**6. Perspective: toward new markers and an integrative approach**

Contamination of aquatic environment is multifaceted as it combines chemical, physical, and biological pollutions, which could be influenced by a wide range of (a)biotic factors. Both short- and long-term cascading negative effects of such pressures on ecosystems still remain difficult to evaluate considering the complexity and the diversity of exogenous inputs in the aquatic environment. While the contamination level in a particular site can be quite readily determined through chemical analysis as defined by the presence of "substances that would not normally occur or at concentrations above the natural background," the assessment of the "pollution status" also integrates the notion of chemicals bioavailability and biological impacts induced by contaminants within the considered environment [102]. Assessing the ecological health of an ecosystem should thus address the following questions, initially enunciated for metallic contaminants by Chapman et al. [103], but which can be generalized to all environmental contaminants: (1) do chemicals/biologicals accumulate in biota above background (or reference) concentrations? (2) once accumulated, are they bioreactive? and if so, (3) what are the incidence and severity of the induced effects (acute, long-term sublethal, individual, population effects, etc.)? Thus, a relevant and successful water quality assessment has to be based on combined analyses integrating the greatest possible number of these various parameters. This also includes the sanitary status of aquatic organisms which are chronically exposed to a cocktail of pollutants in their environment. Despite a large number of publications on invertebrate biomarkers, little of them reached a level of validation allowing to be recommended as efficient tools for an accurate evaluation of the quality of the aquatic compartment. Nonetheless, new integrative approaches could permit to remove scientific (define the reference levels) and technical (integration of several responses) obstacles encountered in environmental-risk assessment strategies. The use of OMICS tools in field studies has gained great interest for some years as it gives broad information on pollutant modes of actions, considerably increasing knowledge and an understanding of such mechanisms. Finally, clustering data in a weight-ofevidence approach represents a powerful and practical tool to facilitate the decision-making processes of environment managers within the framework strategies.

#### **6.1. Genomics and proteomics in ecotoxicology**

Although next-generation sequencing has made great progress over the last decade, genome sequencing of animal models is still far to be a trivial task. This is due to their large genome and the difficulty for assembling too numerous short reads without any scaffold from a closely related species. Furthermore, the delineation of coding open-reading frames and their functional annotation is difficult and subjected to an important bias toward easy annotation of conserved genes but detrimental to species-specific genes. Till now, no complete genome is available for any *Dreissena* and genetic markers are relatively scarce [104–106]. The genomes of marine mussels such as *M. galloprovincialis* are only partially known [107, 108]. Transcriptomics is fully complementary to genomics and adds value in highlighting genes of interest if their expression is regulated. However, this approach has till now been really poorly used for characterizing *Dreissena*. Very few studies based on cDNA microarray have been published, specifying the molecular mechanisms at the early stage of underwater adhesion of the zebra mussel [109], or the effect of seasonal and environmental variations on the physiology and metabolism of *D. polymorpha* [110]. Today, cDNA microarray has been advantageously replaced by RNAseq. High-throughput sequencing of cDNA allows comparing the transcriptome from several conditions and highlighting the most modulated genes in terms of expression levels. In addition, RNAseq data may be used to construct a six-reading frame RNA-translated database that can be used for discovery proteomics. This concept is the basis of numerous proteogenomics analysis of various animal models or even nonmodels [111]. For example, the amphipod *Gammarus fossarum* has been scrutinized in detail to document the reproductive system of amphipods [112] or to understand the response to endocrine disruptors [113].

Till now, *Dreissena* has been worked out in terms of proteomics only with 2D-PAGE traditional approach. For example, [114] analyzed *D. polymorpha* exposed to benzo(α)pyrene and focused on 28 proteins. A set of 16 proteins were found more abundant and 12 were noted as less abundant in exposed mussels. They could be identified after MALDI-TOF/TOF mass spectrometry measurements on their tryptic peptides. Such methodology is known to focus only on the most abundant and soluble proteins [115], thus explaining the relatively low number of protein hits identified by this approach. For sure, high-throughput shotgun proteomics based on next-generation mass spectrometers should allow today documenting thousands of proteins and delineating candidate biomarkers after the analysis of specific exposure conditions. Importantly, these candidate biomarkers require a strict validation consisting in monitoring the candidate biomarkers for a large cohort of animals and various conditions [116]. Such monitoring is ideally carried out with targeted proteomics based on the selected reaction monitoring (SRM) quantitative approach. SRM mass spectrometry assay allows high-throughput multiplex analysis, with fewer samples required. It is efficient in terms of time and cost and is able to reliably detect different proteins across a broad dynamic range of concentrations as recently reported for *G. fossarum* [117]. Thus, in our opinion, OMICS tools are today pertinent for obtaining major insights into the most important molecular mechanisms of mussels.

#### **6.2. Weight of evidence**

biomarkers, little of them reached a level of validation allowing to be recommended as efficient tools for an accurate evaluation of the quality of the aquatic compartment. Nonetheless, new integrative approaches could permit to remove scientific (define the reference levels) and technical (integration of several responses) obstacles encountered in environmental-risk assessment strategies. The use of OMICS tools in field studies has gained great interest for some years as it gives broad information on pollutant modes of actions, considerably increasing knowledge and an understanding of such mechanisms. Finally, clustering data in a weight-ofevidence approach represents a powerful and practical tool to facilitate the decision-making

Although next-generation sequencing has made great progress over the last decade, genome sequencing of animal models is still far to be a trivial task. This is due to their large genome and the difficulty for assembling too numerous short reads without any scaffold from a closely related species. Furthermore, the delineation of coding open-reading frames and their functional annotation is difficult and subjected to an important bias toward easy annotation of conserved genes but detrimental to species-specific genes. Till now, no complete genome is available for any *Dreissena* and genetic markers are relatively scarce [104–106]. The genomes of marine mussels such as *M. galloprovincialis* are only partially known [107, 108]. Transcriptomics is fully complementary to genomics and adds value in highlighting genes of interest if their expression is regulated. However, this approach has till now been really poorly used for characterizing *Dreissena*. Very few studies based on cDNA microarray have been published, specifying the molecular mechanisms at the early stage of underwater adhesion of the zebra mussel [109], or the effect of seasonal and environmental variations on the physiology and metabolism of *D. polymorpha* [110]. Today, cDNA microarray has been advantageously replaced by RNAseq. High-throughput sequencing of cDNA allows comparing the transcriptome from several conditions and highlighting the most modulated genes in terms of expression levels. In addition, RNAseq data may be used to construct a six-reading frame RNA-translated database that can be used for discovery proteomics. This concept is the basis of numerous proteogenomics analysis of various animal models or even nonmodels [111]. For example, the amphipod *Gammarus fossarum* has been scrutinized in detail to document the reproductive system of amphipods [112] or to understand the response to endocrine

Till now, *Dreissena* has been worked out in terms of proteomics only with 2D-PAGE traditional approach. For example, [114] analyzed *D. polymorpha* exposed to benzo(α)pyrene and focused on 28 proteins. A set of 16 proteins were found more abundant and 12 were noted as less abundant in exposed mussels. They could be identified after MALDI-TOF/TOF mass spectrometry measurements on their tryptic peptides. Such methodology is known to focus only on the most abundant and soluble proteins [115], thus explaining the relatively low number of protein hits identified by this approach. For sure, high-throughput shotgun proteomics based on next-generation mass spectrometers should allow today documenting thousands of proteins and delineating candidate biomarkers after the analysis of specific exposure conditions. Importantly, these candidate biomarkers require a strict validation consisting in monitoring

processes of environment managers within the framework strategies.

**6.1. Genomics and proteomics in ecotoxicology**

56 Organismal and Molecular Malacology

disruptors [113].

The WOE approach is based on the packaging of a wide variety of data within several lines of evidence (LOEs) in which the contamination level—assessed through chemical analyses—is combined to bioavailability (bioaccumulation) analysis, and biological responses (biomarkers) on key species and/or model organisms (bioassays) at different levels of biological organization, from the molecular to the community level [118, 119]. The resulting environmental diagnosis is then based on the calculation of a hazard index for each LOE, which is next set out on an evaluation grid allowing a clear and rapid classification of hazard [118, 120, 121]. A global hazard evaluation is also proposed through the compilation of all calculated LOE indexes within a single one that is also finally assigned to a hazard class. The WOE approach is applicable to various matrices such as effluent, water, and soil, as well as for more global environmental diagnosis like aquatic and terrestrial ERA. The WOE model developed by Piva et al. [121] was mainly applied to the quality assessment of harbor area using fish species (European eel, *Anguilla anguilla*) and/or mussels (Mediterranean mussels, *M. galloprovincialis*) as bioindicator organisms. These studies undoubtedly demonstrated the relevance and the performance of the procedure to diagnose ecosystem health status in chronically impacted area (e.g., industrial harbors, natural crude oil, and gas seepage) [122, 123] as well as in accidental pollution events as demonstrated by its use in the Mussel Watch program following the Costa Concordia wreck [124]. Mussels were relevantly used in these studies for their suitability in translocation (caging) procedures and their ability to reflect environmental pollution levels through bioaccumulation measurements (see part 2) and biomarkers analysis (see part 3). The elaboration of each hazard index within the WOE approach relies on the initial calculation of ratio-to-reference (RTR) values. It thus supposes that reference levels are available for every end point integrated in the model. In the abovementioned studies, the reference levels were determined examining the biological responses in control organisms maintained in clean water under laboratory conditions or transplanted at a reference site. However, the laboratory (controlled) conditions are far removed from those during *in situ* exposure as control organisms are not submitted to any variations of their environment that naturally occur in field (e.g., temperature variations, general physico-chemistry of the water column, etc.) and which could modulate the biological responses of the organisms with no link with the contamination status of the environment. Transplantation of organisms at a reference site is also commonly used to avoid such bias. However, a "perfect" reference station would assume that (i) the site is geographically close enough to reflect the natural state of the studied environment; (ii) the exposure conditions are exactly similar to those at the other sites (in terms of temperature, physicochemistry, etc.); and (iii) that there is absolutely no anthropogenic contamination which could induce any modulations in the biological responses, even in a limited way. There is no doubt that the determination of such station in each studied area is not realistic—if not utopic—and that the use of control organisms (at a reference station or maintained in laboratory conditions) to set the reference levels integrated in the WOE approach generates a bias in the ecological health status assessment.

An alternative was proposed by Barjhoux et al. [125] to address these concerns. Briefly, the study proposes an application of the WOE strategy to a freshwater system: the Seine River (France), well known to be submitted to heavy anthropogenic pressures through important industrial, agricultural, and urban activities. The three studied sites were located upstream (Marnay, in a non-urbanized area) to downstream from Paris conurbation (Bougival and Triel, respectively, situated at 40 and 80 km from Paris). The dataset selected for WOE integration included (i) chemical contamination levels, (ii) bioavailability (bioaccumulation) measurements, (iii) biological effects in field-transplanted organisms (biomarkers in gammarids), and (iv) (eco)toxicological responses assessed using laboratory bioassays. The strength of the quality assessment proposed in this study lies on the use of the same population of gammarids for bioaccumulation and biomarkers measurements. Reference and threshold values were established using modeling developments quantifying the natural variability of the studied markers in relation to identified confounding factors. These reference/threshold levels were integrated in the WOE approach as they clearly enhance the reliability of *in situ* methodology and allow its implementation at a large spatial and temporal scales [126, 127]. The calculated WOE indexes clearly reflected the anthropogenic gradient along the Seine River,

**Figure 2.** WOE indices and associated hazard classes integrating the results of each LOE calculated for the three stations during four sampling campaigns (C1–C4) and annual average (AA) values. The hazard class attributed to each LOE hazard quotient (HQ) is summarized in the table below. ChemHQwater/sed, water/sediment contamination HQ; BioavHQ, bioavailability (bioaccumulation) HQ; BiomHQ, biomarker HQ; ToxHQwater/sed, bioassay-based HQ on water/sediment samples. Note that in the C2 campaign, only data on water contamination and bioavailability are evaluated. The integral version of the article is available at: http://link.springer.com/article/10.1007/s11356-016-6993-6. The figure was reproduced with permission of SpringerNature publisher.

with values increasing from upstream to downstream of Paris (**Figure 2**: [125]). The results also highlighted some seasonal variations in the hazard class attributed to each site with the winter campaign showing lower level of perturbation than the three other campaigns.

Accordingly, the use of external reference values and thresholds eliminated the need for a reference site in the study area, which could be very problematic in large rivers subjected to multiple and diffuse pressures. The results of this study also reveal that at the upstream site, generally used as a relative reference or control site in previous investigations in this area, the low contamination levels nonetheless resulted in low but significant biological effects. The WOE approach applied in this study proved to be efficient and relevant in terms of both global environmental hazard diagnosis and seasonality analysis. The in-depth characterization of the baseline levels and relevant effect thresholds for environmentally relevant end points is thus a challenge and might be vigorously pursued and developed further over the coming years to lead to homogenized ERA procedures between the various environmental institutions. In particular, several research programs are in progress to define basal reference levels, effect thresholds, and confounding factors of the biological responses in *D. polymorpha* (from molecular to population) in order to routinely include this promising species in WOE approaches dedicated to freshwater environment quality diagnosis.
