In 1998, a group of scientists proposed approaches for developing alternative tests for ecotoxicity. These included the use of lower organisms as surrogates for vertebrates; nondestructive techniques in vertebrate testing such as the use of biomarkers, excreta, or eggs; (Quantitative) Structure Activity Relationships ((Q)SARs); and in vitro test systems (Walker et al., 1998). This section will describe progress in these and other approaches for replacing animals with non-animal methods for aquatic toxicity testing.
Biomarkers are an approach used in ecotoxicity testing to reduce and refine animal experiments. Walker (1998) defined a biomarker as “a biologic response to an environmental chemical at the individual level or below which demonstrates a departure from normal status.” At the present time, biomarker assays are not available for many classes of chemicals. The identification of relevant in vivo biomarkers of toxicity is critical to their use in cell-based assays. On the other hand, cell assays may be used to identify biomarkers indicative of a toxic response later found to be a useful indicator of an organism’s environmental exposure. One example of biomarker use in ecotoxicology is the use of vitellogenin expression in male fish as a marker for endocrine-related reproductive effects (e.g. Cheek et al., 2001). Walker (2006) proposed the development of biomarker assays for neurotoxicity and endocrine disruption for testing the ecotoxicity effects of pesticides, based on the fact that these two mechanisms of toxicity are responsible for population declines. Despite the promise of biomarkers, there are many complexities to their identification and use. For example, dioxin was found to induce gene expression changes in the brain, liver, and testis of medaka fish, but different genes were up or down regulated in the three fish tissues (Volz et al., 2005).
Complementary to the biomarker approach is to evaluate an array of gene (genomics), protein (proteomics), or small molecule (metabolomics) changes in cells, tissues, or organisms. Ankley, et al. (2006), discussed the utility of these types of molecular methods, especially toxicogenomic approaches for regulatory ecotoxicology. Exposure of cells or organisms to toxic substances (or any stressor) results in changes in their normal gene expression pattern. Specific patterns of gene expression can reflect different mechanisms of action of toxicity. The use of toxicogenomics in regulatory toxicology is complicated by many issues, including the fact that other stressors can contribute to changes in gene expression patterns; the vast amount of data generated by the experiments; and the lack of experience and know how in translating the genomic data into regulatory decisions. The major benefits identified for investing in toxicogenomics for ecotoxicology decision making are its potential to reduce uncertainty (better science for decision making), and to optimize testing resources (more samples tested faster and cheaper) (Ankley et al., 2006).
Toxicogenomic microarrays to assess endocrine disrupting chemicals (EDC) are now commercially available from EcoArray, Inc. EcoArray scientists worked with the US Environmental Protection Agency (EPA) to develop a 2,000-gene microarray for ecotoxicology testing applications (Larkin et al., 2007). Since that time, they have developed a 22,000-gene array biochip for the fathead minnow that is useful for research and testing in most common freshwater species in the US.
Fish embryos have been found to be a sensitive measure of acute aquatic toxicity (Escoffier et al., 2007). Toxic compounds can impact embryo anatomy and survival. Assays using fish embryos have been particularly good at detecting chemicals that cause developmental delays or pathology. Fish embryo models, while not a true in vitro model, represent one emerging approach to replacing the fish LC50 test. In the European Union, fish embryos are not considered animals, and tests on embryos up until the time they become free-eating larvae, are considered non-animal tests.
Fish cell cultures, mammalian cells, and transgenic cell lines have all been used to predict the ecotoxicity of chemicals. A common endpoint in cell-based assays is cytotoxicity, or cell death. The goal of fish cell-based assays is to be able to predict the in vivo fish LC50 value for the chemicals being tested. Both primary fish cell cultures and fish cell lines from various types of fish organs and tissues have been used (Castaño et al., 2003). Cellular cytotoxicity assays are good screening assays and many types of primary fish cells and fish cell lines can correctly rank the in vivo toxicity of chemicals. However, the cell assays are less sensitive than the in vivo fish test and may incorrectly rank chemicals with specific modes of action (Segner, 2004). Therefore, the proposed role of cytotoxicity assays in a tiered alternative testing strategy has been in priority setting for further hazard testing and/or for the toxicity classification of chemicals and environmental samples.
