transcriptional Effect Level may play a role in determining if a predicted environmental concentration poses a risk to a sensitive species within an ecosystem. Additionally, DNA microarrays may add a complementary approach to Toxicity Identification Evaluations and help characterize causal agents in complex effluents.

Introduction

Emerging contaminants are a complex and pressing concern in environmental health, and new approaches, such as genomics, are needed to evaluate the risk of our aquatic ecosystems and water supplies. Review articles have articulated different definitions, but most agree that emerging contaminants are unregulated chemicals, which have become of environmental concern either through advances in analytical techniques to detect them, their association with a newly reported effect, or through their recent introduction in the environment (Kolpin et al., 2002; Muir and Howard, 2006; Petrovic and Barcelo, 2006; Richardson, 2007; Ternes, 2007). They may represent a distinct class of chemicals with a common structure such as brominated flame retardants, or a class that is defined by the effect it causes in organisms such as endocrine disrupting compounds. In general, the risk of these compounds to aquatic ecosystems is largely unknown while their environmental release is not regulated. Additionally, their number is likely to continue to grow unabated, given that only a small fraction of above 30,000 chemicals in commerce have been studied (Richardson, 2007).

Because of the scope of the problem of emerging contaminants, we suggest that holistic genome-based approaches may aid in understanding how these chemicals interact with aquatic organisms. Ecotoxicology has in recent years embraced the genomic technologies to create the rapidly growing field of ecotoxicogenomics (Snape et al., 2004). Genomic tools, including microarrays, target the molecular responses the organism experiences in reaction to the pollutant, and provide an illustrative picture suggestive of the toxic effects experienced by the organisms and the compensatory mechanisms the organism has mobilized in its defense. Microarray-based methods may assist in characterizing the adverse effects of emerging chemicals and offer a novel approach to detect them in the environment. The purpose of this review article is not to compare genomics and traditional ecotoxicology, but to describe how microarray technologies work and propose areas in emerging contaminant research where they may contribute, as complementary tools, to understanding the risks these contaminants pose.

Examples and Challenges

We include a few examples of emerging contaminants to illustrate the unique challenges presented by this broad category of pollutants (also see Box 1). For a more detailed discussion of the chemical properties and other classes of emerging contaminants, we refer the reader to the recent review by Field et al. (2006) and Richardson (2007).

Table BOX 1..   Challenges When Confronting Emerging Contaminants.
Unknown or Untraditional Toxicity: For many emerging contaminants, their toxicity to aquatic organisms is largely unknown. Even pharmaceuticals, which undergo extensive testing in mammalian models, may exhibit different toxicity on aquatic species (Crane et al., 2006). In addition, many pharmaceuticals and EDCs are not responsive to traditional toxicity assays that measure lethality or reproduction over a single generation (Sumpter and Johnson, 2005) and are requiring regulatory agencies to rethink testing requirements (Gray, 1998). This could also be true for other emerging chemicals including PBDEs and nanomaterials whose mechanism of action is not known.
Fate and Transport: A few classes of emerging pollutants are highly persistent in the environment and monitoring efforts have shown their global distribution. For other pollutants, there is limited information regarding their fate and transport. Nanoparticles in particular pose a unique challenge. Because of their unusual size, and the diversity of their chemical properties, very little is known about where nanomaterials are likely to exist in the environment (Moore, 2006).
Complex Effluents and Mixtures: Often these contaminants are found in complex effluents and are therefore, found in combination with other chemicals. For example, pharmaceuticals are primarily found in wastewater effluent, which is comprised of a mixture of many different PPCPs and other contaminants (Dorne et al., 2007).
Regulation: For many of these chemicals, the current or anticipated regulatory situation remains uncertain. For a discussion of regulatory issues related to emerging chemicals, see Richardson (2007).

Pharmaceuticals and Personal Care Products

Pharmaceuticals and personal care products (PPCPs) represent a category of emerging pollutants that are released in the environment through personal activities. Small but cumulative usage of these products by a large number of individuals results in widespread but significant pollution, which is difficult to control. The primary route of pharmaceuticals into the environment is through human excretion, disposition of unused or expired products, or through agricultural usage in livestock (Fent et al., 2006). Because these are often the active ingredients in medicines, they are engineered to be bioavailable and potent and may result in adverse environmental effects at low concentrations. While acute toxicity data are available for a number of pharmaceuticals (Fent et al., 2006), sublethal effects to organisms due to chronic exposure to low, environmentally relevant concentrations are not known (Daughton and Ternes, 1999). However, pharmaceutical and PPCP residues have been detected in fish tissues downstream of wastewater treatment facilities leading to bioaccumulation in muscles and critical organs (Schwaiger et al., 2004; Brooks et al., 2005). The synthetic steroid ethynylestradiol (EE2) is an example of a well-studied pharmaceutical, shown to cause sublethal effects in fathead minnow leading to population decline at very low concentrations (Kidd et al., 2007).

