“People often ask, ‘What is the single most important environmental problem facing the world today?’ The single most important problem is our misguided focus on identifying the single most important problem! … because any of the dozen problems, if unsolved, would do us great harm and because they all interact with each other.” Jared Diamond [1].

The field of ecotoxicology studies the effects of anthropogenic chemicals on ecosystems at different levels of biological organization, from the molecular and cellular level to entire ecosystems. Aquatic ecosystems are often impacted by chemical pollution, originating from municipal and industrial wastewater effluents (point sources), airborne deposition as well as runoff from urban and agricultural areas (diffuse sources). Ecotoxicology is closely related to, but sometimes distinguished from stress ecology, which considers a broader range of natural stressors such as the effects of temperature or oxygen depletion on individuals, populations and communities, however these parameters directly impact on toxicity. There is further separation between ecotoxicology and environmental toxicology, which is often rather artificial, with a tendency of environmental toxicology being more focused on the level of individual organisms, including humans, or cells, and ecotoxicology being more focused on the level of communities and ecosystems.

One of the core missions of ecotoxicology is to understand the mechanisms by which contaminants perturb normal biological performance (their mode of action), in order to develop appropriate measures to prevent adverse outcomes resulting from environmental contaminants. There is a wide range of possible contaminant effects that can compromise the ecological fitness of individual organisms or populations. Ultimately, the impact of a toxic contaminant or contaminant mixture depends on the relative sensitivity of a species, community or ecosystem, and the intensity and timing of exposure. Acutely toxic events—most notably fish kills—which were relatively common a few decades ago, are now rarely observed in most industrialised countries, however, even sublethal toxicity can lead to severe impacts on entire populations.

Of the vast number of substances that have been introduced to aquatic systems around the world, a number posing serious environmental threats have primarily been identified either through single-species toxicity testing in the laboratory, or they have been brought to light as a result of observing biological effects in situ. Examples include dichlorodiphenyl-dichloroethylene (DDE) and eggshell thinning in birds [2], sex changes in freshwater fish associated with endocrine disrupting chemicals, EDCs [3], and tumors in marine fish associated with polycyclic aromatic hydrocarbon (PAH) accumulation [4]. These examples highlight the need for techniques that not only detect overt damage to organisms exposed to pollutants, but also the less obvious biochemical and physiological impairment that might ultimately result in ecological damage [5].

With more than 67 million organic and inorganic substances known to date [6], monitoring and assessing effects of chemical pollution necessarily faces great challenges. In light of recent legislation by the European Union, such as Registration, Evaluation, Authorisation and Restriction of Chemical substances (REACH) and the Water Framework Directive (WFD), ecotoxicology needs to become more proactive. Consequently, the development of reliable predictive tools to assess the impacts of chemicals as well as robust and sensitive retrospective tools to monitor water quality will be important for advancing this field.

Toxic substances can cause effects at different levels of biological organisation, from molecular to ecosystem levels. Of primary concern is the protection of aquatic organisms at the population or ecosystem level; it is therefore important to bridge the gap between the relatively short-term (acute) effects that can realistically be quantified in laboratory or field experiments, and longer-term (chronic) ecological effects.

Forty four years after the word “ecotoxicology” was coined by René Truhaut in 1969 [89], technological and computational advances have enabled a number of assessment methodologies to arise, as well as new technology to continually evolve; e.g., gene expression assessments using differential display reverse transcription PCR (dd-rtPCR) to whole transcriptome assessments using microarrays and more recently direct sequencing (RNASeq), or single protein analysis using western blots, to high-throughput proteomics. The rapid development of new technologies such as transcriptomics, proteomics and metabolomics are revolutionizing ecotoxicology [90], however, with this development, the volume of generated data has also increased exponentially, often beyond the capacity of what can realistically and coherently be analyzed. These “new” technologies have had a significant impact on medical research and are gradually being incorporated into ecotoxicological studies. This is a challenging task since medical research primarily focuses on responses at the level of the individual and one species, whereas ecological research attempts to utilize these systems to encompass populations, species communities and ecosystems. Nevertheless, a number of successful ecological case studies have demonstrated the usefulness of novel technologies, in complement with traditional approaches, in generating valuable information on the effects of, and responses to, environmental stressors upon biological systems (e.g., [91–94]). The availability and accessibility of numerous genomic techniques and their approaches have made them favorable tools that are growing in use within interdisciplinary research fields. High-throughput gene and protein techniques are currently encompassed into systems biology [90,95], attempting to highlight the complex interactions between responses that span from cellular mechanisms to the whole animal, and extrapolating these, via life cycle models to population level effects.

Genomic approaches, or “omics” as they are currently referred to, encompass transcriptomics (mRNA), proteomics (protein), and metabolomics (metabolic signatures from resulting activity), and incorporate epigenetics (heritable changes in expression), and genotyping (DNA). The proliferation of different genomic approaches over the past decade, through a combination of advances in biological, instrumental and bioinformatic techniques, has thus generated a means of characterizing biological processes in response to stressors, yielding unparalleled information on the molecular and biochemical status of organisms. Numerous reviews are currently available on the perspectives and future directions for the inclusion of mechanistic research in aquatic toxicology (e.g., [96]).

Transcriptomics, otherwise known as “global analysis of gene expression”, examines the genes and corresponding biochemical pathways that are involved in various biological processes. These techniques are highly sensitive indicators of an organisms' interaction with their environment, because they capture the physiological responses initiated at the molecular level. However, it is important to highlight that RNA is not the final product of a gene and that the functional expression of a gene is carried out by the synthesized protein. Protein expression, however, is influenced and regulated by many processes downstream of mRNA synthesis. Transcriptional expression levels can be utilized to understand the functional network of the biochemical pathways they are involved in, and thus are useful in the characterization of biological processes of interest. Notable examples are those of Spromberg and Meador [97], who were able to associate immune suppression, reproductive dysfunction and somatic growth impairment in salmonids, to population level responses using life history models, Heckmann et al. [98] and Connon et al. [99] who were successful in combining molecular and organismal stress responses in Daphnia magna, by assessing mechanistic responses to exposure indicative of chronic consequences for population growth rates, and Garcia-Reyero et al. [92,100] who investigated synergistic and antagonistic effect of EDC mixtures, identifying specific signatures relating to the modes of action of each chemical, and highlight the complexities of assessing multiple compound in wastewater treatment effluent. Studies of this kind emphasize the importance of bridging the assessment of toxicity responses at environmentally relevant (or low) concentrations that can mechanistically explain the effects contaminants may have at higher levels of biological organization and resulting population dynamics.

Proteomics represents the high-throughput assessment of the functional responses of gene expression; the proteins and peptides, as well as protein-protein interactions. Combined with bioinformatic analyses, proteomics is used to assess functional biochemical pathways that respond to environmental and/or contaminant stimuli. As indicated above, these responses may not necessarily correspond with transcriptional changes due to factors, including rates of degradation, translational inefficiencies, and relativity to the status of the cell at the assessed time [90]. As such they may offer a more robust approach for ecotoxicological risk assessments [93], however analysis is somewhat limited by the relatively large sample volume requirements. Proteomic tools have been successfully utilized in studies identifying hypoxic effects in fish brains [101], chemical and site-specific signatures [94,102], specific effects from exposures to microcystins [103,104], combinations of transcriptomics and proteomics in the assessment of neurological damage in response to dieldrin exposure [105], as well as in numerous studies investigating the effects of endocrine disrupting chemicals (as reviewed by Martyniuk and Denslow [106]).

Metabolomics measures the concentrations of metabolites that represent enzymatic activity upon xenobiotics, and associates these, through bioinformatic techniques, with changes in biological functions in the exposed organism, thus its development and continual enhancement has contributed significantly to the field of ecotoxicology. A major advantage that metabolomics has over transcriptomics and proteomics, is that resulting metabolites do not differ between taxa, thus this approach is not species-specific, permitting for the same toolbox to be used a broad range of species.

Lin et al. [107] summarize the emergence of metabolomics within the “omics” field and highlight how these methodologies can be used within regulatory ecotoxicology, and Ludwig and Viant [108] present a valuable review of advances in metabolomics methodologies that enhance the confidence in metabolite identification and reduce batch-to-batch variation. One of the main benefits of metabolomics is that it provides a highly suitable assessment approach for field studies. Prominent studies in metabolomics include those of Viant [109] who describes the use of metabolite profiles to determine Chinook salmon exposure responses, differentiating between amongst three classes of pesticides, Ekman et al. [110], who used metabolite analyses linking feminization of male fathead minnows on exposure to EE2, and Poynton et al. [111] who utilized a combination of transcriptomics and metabolomics on D. magna haemolymph, to assess effects of sublethal cadmium exposure, identifying decreased levels of digestive enzymes, elevated transcription of cuticular proteins, and genes involved in oxidative stress that corresponded with effects on amino acids and fatty acids determined via metabolomics. Recent advantages, capabilities and applications of metabolomics tools within ecological research fields ranging from ecophysiology to ecotoxicology are reviewed in Viant et al. [112].