The report from the ECVAM workshop on the use of fish cells in ecotoxicology describes the use of fish cells and fish cell lines for research into the mechanisms and biomarkers of chemical toxicity in fish, for predicting the fish LC50 test, for genotoxicity testing, for testing environmental samples composed of complex mixtures, and for detecting endocrine disrupting and dioxin-like chemicals (Castaño et al., 2003). The fundamental question of whether mammalian cells could be used in place of fish cells was discussed. Fish and mammalian cells performed comparably in cytotoxicity tests of 50 chemicals; however, further information regarding the species-specific mechanisms of chemical toxicity to fish requires more study in fish cells. One approach to improving the performance of fish cell assays is to use relevant mechanistic-based endpoint assays rather than testing only for cytotoxicity. When a neutral red cytotoxicity assay using a fish hepatoma cell line did not correlate well with in vivo LC50 data for 18 pesticides, the researchers explained that better endpoints in the cell assays based on “cell-specific functions…related to the mode of toxic action of the compound” would provide a more predictive assay than cytotoxicity (Knauer et al., 2007).
In vitro fish cell cytotoxicity assays have consistently been less sensitive than the in vivo fish LC50 test, a drawback that restricts their use in replacing the fish LC50 bioassay. Gülden & Seibert (2005) found that there is a reduced availability of chemicals in the in vitro test systems. Some chemicals are reduced by partitioning into lipids or binding to serum albumin. The lower sensitivity of the fish cells has been found to be a general effect also seen in mammalian cell assays (Gülden et al., 2005). Gülden & Seibert (2007) recently reported the following methods for increasing both the sensitivity of the in vitro assays as well as the correlation of the in vivo-in vitro responses: using inhibition of cell growth as the assay endpoint (rather than cytotoxicity) and measuring cytotoxic potency using the bioavailable free cytotoxic concentration of chemicals in vitro rather than the nominal cytotoxic concentration. Several review articles also include a discussion on methods for improving the sensitivity of cell-based assays (Castaño et al., 2003; Combes et al., 2006).
The replacement of fish in aquatic toxicity testing will require the use of well-defined in vitro test batteries based on the mechanisms of the in vivo biological toxicity responses to a chemical. Repetto, et al. (2001) tested one chemical in a battery with the following results: “The system most sensitive to pentachlorophenol, was micronuclei induction in A. cepa [a plant], followed by D. magna immobilization, bioluminescence inhibition in V. fischeri bacteria at 60 min and cell proliferation inhibition of RTG-2 cells at 72 h. Inhibition of cell proliferation and MTT reduction on Vero monkey cells showed intermediate sensitivity.” Assay endpoints evaluated by Repetto in the salmonid fish cell line RTG-2 were neutral red uptake, cell growth, MTT reduction, and lactate dehydrogenase leakage and activity. In another study, 6 model systems and 18 assay endpoints were assessed to evaluate the potential aquatic toxicity of the commonly used preservative butylated hydroxyanisole (BHA) (Jos et al., 2005). Sensitivity of the different assays (EC50s) ranged from 1.2 to >500 microM, supporting the importance of a test battery approach for obtaining sufficient information on a chemical’s potential ecological impact. The Repetto lab in Spain has continued to publish aquatic toxicity evaluations of specific substances using a battery of test methods (Zurita et al., 2007; Zurita et al., 2007), only a few of which are cited here. Identification and validation of the most useful cellular assays are needed so that they can be incorporated into predictive tiered testing schemes and test batteries that can be considered for regulatory acceptance.