Endocrine Disrupting Chemicals

Endocrine disrupting chemicals (EDCs) are pollutants, which interfere with the normal functioning of the endocrine system resulting in adverse effects of reproduction, development, and immune function (Vos et al., 2000). EDCs consist of many classes of chemicals including natural and synthetic hormones and other pharmaceuticals, pesticides, plasticizers, and organometallic compounds. Effects from EDCs are not a new phenomenon, and several EDCs such as DDT and tributyl tin are currently regulated.

Polybrominated Flame Retardants

Polybrominated diphenyl ethers or PBDEs are a group of chemicals that are used as flame retardants in a variety of polymer resins and plastics. They are found in many products such as furniture, televisions, stereos, computers, carpets, and curtains, which are used in homes and businesses. Because of their persistence in the environment, they are ubiquitous, global pollutants which readily bioaccumulate in living organisms (de Wit, 2002; Ueno et al., 2004; Ramu et al., 2007). PBDEs can cause neurotoxicity in mammals (Eriksson et al., 2006) and can act as thyroid hormone agonists or antagonists in frogs (Veldhoen et al., 2006).

Perfluorinated Compounds

Perfluorinated compounds (PFCs), including the two most common compounds, perfluorooctane sulfonate and prefluorooctanoic acid, have been used in industries for over 50 years because of their unique chemical properties such as hydrophobicity, lipophilicity, and moderate solubility. They are found in food packaging, as coatings on cookware, in paints, as surfactants, and also have many other applications. Like PBDEs, they are very stable compounds, which make them extremely persistent in the environment (Prevedouros et al., 2006; Richardson, 2007). They have been detected in the arctic atmosphere (Shoeib et al., 2006), throughout the global oceans (Yamashita et al., 2005), in the wildlife of Antarctica (Tao et al., 2006), and in animals and humans worldwide (Houde et al., 2006). A recent review by Lau et al. (2007) summarized the known effects associated with PFCs, for example, effects to the liver, thyroid hormones, and development.

Nanomaterials

Nanotechnology is concerned with supra-molecular compounds in the “nano” range (0.1-100 nm in diameter) and their potential commercial and scientific applications. Besides their common size, nanoparticles constitute a very diverse range of compounds. They include those made primarily from carbon, such as the buckminsterfullerene or C60 and carbon nanotubes, to metal based spherical particles. Nanoparticles may be coated to enhance some physical characteristic such as solubility and other functional groups may be attached to the original particle to alter its chemical properties. Overall, their diversity poses unique challenges to environmentalists who must assess the risks posed by each of these particles and understand how chemical alterations affect this risk (Moore, 2006). Currently, there are over 500 commercial products and the industry is expected to grow exponentially over the next decade (http://www.nanotechproject.org).

Organisms and Nonliving Pathogens

Emerging contaminants also include organisms and nonliving pathogens such as some infectious agents and invasive species. Richardson (2007) provided a thorough description of algal toxins and microbes of concern. Recent evidence suggests that prions too may be transferred through environmental sources (Miller et al., 2004) including the soil (Johnson et al., 2006). Increased global trade and travel have also facilitated the occurrence of invasive species introductions with ecological and economic consequences (Strayer et al., 2006). Genomic techniques have been developed and employed for species identification and have applications for emerging pathogens and invasive species. However, as the methodologies and considerations of these technologies differ somewhat from gene expression microarrays, this subject is beyond the scope of this review. Several papers have addressed the use of genomics for detecting adenovirus (Jiang, 2006), specific bacteria strains using 16S rRNA (Pontes et al., 2007), pathogens using environmental arrays (Heinemann et al., 2006), and invasive species (Darling and Blum, 2007).

The Ecotoxicogenomics Approach

Genomic analysis, such as expression profiling, can assess an organism’s reaction to an environmental stressor. Figure 1 illustrates how a pollutant can lead to specific gene expression changes in an exposed organism. First, the organism is exposed to a chemical pollutant which enters and distributes throughout its body. The pollutant interacts with cells and cellular components in a manner dependent on its chemical properties resulting in specific cellular damage. In response, the organism reacts to the pollutant at multiple levels which includes altering the expression of genes, protein levels, or metabolite concentrations. These changes could, for example, aid in protecting the organism from the particular stressor or mitigate adverse effects of the stressor. The particular set of genes (or proteins or metabolites) which are altered will be dependent on and specific for the pollutant’s mechanism of action. The particular pattern of response therefore can represent a fingerprint for a specific mode of action (MOA) and pollutant.

Figure 1.

 DNA Microarrays as a Tool to Study Organism Responses to Pollution.

The DNA microarray is one of a host of genomic tools available to the ecotoxicologists. Gene expression microarrays generally consist of an ordered array of DNA spots in close proximity (e.g., 100 microns) to each other on a substrate (such as a glass slide). Each spot usually contains a different DNA molecule which represents a probe for a specific gene. Microarrays allow an investigator to interrogate the expression of all the genes represented on the array with a single RNA sample. There are multiple variants of this general approach which vary in the origin of the DNA probe and the spotting procedure. cDNA microarrays contain DNA which originates from PCR amplification of individual cloned cDNAs. As a result, the length of the probe can vary considerably. Robotic printers are used to print the DNA spots in high density on the slides. A more recent type of DNA microarray consists of 50-70mer oligonucleotide probes for specific transcripts. In contrast to the cDNA arrays, sequence information for each transcript is needed. These oligonucleotides can be synthesized directly on the glass slide, or prefabricated oligos may be spotted on a coated glass slide (Marshall and Hodgson, 1998). Affymetrix arrays represent another approach in which multiple short oligos are used as probes for a particular gene. All of the platforms have been used in ecotoxicology studies and standardization of platform remains an issue of concern. For more information including an experimental comparison of different microarray platforms see Yauk et al. (2004) and the Microarray Quality Control Program (http://www.fda.gov/nctr/science/centers/toxicoinformatics/maqc/index.htm).

There are multiple experimental approaches for microarray studies (MAQC Consortium, 2006). We illustrate a common approach in Figure 2 in which two pools of cDNA representing exposed and control samples are competitively hybridized to a microarray. The comparative hybridization method was developed by Schena et al. (1995), DeRisi et al. (1996), and Shalon et al. (1996). Subsequently, multiple alternative methods have been developed such as using cRNA and linear amplification methods for RNA samples to increase the sensitivity of the signal and allow microarray investigations of very small tissue samples (Pabon et al., 2001).

Figure 2.

 Illustration of a Competitive Hybridization Microarray Experiment. In this approach, the test organism(s) is exposed to the contaminant or an appropriate control condition. RNA is extracted from both exposed and control, reverse transcribed to cDNA, and each is labeled with a different colored fluorescent dye (e.g., red and green). The two labeled samples are pooled and simultaneously hybridized to the microarray. Complementary DNA strands of each label color to the probe DNA spotted onto the glass slide will bind in proportion to their abundance in the pool. Following hybridization, the microarrays scanned are used to determine the relative intensity of the two dyes bound to each DNA spot and by extension the relative abundance of the transcript corresponding to the DNA in the control and experimental samples. This figure was adapted from Nuwaysir et al. (1999).

The analysis of data derived from gene expression microarrays can be complex and remains an area of active research. A key first step is to identify differentially expressed genes (e.g., increased transcript level in one condition variation). There has been some consensus about analysis approaches (Allison et al., 2006; Quackenbush, 2006), but lack of standardization in approaches has introduced difficulties when comparing results between laboratories. As originally suggested by Nuwaysir et al., genes identified as differentially expressed in response to a particular exposure can represent a gene expression signature or fingerprint for that exposure condition (Nuwaysir et al., 1999). Similarly, gene expression signatures may help in classifying unknown toxicants and will aid in the molecular and biochemical understanding of the toxic mechanism (Nuwaysir et al., 1999). As described later, there has been considerable validation of the “expression signature” approach in multiple organisms including ones of relevance to ecotoxicology and water quality.

A major obstacle to the application of microarrays to emerging contaminants is the lack of genomic or cDNA sequence data for ecotoxicological species (Snape et al., 2004). Genomic or cDNA clone sequence data are used to generate nonredundant microarrays in which each spot represents a different gene. In ecotoxicology, many investigators have used alternative strategies for constructing microarrays with limited sequence information (Snell et al., 2003). One method is to create cDNA libraries enhanced for transcripts that respond to stress or represent a particular developmental stage of interest (Bultelle et al., 2002; Snell et al., 2003). These libraries decrease the amount of sequencing needed for microarray construction by increasing the relevance of chosen transcripts. Another method explored in our lab is the use of anonymous microarrays. Microarrays are printed with unsequenced cDNAs from a normalized library. Only the transcripts responsive to a particular treatment are sequenced, thus drastically reducing the cDNAs sequenced without reducing the number of transcripts investigated for differential expression (Wintz et al., 2006; Poynton et al., 2007).

Recent advances in high throughput parallel sequencing and the availability of commercial cost-effective printing capabilities will make comprehensive arrays accessible for any organism of interest. Until recently, oligonucleotide arrays were only commercially available for a small subset of species. Although a few investigators constructed their own oligo sets and printed them in-house for their species of interest (Iguchi et al., 2006; Kim et al., 2006; Watanabe et al., 2007), the upfront costs of such a project are large and not feasible for most laboratories. The development of high throughput pyrosequencing, such as that available from 454 life sciences (Margulies et al., 2005), enables a researcher to simultaneously obtain sequence information from a cDNA library sufficient to generate oligo probes representative of the library (Gowda et al., 2006; Weber et al., 2007). Commercial suppliers such as Agilent and Nimblegen recently began offering custom oligonucleotide microarrays making it possible for any laboratory to use the pyrosequencing data to construct an oligonucleotide array for their species of interest at a reasonable cost. The feasibility of this approach to develop and utilize oligo arrays for the large mouth bass (Micropterus salmoides) was recently demonstrated by Dr. Denslow’s group at the 2007 SETAC meeting (Barber et al., 2007).

Applications of Ecotoxicogenomics to Study Emerging Contaminants

Ecotoxicogenomics can assist in addressing the challenges of emerging contaminants, as discussed in Box 1, including unknown toxicity, persistence, environmental fate, transport, and presence in complex mixtures. In the following paragraphs, we describe how ecotoxicogenomics has demonstrated the ability to inform MOA, and predict exposure to chemicals in the environment. Table 1 also provides a list of recent studies that have applied genomic techniques to better understand the exposure to emerging contaminants. This section ends with two potential applications for genomics in monitoring of emerging contaminants, the No Observed Transcriptional Effect Level (NOTEL) and genomic-based Toxicity Identification Evaluation (TIE). The development of these applications is still in their infancy and further studies are needed to evaluate their potential for water quality monitoring; however, preliminary studies suggest that these techniques may become valuable tools in the future.

Table 1.   Studies Applying Genomics Technologies to Study Emerging Contaminants.
ContaminantOrganismGenomic ApproachReference
Pesticides
 Chlorpyrifos, DiazinonRattus norvegicus252 Gene arraySlotkin and Seidler, 2007
 DiazinonHomo sapiens600 Gene arrayMankame et al., 2006a,b
 DiazinonOryzias latipesDifferential displayYoo et al., 2007
 FenarimolDaphnia magnaSSH PCR, cDNA arraySoetaert et al., 2007
 MianserinDanio rerioBrain specificvan der Ven et al., 2006b
Emerging Contaminants
 2,4-DNTPimephelas promelascDNA arrayWintz et al., 2006
 Bis (Tri-N-butyltin) oxide (TBTO)Rattus norvegicusOligo arrayBaken et al., 2007
 BromobenzeneRattus norvegicusAffymetrix arraysTanaka et al., 2007a
 BromobenzeneRattus norvegicusMetabolomics/gene expression Heijne et al., 2005
 Nanoparticles (C60)Danio rerioAffymetrix arraysHenry et al., 2007
 PerchlorateXenopus laeviscDNA array, Q-PCRHelbing et al., 2007
 Perfluorooctanoic acid (PFOA)Rattus norvegicusAffymetrix arraysGuruge et al., 2006
 PFOAGobiocypris rarus cDNA array, Q-PCRWei et al., 2008
 PFOAMus musculusAffymetrix arraysRosen et al., 2007
 PFOA, perfluorooctane sulfonate acid (PFOS)Gallus gallusGenome arraysYeung et al., 2007
 RDXPopulus deltoides X nigra DN34RT-PCRTanaka et al., 2007b
 RDXRattus norvegicusOligo arraysPerkins et al., 2006
 RDXArabidopsis thalianaSAGEEkman et al., 2005
 TributyltinSalmo salarReal time PCRMortensen and Arukwe, 2007
 TributyltinTetrahymena thermophilaSSH, Q-PCRFeng et al., 2007
 TrimethylbenzeneRattus norvegicusMicroarrayMcDougal and Garrett, 2007
 VanadiumRattus norvegicusAffymetrix arraysWillsky et al., 2006
Pharmaceuticals
 ChlorpromazineDanio rerioBrain specific arrayvan der Ven et al., 2005
 Mixture of 13 pharmaceuticals Danio rerioOligo arrayPomati et al., 2007
 PropiconazoleDaphnia magnacDNA arraySoetaert et al., 2006
Complex Mixtures
 EffluentsCyprinus carpicDNA microarrayMoens et al., 2007a
 Herbicide mixturePlatichthys flesusSSHMarchand et al., 2006
 Multiple contaminantsPhalacrocorax carbocDNA microarrayNakayama et al., 2006
 Paper mill effluentMicropterus salmoidesDifferential displayDenslow et al., 2004
Endocrine Disruptors
 17α-ethinylestradiol (EE2)Danio rerioOligo arraySantos et al., 2007
 EE2Danio rerioOligo array, RT-PCRMartyniuk et al., 2007
 EE2Carassius auratuscDNA microarrayMartyniuk et al., 2006
 EE2Oncorhynchus mykisscDNA microarrayHook et al., 2006
 17β estradiol, 4-nonylphenol (4NP), 1,1-dichloro-2,2-bis (P-chlorophenyl) ethylene (P,P′-DDE)Micropterus salmoidesMacroarrayLarkin et al., 2003
 17β-estradiolPimephales promelasOligo arrayLarkin et al., 2007
 4-nonylphenolOryzias latipesOligo arrayKim et al., 2006
 E2, 4NP, bisphenol A, EE2Cyprinus carpiocDNA microarrayMoens et al., 2007b
 FadrozolePimephales promelasOligo arrayVilleneuve et al., 2007
 Flutamide, EE2Pimephales promelasQ-PCRFilby et al., 2007
 Progesterone, estrogen, testosteroneCaenorhabditis eleganscDNA microarrayCustodia et al., 2001

Genomic Approaches to Understanding MODE OF ACTION

Expression profiles, and other genome wide approaches, have helped generate testable hypotheses of the MOA (Hamadeh et al., 2002; Waring et al., 2002) of toxicants and enabled classification of chemicals based on their MOA. For emerging chemicals, MOA data will be important to determine what analysis may be most relevant for a pollutant and will help prioritize testing of chemicals. Additionally, mechanistic data are necessary to make informed prospective risk assessments of emerging chemicals, and microarrays have the potential to play an important role in providing this information (Iguchi et al., 2006).

There are several methods to infer MOA information from microarray studies. Recent studies investigating the effects of the neuropharmaceuticals on zebra fish brain correlated the effects seen in the fish species with effects known to occur in mammals. Thus, through species extrapolation, they were able to make predictions about the long-term effects of these drugs in fish (van der Ven et al., 2005, 2006a,b). Other techniques for investigating MOA from microarray studies were reviewed recently (Moggs, 2005). Moggs suggested using hierarchical clustering coupled with Gene Ontology (GO) to understand the categories of genes affected and the functions that may be impaired by exposure. Also integrating the gene products into known biological pathways can reveal the metabolic, signaling or other responses to exposure. Finally, phenotypic anchoring in which gene expression is linked to physiological/phenotypic effects is vital to provide support for causal inference. A goal is identification of clusters of genes that predict an outcome and can be used as biomarkers. These approaches will be described in the paragraphs below.

GO is a systematic method to describe the molecular function, biological process, and cellular component of each gene based on a structured vocabulary (Harris et al. 2004). This public resource is available through the GO Consortium (http://www.geneontology.org). The investigator can use this classification system to determine which cellular components or biological processes are affected by a stressor. The assignment of GO terms to a list of differentially expressed genes provides insight into the higher order effects and the biological processes that could be involved in the MOA. Multiple recent studies of estrogenic compounds have used GO approaches to predict the physiological effects caused by estrogens and pharmaceuticals on different organs in zebrafish, Danio rerio (van der Ven et al., 2006b; Martyniuk et al., 2007; Pomati et al., 2007; Santos et al., 2007). In a recent microarray study of 17-β estradiol in fathead minnow, Pimephales promela, the authors used GO terms to identify candidate targets of toxicity. These include blood coagulation, metabolism, protein biosynthesis, electron transport, and regulation of cell growth (Larkin et al., 2007). However, for most ecotoxicology species, assignment of GO terms depends on a GO annotated homolog in well-studied organisms which adds additional uncertainty.

Another method to investigate the biological outcomes of a chemical exposure is through Pathway Mapping using pathway databases such as Kyoto Encyclopedia of Genes and Genomes (http://www.genome.jp/kegg/) (Ogata et al., 1999) and Gene Map Annotator and Pathway Profiler (http://www.genmapp.org) (Dahlquist et al., 2002). These tools allow the investigator to map differentially expressed genes identified in an exposure study to biological pathways. An overview can emerge of the potential targets and allow prediction of effects on the function of the pathway. Villeneuve et al. used pathway analysis to understand the effects of estrogenic compounds. They constructed a graphical model of the teleost brain-pituitary-gonadal axis to drive a hypothesis-based ecotoxicogenomic study of the effect of the androgen fadrozole on the fathead minnow. Their model, derived from a thorough literature search on the effects of endocrine disruption on the brain, pituitary, liver, and gonads, provides an essential tool for linking the molecular response to an adverse effect that may drive prospective risk assessment (Villeneuve et al., 2007).

Finally, as suggested by many investigators, phenotypic anchoring of molecular events to adverse physiological outcomes is essential for reliable MOA predictions from microarray studies (Paules, 2003; Moggs, 2005; Ankley et al., 2006). Phenotypic anchoring refers to the ability to demonstrate that a molecular event causes or is associated with a toxicological outcome or disease state. Without this crucial line of evidence, a change in the expression level of a gene has little toxicological significance. To show a phenotypic linkage, investigators have incorporated traditional toxicological endpoints into their microarray studies. A recent study of EE2 effects at an environmentally relevant concentration in zebrafish revealed that EE2 adversely affects gamete production in both males and females. The dysregulation of several gene families provided a potential mechanism for the disruption of gamete production in both sexes (Santos et al., 2007). Similarly, Iguchi et al. used genomic approaches to study endocrine disruption in a number of diverse, ecologically important species (Iguchi et al., 2006; Watanabe et al., 2007). In their studies, they have followed up their gene expression studies with detailed molecular characterization of the receptors involved in the endocrine disruption process to make progress toward understanding the molecular events involved in producing a phenotypic response.

Genomic Biomarkers of Exposure

Because the fate and transport of many emerging contaminants are unknown and several of these compounds have been shown to bioaccumulate and biomagnify, simple models are not available for predicting environmental exposure. Biomarkers of exposure are important tools for determining bioavailability of environmental contaminants and may play a role in detecting exposure to emerging contaminants. However, biomarkers are not presently available for many emerging contaminants. DNA microarrays have shown the ability to uncover novel biomarkers of exposure and may allow identification of new biomarkers to emerging contaminants. Additionally, gene expression signatures could aid in identifying the causal agents responsible for an observed toxicity (Nuwaysir et al., 1999; Miracle and Ankley, 2005).

Different strategies have been used to identify novel biomarkers using microarray studies. A strategy employed by Kim et al. was to compare the response to different estrogenic compounds and identify genes that were reproducible across chemical treatments (Kim et al., 2006). Another method is to select biomarkers related to the MOA. Biomarkers rooted in the MOA of contaminants are more informative, more reliable in the field, and likely to predict both exposure and effects (Mayer et al., 1992; Handy et al., 2003). This approach is exemplified in two zebrafish studies of estrogenic compounds, where the investigators selected genes predicted to be important players in endocrine disruption as potential biomarkers of exposure (Martyniuk et al., 2007; Pomati et al., 2007). Using a different approach to select candidate biomarkers, Venier et al. compared expression profiles from laboratory and field-exposed mussels, and uncovered over 40 novel biomarkers whose expression levels were regulated similarly in the laboratory and field exposures (Venier et al., 2006).

Bioinformatics can aid in the identification of the most robust set of biomarkers. Natsoulis et al. used an iterative trimming method, where they started with a list of 10,000 genes (or variables) and trimmed the gene set until they were left with a set of genes that were able to predict drug class (Natsoulis et al., 2005). In another study, feature selection tools for candidate gene identification was used alongside an algorithm for making class predictions to define a set of gene expression biomarkers for exposure to different endocrine disrupting compounds in zebrafish (Wang et al., 2008).

In addition to identifying biomarkers, many have suggested that the gene expression profile may be used to predict exposure to pollutants in the environment (Ankley et al., 2006). Recent studies in diverse ecotoxicological organisms have been conducted to prove that different chemicals produce distinct gene expression profiles including a few studies investigating emerging contaminants. In a gene expression study of 22 genes, Filby et al. investigated the effects of an estrogen (EE2) and an anti-androgen (Flutamide) on fathead minnow. Both chemicals act to suppress male secondary sex characteristics, but through different mechanisms. However, exposure to these two chemicals may be confused if only phenotypic characteristics are considered. Gene expression studies revealed clear differences in affected genes. The authors suggested that microarray screening tools could be used to distinguish exposure to these two chemicals in the environment (Filby et al., 2007). Moens et al. also showed the ability to distinguish among four compounds (17-beta-estradiol, 4NP, Bisphenol A, EE2) with estrogenic activity using microarray in the common carp (Cyprinus carpio) (Moens et al., 2006).

Field studies have demonstrated the capability to distinguish between reference sites and contaminated sites using gene expression profiling (Williams et al., 2003; Denslow et al., 2004; Maples and Bain, 2004; Roling et al., 2004; Meyer et al., 2005). Nakayama et al. correlated gene expression changes in wild cormorants (Phalacrocorax carbo) to measured tissue concentrations of several persistent organic pollutants (Nakayama et al., 2006). In our laboratory, we compared expression profiles in Daphnia magna exposed to blinded field samples from an abandoned copper mine to expression profiles of D. magna exposed in the laboratory to different metals. We were able to predict the primary contaminant in the field samples based on comparison of expression profiles (Poynton et al., 2008b). Although these studies are preliminary, these examples illustrate that gene expression profiling is able to distinguish between clean and contaminated sites. In addition, the causal agent in a field sample can be identified by comparison of expression profiles between laboratory and field exposures.

In addition to monitoring exposure, it has been proposed that DNA microarrays may play a role in monitoring for effects and act as an early warning that more severe effects will arise (Ankley et al., 2006). Although, phenotypic anchoring studies have suggested that genomic responses are able to predict chronic outcomes (Heinloth et al., 2004; Moggs et al., 2004), this has not yet been illustrated in the field. More studies are needed to further define the potential applications and limitations of genomics in ecotoxicology.

No Observed Transcriptional Effect Level

The NOTEL is defined as the dose of chemical which results in no significant changes in gene expression (Lobenhofer et al., 2004). Lobenhofer et al. first suggested this concept in a study of MCF-7 cells, an estrogen responsive cancer cell line, exposed to four different concentrations of estrogens including two low doses, a physiological dose and a cytotoxic dose. The authors noted that the two lowest doses did not significantly change expression in any of 2000 genes assayed, and suggested NOTEL as a threshold concentration for transcriptional responses to a stressor (Lobenhofer et al., 2004). Any significant cellular perturbation should cause some change in gene expression; therefore, the NOTEL represents a true No Observed Effect Concentration. Some have suggested that the NOTEL could be used to monitor contamination and even used to set regulatory standards (Ankley et al., 2006). A field sample which did not elicit changes in gene transcription would be unlikely to cause toxicity to the exposed organism. However, the converse is not necessarily true, gene expression changes, while indicative of exposure, may not always indicate toxicity but could represent compensatory responses which do not result in decreased fitness.

A few studies have illustrated the potential role of the NOTEL in environmental monitoring. A recent study conducted in our laboratory showed that at low metal concentrations, there were very few genes differentially expressed in D. magna. In addition, the gene expression profiles from the low metal concentrations did not overlap with the expression profiles of higher metal concentrations (Poynton et al., 2008a). We next compared the gene expression patterns resulting from exposure to field samples collected at two abandoned Cu mines. The gene expression patterns upstream from the Cu mines, with very low metal concentrations, resulted in very few differentially expressed genes (Poynton et al., 2008b). In another example, a Mummichog, Fundulus heteroclitus, microarray was used to monitor the effectiveness of remediation at a Superfund site. Roling et al. found that gene expression was correlated with tissue concentrations of chromium and predicted the tissue concentrations better than sediment chromium concentrations (after dredging). They also showed that when tissue concentrations were low, few genes were differentially expressed, providing additional evidence for a NOTEL (Roling et al., 2007).

Genomic Toxicity Identification Evaluation Approach to Mixtures

In field situations, organisms are exposed to not just one compound but a mélange of contaminants, which can interact within the environment and individual organisms. As mentioned earlier, many emerging contaminants are found in the environment in complex effluents containing a mixture of contaminants. The TIE process is often utilized to separate out the toxicity associated with individual chemicals in complex mixtures and determine the causal agents. Current methods for toxicity identification use biophysical separation methods and treatments which remove one or more toxicant classes, coupled with effluent toxicity testing following each manipulation. The time and cost associated with conducting a TIE in this manner can be substantial (Pillard and Hockett, 2001), and in general, these approaches have been successful in differentiating broad classes of toxicants in some effluents, but not individual pollutants. In addition, separation approaches do not consider synergistic or antagonist effects that may occur in mixtures. While this method has been a mainstay of toxicant identification for years, it is clear that a TIE approach that is more sensitive (able to identify toxicants not identified by current approaches), specific (able to identify individual toxicants vs. broad classes), timely (able to identify toxicants in days rather than weeks), and cost-effective (less than the current costs) would greatly facilitate the TIE process.

Genomics may offer alternative and complementary tools for TIE. Gene expression profiles may be able to determine the causal agent in a complex effluent. However, the question remains as to how gene expression signatures are influenced by the presence of other chemicals. It is known that chemicals may interact in mixtures, causing unexpected outcomes to survival and reproduction (Walker et al., 2006). It is likely that combinations of chemicals will have different effects than individual chemicals on the gene expression profiles of that organism. Chemicals that cause an additive response in acute or chronic bioassays may also have an additive expression profile. Chemicals that show synergistic or antagonist effects in standard bioassays may have distinctive expression profiles, not resembling the expression profiles of the single chemicals. For example, in a study of the effects of a mixture of chromium and benzo[a]pyrene (B[a]P) in a hepatoma cell line, Wei et al. (2004) found that the gene profile of the mixture was distinct from the profiles of the individual contaminants. However, these two known carcinogens target similar transcriptional pathways, in opposing directions causing an antagonistic response. Therefore, the suppression of many of the (B[a]P)-induced genes in the mixture was predicted. Krasnov et al. (2007) investigated three model toxicants, Cd, chlorotetrachloride (CCl4), and pyrene, and showed that the expression profiles were additive at low level exposures, and that the Cd and pyrene expression profiles could be dissected from the mixture profile. A final study involving nanoparticle research has shown the ability to distinguish between effects due to the nanoparticle and its co-solvent. Because of their hydrophobicity, the co-solvent tetrahydrofuran (THF) is often used to solubilize fullerenes (C60). The THF is removed which leaves the C60 solubilized in water. To examine the influence of the THF on C60 toxicity, Henry et al. prepared two formulations of C60, in water or solubilized with THF in water. Gene expression profiles were similar in the THF-C60 and the THF alone exposures suggesting that toxic effects seen and attributed to C60 may be a result of THF degradation products remaining in the water (Henry et al., 2007). This study demonstrates a method using DNA microarrays to decipher between the effects caused by a toxicant and its co-solvent, and shows promise for the use of genomics to determine causal agents. Despite these preliminary results, it is apparent that more studies are needed to understand the effects of mixtures on gene expression before DNA microarrays can be fully integrated into a TIE approach.

Conclusion

New chemicals and drugs are continuously developed and released in the environment. New approaches are needed for environmental risk assessment to catch up with the backlog of contaminants and keep pace with the increasing surge of new potential risks. Genomics provide the tools to remain competitive in the race. Genomic approaches including DNA microarrays will continue to help us understand the effects of conventional pollutants. Proof-of-principle studies have shown that gene expression profiling is able to suggest potential modes of action and predict exposure to pollutants in the environment. Genomics similarly can provide a tractable solution to the challenges presented by emerging contaminants including unknown toxicity, and uncertain environmental fate.

Acknowledgments

Special thanks to Dr. Henri Wintz for his thoughtful and insightful discussions, that contributed to this article. We would also like to thank the agencies that supported our research including the USEPA STAR fellowship program (FP-91644201-0), National Science Foundation (BES-0504603), and U.S. Army Engineer Research and Development Center (BAA053799).

Ancillary