Whilst these technologies have proven to be useful in elucidating modes of action of toxicants and can contribute to the risk assessment process as part of a weight-of-evidence approach, there is a strong need for critical studies that relate molecular changes to ecological adverse outcome [113], as well as for studies addressing mixture effects [114]. There are several examples, however, which suggest that mechanistic links between omics responses and effects at other levels of biological organization (behavior, growth, predation risk, fitness, and mortality) can be established (e.g., [115–118]). Omics technologies continue to advance, becoming more affordable, efficient and generating rich datasets, which is facilitating the incorporation of these techniques into risk assessments and assisting regulatory decisions [113].

Epigenetics is an emerging field that is rapidly being incorporated into ecotoxicological studies. It investigates the alterations in gene function or cell phenotype, without changes in DNA sequences, that may result, for example from methylation, or histone modification, and can be triggered by alterations in environmental stressors [119,120]. Vandegehuchte and Janssen [121] and Head et al. [120] review the effects of epigenetic induced changes in unexposed offspring, stressing that populations can experience the effects of their ancestors' exposure to chemicals, and that there is a need to also evaluate the persistence of chemical exposure-induced epigenetic effects in multiple subsequent generations. Harris et al. [122] described Daphnia magna as being an ideal model organism for epigenetic research as its parthenogenetic life-cycle permits assessments to be conducted in the absence of confounding genetic differences. Additional attributes that make D. magna an optimal and internationally recognized model organism for ecotoxicological studies are its rapid life-cycle, ease of culture and low cost, as well as the occurrence of multiple phenotypic differentiations, so-called polyphenisms, leading to modifications such as neckteeth and helmets in response to hormonal cues from predators (e.g., Chaoborus spp). Vandegehuchte et al. [123,124] evaluated DNA cytosine methylation transgenerational effects in populations of D. magna, which resulted in decreases in body length and reduced numbers of offspring following contaminant exposure.

Finally, another genomic approach that will undoubtedly be increasingly incorporated into ecotoxicological assessments is genotyping. It is potentially genotyping techniques that will assist comparative species toxicology, and help understand what sensitivity means from one organism to another. This is an especially important parameter that must be considered within the context of ecosystem resilience, as differences in intra-species sensitivity will affect both genetic structure of populations as well as community structure, especially in the presence of a multitude of contaminants within aquatic ecosystems. Morgan et al. [125] highlight the need for the awareness of microevolution in ecotoxicological studies, using high levels of genetically differentiated metal resistance in invertebrate populations as an example. This review advocates the inclusion of genotyping tools, particularly in the assessment of field populations, in order to understand and incorporate likely adaptations of sampled organisms to polluted sites. Babin-Fenske et al. [126] conducted a phylogenetic analysis of Hyallela azteca species, contrasting the different lake habitats, revealing two distinct populations associated with historically contaminated and low impacted lakes, respectively. Thus, when attempting to conduct comparisons using field-caught organisms, or for laboratory studies using sentinel organisms, it is essential to include genotype assessments in conjunction with epigenetic studies in order to understand relative intra- and inter-species sensitivities.

None of these techniques would be of use without bioinformatic approaches and databases used to store gained knowledge. The bioinformatic technology is continually attempting to assess the ever-increasing datasets, and has also expanded exponentially in recent years, but is somewhat restricted by the provision of limited data through existing databases (e.g., Toxnet, Comparative Toxicogenomics Database) on interactions between disease contaminant effects and adverse outcomes that are based on published data.

It is not the use of individual tools, determining individual endpoints, but rather an integration of many fields of expertise; multidisciplinary efforts, which will permit the generation of more accurate toxicological information to effectively guide ecotoxicology and environmental management in the future. Ecotoxicologists are especially aware of this, and thus are rapidly moving away from single endpoint assessments to investigating biological systems or pathway effects that may lead to adverse outcomes. Numerous reviews and studies are addressing this need for integration of many schools of thought. Kramer et al. [127] reviewed the application of a suite of techniques within ecotoxicology that will permit bridging of mechanistic approaches with population-level effects. Ankley et al. [128] report the use of a multidisciplinary approach in which mechanistic assessments, in conjunction with bioinformatics, modeling, and whole organism tests, successfully determine links across biological levels of organization. Not only are multidisciplinary approaches suitable for the assessment of contaminant exposure, but are also directly applicable to disease, which will directly affect sensitivity to contaminants and vice versa [129]. Hostetter et al. [130] and Connon et al. [131] describe the use of a multidisciplinary approach synchronously utilizing molecular indicators of health status, with histopathology, external parameters of disease, and infection classes, which were strong predictors of survival probability of out-migrating salmonids. Similarly, Miller et al. [132], assess similar responses using genome-wide studies that predict salmonid migration and spawning failure.

In order to link the effects of pollutants through various levels of biological organization to ecosystem health, an integrated and multifaceted approach is required, that is based on existing knowledge that correlates processes that include the route of uptake, detoxification and pathology in between each of the assessed levels [133]. Portfolios of carefully selected biological indicators, including standard whole organism bioassays as well as biomarkers and in vitro tests, are required to evaluate toxicity. There is also a need to directly link chemical pollution in ecosystems to measurable toxic effects [134]. Chemical concentration data can only be of value for environmental impact assessments when effect levels and mechanisms of action of chemicals are already known. Thus the value of biological testing cannot be underestimated when attempting to detect toxicity before populations are negatively impacted. The sensitivity of an organism to a particular contaminant relates to its ability to mechanistically cope with the exposure; the ability of an organism to sequester or eliminate resulting metabolites from its body, and its overall health status. Recent mechanistic research approaches, such as the omics techniques described above, have permitted the assessment of sublethal toxicology, driving experimental designs to encompass environmentally relevant, low dose contaminant concentrations that are often below the limit of detection of current analytical chemistry protocols.

The majority of omics studies published within the last decade (too many to cite) have relied on more than the interpretation of genes, proteins or metabolite data, that indicate mechanistic effects, or modes of action of contaminants. Most of these studies have also integrated at least one other endpoint, at higher levels of biological organization; associating mechanistic responses with reproductive growth rate, offspring viability and sex ratios, masculinization or feminization in fish, disease, growth and development or survival, or with other physiological parameters such as swimming performance and mating behavior. The greater the integration of the above endpoints, the greater will be the need for multidisciplinary collaborations, and greater will be our understanding of the adverse effects caused by single contaminants, their mixtures, and their interactions with biotic and abiotic parameters within a particular ecosystem.

Adverse outcome pathways (AOPs) have been defined by Ankley et al. [135] as “a conceptual construct that portrays existing knowledge concerning the linkage between a direct molecular initiating event (e.g., a molecular interaction between a xenobiotic and a specific biomolecule) and an adverse outcome at a biological level of organization relevant to risk assessment”. This approach establishes a basis upon which concepts and testable hypotheses, to be tested by multi-disciplinary fields of study, can be constructed. Examples of such studies include those of Villeneuve et al. [136] who provide a review of the integration of toxicogenomics, as a complementary tool to traditional methodologies, when applied to risk assessment. Altenburger et al. [114] highlight the use of omics technology for mixture studies, stressing the need for dose and time mixture toxicity models that include bioinformatics and stress response concepts that can integrate the multiple responses into testable hypotheses. Testable hypotheses will permit the understanding and thus the prediction of an adverse effect. In order to adopt this approach, researchers need more than endpoints at the mechanistic and population levels within a study; assessments at associated and subsequent levels of biological organization are required.

As indicated above, biological systems do not respond to perturbations as single molecules, nor through single pathways, but rather as multiple-level interactions, within networks of cellular components [137]. While AOP assessments aim to establish associations between different levels of biological organization, systems biology aims at understanding impacts of contaminants upon the whole (Figure 5). Systems biology is thus defined as the assessment of interacting networks in response to perturbations in order to discover, understand and predict the emerging properties of the system [137], by examining the dynamics of cellular to organismal functions, rather than its individual components [138].

By far, the greatest benefit of systems biology is that it is based on conceptual, descriptive and mathematical models used to describe these perturbations, and their value in determining effective outcomes at higher level of organization. A recent example of such approach involves the use of reverse engineering; employing mathematical and statistical algorithms to determine relationships between genes, proteins and metabolites. Perkins et al. [139] describe this approach utilizing high-density transcription data to identify AOPs that relate to the disruption of the hypothalamus-pituitary-gonadal endocrine axis in fathead minnows. The authors then assessed this approach in contrast with biochemical biomarkers; vitellogenin, testosterone and estrogen, as surrogates of fecundity, inferring effects at higher levels of biological organization.

Current and planned studies based on AOP approaches are expected to strengthen the knowledge base required for systems biology to become a successful predictive tool in the ecotoxicological context, and for this information to be used towards understanding pollution effects on community and ecosystem functioning.

One of the greatest challenges in ecotoxicology ultimately lies in linking the effects of chemical exposure to ecosystem-level effects and ecosystem services [140]. Generally, effects on ecosystem functioning and effects on ecosystem services need to be distinguished. Ecosystem effects comprise all abiotic and biotic changes that exceed the natural change rate or frequency. In some cases, such changes can strongly affect the functioning of entire ecosystems, so-called phase shifts. On the other hand, the concept of ecosystem services relies on all values and functions that ecosystems provide for mankind, e.g., including provision of food, recreation, aesthetic and ethical values [140,141]. Ecosystem services are particularly strongly valued if direct links between contamination and human health are evident, e.g., in the case of fishes that exceed certain levels of methylmercury and cannot be eaten or sold. From the perspective of ecosystem functioning, the interaction of chemical stressors with biodiversity is most important, comprising the range from genetic diversity, to species, communities, and habitat types. Whilst effects on higher levels of biological organization are more obvious, contaminant pollution also has the potential to exert subtle selective power on the gene pool, e.g., by selecting or favoring individuals that are more resistant to certain types of contaminants than others. Such selection can quickly (i.e., within few generations) result in evolution of gene pools, e.g., the reduction of genetic variation.

Traditionally, most studies in the context of assessing ecosystem effects rely on time-series correlations between contaminant concentrations and monitored community-level changes in particular types of habitat. One major weakness of such an approach is that correlation-based analyses are not suitable for verifying cause and effect, that changes on an ecosystem-level often lack replication and are typically induced by many other factors (e.g., temperature or salinity shifts) that cannot be separated from the effects of contaminants. At the same time, such approaches require a high monitoring effort in the field and, with few exceptions (e.g., [27]), contamination of natural ecosystems for scientific experiments are not considered ethical. Consequently, ecosystem level effects are typically being studied in small-scale mesocosms or microcosms that resemble natural ecosystems, containing at least a minimum number of representative species, ideally also keystone species whose presence or absence critically influences the entire functioning of such systems. Mesocosms have been widely applied to identify long-term community effects of low toxicant concentrations but often face the problem that results are obscured by confounding factors and high variability between replicates (e.g., [142,143]). This variability has been identified as a major source of false negative results, i.e., concluding a chemical has no effect when it really has [143–146]. Whether community responses can better be evaluated using a species-based principle response curve (PRC) approach [147] or with the SPEAR_mesocosm approach [142,148] is still controversial [149,150].

In cases in which no data on ecosystem effects are available, modeling approaches are typically being applied. Traditionally, these approaches derive “safe” environmental concentrations of toxicants by extrapolation factors or by statistical extrapolation from a set of single species toxicity data. One major weakness of such approaches is that species' interactions which can be important in natural ecosystems (e.g., predator-prey interactions) are not considered. Ecology-based ecological modeling alternatives have been suggested to account for ecological interactions, but it is largely unexplored how well predictions from these models quantitatively match with large-scale experimental studies [151].

Chemical characterization by itself does not provide specific biological information about potential hazards to organisms. Traditional ecotoxicological approaches that utilize individual endpoints lead to assumptions and uncertainty in factors that can erroneously infer effects at higher levels of biological organization. By truly integrating monitoring strategies, involving chemical analyses, a suite of ecotoxicological tools, the study of population/community responses in the same water body and, preferably, time of the year, a more holistic picture can be obtained. Such an approach can tell us whether the chemicals analyzed are bioavailable and giving rise to negative health effects in aquatic organisms, and whether population effects are observed. The concept of weight of evidence (WOE) refers to the integration of data generated within such a multidisciplinary approach including data from different studies, or lines of evidence (LOEs). Thus in the 21st century, ecotoxicology needs to evolve, moving away from traditional toxicology approaches that are still in use by most regulatory agencies, and encompass integrative tools and models to evaluate the risk posed by existing and new contaminants as part of their registration, potential use, and application. Ideally, integrative approaches of linking mechanistic responses on the cellular level with whole organism, population, community and ecosystem effects and services would be desirable for a comprehensive assessment of ecotoxicological risk. Since such approaches are difficult and costly, the identification of key links between responses at different levels of biological organization with ecosystem functioning, and associations with ecosystem services, can be a first useful step towards making ecotoxicology more fundamental and effective.