In vitro assays are also being developed for the assessment of endocrine disrupting chemicals. Since EDCs act by a number of mechanisms, in vitro assays will be considered only as screening assays until the complexity of the in vivo responses can be replicated in some in vitro/computational test system. Additionally, there are interspecies differences in the actions and responses to EDCs. Two endpoints for detecting EDCs in fish cells have been identified: vitellogenin production in hepatocytes and estrogen receptor activation/binding (Castaño et al., 2003). A study to compare EDC effects across species and tissues used the H295R human adenocarcinoma cell line and fathead minnow ovary explants (Villeneuve et al., 2007). Both models responded to all six chemicals, and each was more sensitive than the other to some of the chemicals. The H295R assay response was more sensitive and less variable, but the minnow ovary explant assay was more predictive of in vivo minnow effects. The researchers concluded that both assays “have utility for identifying endocrine-active chemicals in screening-type applications.” Another approach using in vitro assays was to use two methods to develop complementary data. Schmieder, et al. (2004), tested 16 chemicals in a trout estrogen receptor binding assay and a fish liver slice assay. The combination of the assays was found to be useful for interpreting the relevance of metabolism in vitellogenin induction and for detecting low affinity chemical binding to fish receptors.
(Q)SARs and other computational approaches will be an important component of any non-animal ecotoxicity test scheme. Villeneuve, et al. (2007), developed a graphical model for ecotoxicogenomics research on EDCs in small fish that “incorporates six compartments representing the major organs involved in the fish reproductive axis and depicts the interactions of over 105 proteins and 40 simple molecules, transcriptional regulation of 25 genes, and over 300 different reactions/ processes.” This model, and microarray data from the experiments (Villeneuve et al. 2007), were used to compare gene expression changes in minnow tissues with the mechanism of action of the EDC being tested.
(Q)SAR predictions are commonly used for some specific applications in ecotoxicology, especially to fill in data gaps for chemicals with structures similar to ones of known toxicity. The US EPA makes available ECOSAR, a Structure Activity Relationships (SAR) program that can be used to predict the acute (and sometimes the chronic) toxicity of chemicals to fish, invertebrates, algae, and other aquatic species. Several review articles describe the various (Q)SAR models that are used to predict acute aquatic toxicity, the basis for their predictions, and expert systems that integrate (Q)SAR selection (Cronin et al., 2003; Bradbury et al., 2003). A (Q)SAR-based decision support system has been developed that can rank chemicals on the basis of their structure alone for persistence, bioaccumulation potential, and toxicity (Mekenyan et al., 2005). Papa, et al. (2007), have also developed and validated a (Q)SAR prediction model for fish bioconcentration factor (BCF) for a wide range of chemical classes. (Q)SAR models have also been developed to predict estrogen disruptor chemicals (Schmieder et al., 2003; Liu et al., 2006; Liu et al., 2007). (Q)SAR models should be particularly helpful in screening databases of chemicals for their potential estrogen activity. Perhaps most useful for reducing the number of fish in regulatory testing would be a (Q)SAR model that could predict the response of fish to a chemical in the 96-hour fish lethality (LC50) assay; Papa, et al. (2005), report progress in developing and statistically validating such a model for the fathead minnow.
Databases of toxicity test data are useful in reducing duplication of studies and therefore unnecessary animal testing. The US EPA’s ECOTOXicology database (ECOTOX) provides toxicity data for 5,900 aquatic and terrestrial species and 8,400 chemicals and includes information on the species, test methods, and test results. The US EPA has also developed the Pesticide Ecotoxicity Database, which contains more than 18,000 records for about 805 pesticide active ingredients compiled primarily from studies conducted by commercial laboratories, but also contains government studies and published data. The US EPA’s Toxic Substance Control Act Test Submission database (TSCATS) is a resource for unpublished technical reports submitted by industry to the EPA, and contains studies on more than 8,000 chemicals, including their environmental effects and environmental fate. The High Production Volume Information System (HPVIS) is another EPA database that provides access to health and environmental effects data submitted on HPV chemicals. Other useful databases include: INCHEM; Environmental Health Criteria Monographs; and Hazardous Substances Data Bank (HSDB).
The following invited commentary provides the AltTox community with a new perspective on ecotoxicity testing: