lable at ScienceDirect

Environmental Pollution 197 (2015) 90e98

Contents lists avai

Environmental Pollution

journal homepage: www.elsevier .com/locate/envpol

Transcriptomic resources for environmental risk assessment: a casestudy in the Venice lagoon

M. Milan a, *, M. Pauletto a, L. Boffo b, C. Carrer c, F. Sorrentino d, G. Ferrari d, L. Pavan e,T. Patarnello a, L. Bargelloni a

a Department of Comparative Biomedicine and Food Science, University of Padova, Viale dell'Universit�a 16, 35020 Legnaro, PD, Italyb Azienda ULSS n. 14 Chioggia, Italyc Thetis S.P.A. e Magistrato alle Acque, Laboratorio Centro Studi Microinquinanti Organici (C.S.M.O.), Via Asconio Pediano, 9, 35127 Padova, Italyd Ufficio Tecnico per l'Antinquinamento, Magistrato alle Acque di Venezia, San Polo 737, Riva del Vin, 30125 Venezia, Italye G3 Industriale, Via Milano 18, 30020 Marcon, VE, Italy

a r t i c l e i n f o

Article history:Received 11 September 2014Received in revised form24 November 2014Accepted 1 December 2014Available online

Keywords:Environmental risk assessmentGenomicsVenice lagoonRuditapes philippinarumGood environmental status

* Corresponding author.E-mail address: massimo.milan@unipd.it (M. Mila

http://dx.doi.org/10.1016/j.envpol.2014.12.0050269-7491/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

The development of new resources to evaluate the environmental status is becoming increasinglyimportant representing a key challenge for ocean and coastal management. Recently, the employment oftranscriptomics in aquatic toxicology has led to increasing initiatives proposing to integrate eco-toxicogenomics in the evaluation of marine ecosystem health. However, several technical issues needto be addressed before introducing genomics as a reliable tool in regulatory ecotoxicology. The Venicelagoon constitutes an excellent case, in which the assessment of environmental risks derived from thenearby industrial activities represents a crucial task. In this context, the potential role of genomics toassist environmental monitoring was investigated through the definition of reliable gene expressionmarkers associated to chemical contamination in Manila clams, and their subsequent employment forthe classification of Venice lagoon areas. Overall, the present study addresses key issues to evaluate thefuture outlooks of genomics in the environmental monitoring and risk assessment.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Ecosystems are complex entities subjected to sudden changesdue to environmental variables, as well as to anthropogenic impact.Over the past century, the land surface has been deeply trans-formed by human action, leading to habitats impoverishment andworrying consequences on living organisms (Denslow et al., 2007;Vitousek et al., 1997). As a result, several countries have beencompelled to implement legislations aiming to achieve sustainablemanagement of water resources and to protect and improve theecological status.

The European Water Framework Directive (WFD, 2000/60/EC)and the European Commission Directive (Directive, 2008/56/EC)established a legislative framework to define “good ecological sta-tus” and delineate the corresponding community actions in termsof water policy to protect and improve the ecological status ofEuropean surface waters (rivers, lakes, transitional waters, and

n).

coastal waters) by 2015. The WFD defines the ecological status as“an expression of the quality of the structure and functioning ofaquatic ecosystems associate with surface waters” (Ar. 2.21) andrequires site-specific evaluations of water bodies by integratingbiological, physicochemical, chemical, and hydro-morphologicaldata. Accordingly, in order to assess the ecological status of ma-rine environments, water policies previously based on the estab-lishment of emission limits, have been updated through thedevelopment and integration of approaches and knowledge fromdifferent disciplines able to detect changes in ecosystem func-tioning (Hagger et al., 2006; Solimini et al., 2006, 2009).

Recently, transcriptome profiling has benefited from techno-logical improvements and has become more efficient andeconomically affordable for a wide range of species, thus over-coming the initial concerns about its high costs (Van Aggelen et al.,2010). The employment of transcriptomic techniques, such as DNAmicroarrays, in aquatic toxicology already allowed the identifica-tion of early biomarkers of exposure and the definition of Mode OfAction (MOA) or gene patterns associated to chemicals or envi-ronmental stressors (e.g. Falciani et al., 2008; Leaver et al., 2010;

M. Milan et al. / Environmental Pollution 197 (2015) 90e98 91

Milan et al., 2013a; Pujolar et al., 2013; Schirmer et al., 2010).Therefore, international initiatives proposing to integrate eco-toxicogenomics in Environmental Risk Assessment (ERA) and inthe evaluation of marine ecosystem health have been recentlyincreased (Van Straalen and Feder, 2012; Van Aggelen et al., 2010).

Currently, the main challenges that must be addressed for asuccessful implementation of genomics in regulatory ecotoxicologyare i) the standardization of data collection and analysis (Bourlatet al., 2013); ii) the definition of transcriptional features/perturba-tions reflecting the response to specific environmental conditions,and the establishment of thresholds associated to “non-significant”and “adverse” responses (Blunt et al., 2007); iii) the generation ofwell-defined MOA for reference chemicals; iv) the identification oflinks between molecular and biochemical responses and adversealterations in survival, development, and reproduction (VanAggelen et al., 2010); v) increasing exchange of information be-tween scientists and end-users, such as government policy, non-government organizations and environmental regulators (Bluntet al., 2007).

Fig. 1. Map of the Venice lagoon indicating the Manila clam samp

The Venice lagoon constitutes an excellent case study for envi-ronmental risk assessment due to the severe impact of differentanthropogenic pressures, such as the industrial activities of PortoMarghera, maritime traffic, and agricultural pollution (Apitz et al.,2007; Matozzo et al., 2010; Micheletti et al., 2011). Regulatorybodies in the Venetian region carried out extensive studies to assessthe environmental status of the Venice lagoon. This area is also ofmajor importance for bivalves production, especially Manila clams(FAO data; Turolla, 2007). Currently, to ensure that seafood fromlagoon areas is absolutely safe for the consumer, one third of theVenice lagoon is classified as not suitable for shellfish harvesting(DGR 3366/2004). In addition, upper limits of whole-bodies dioxinsand dibenzofurans (PCDD/Fs) imposed for bivalves commerciali-zation (0.5 pg/g WHO-TEQ) are very stringent and far below thoseimposed by the European Commission (Regulation No 1259/2011)for fish and fishery products including molluscs species (3.5 pg/gWHO-TEQ).

Given the high economic importance of clam culture and har-vesting in the Venice lagoon (Solidoro et al., 2003), the restrictions

ling sites: Marghera (PM), Chioggia (CH), and Colmata (CO).

M. Milan et al. / Environmental Pollution 197 (2015) 90e9892

imposed by the regional Directive led to a fierce debate upon the re-classification of certain areas. More in general, a proper assessmentof the ecological status of the Venice lagoon would be necessary toguide its sustainable management. To this end, several studies werecarried out over the last 15 years, most of them based on the use oflimited sets of biochemical, cellular, and organism biomarkers(Matozzo et al., 2010; Moschino et al., 2010).

In the present study, the potential for genomic tools to beincorporated in a multi-disciplinary approach for environmentalmonitoring and risk assessment has been evaluated within thecomplex context of the Venice lagoon, through transcriptomicanalysis of Manila clams collected in areas subjected to differentanthropogenic impact.

2. Materials and methods

2.1. Samples collection

Approximately 150 adults Manila clams were collected inJanuary and February 2011 in each of three sampling sites in theVenice lagoon (Fig. 1): inner Porto Marghera (PM), Chioggia (CH),and Casse di Colmata (CO). Despite clam harvesting for humanconsumption is strictly forbidden in both CO and PM areas (DGR3366/2004), the different classification as “forbidden area 1” and“forbidden area 2” (see Fig. 1) refers to seed recruitment, that isallowed only in “forbidden area 2” (CO). Clams were sampledduring the resting stage (for Ruditapes philippinarum fromNovember to February as demonstrated by Meneghetti et al.(2004)), in order to avoid physiological changes due to differentreproductive stages. Water temperature and salinity weremeasured on-site with a field thermometer and a hand-heldrefractometer. After sampling, clams were transferred to the labo-ratory and, for each sample set the digestive gland was dissectedfrom 50 bivalves. To increase homogeneity across samples, Manilaclam of similar average size and weight have been considered forgene expression analyses. Clams collected in January were analysedby DNA microarray and results validated through RT-real time PCR.In addition, RT-real time PCR analyses of 6 target genes were per-formed in clams sampled in February (see “Real-time RT-PCR”section). Environmental parameters, geographical coordinates,number of individuals and experimental analyses performed foreach sampling site are summarized in Table S1.

2.2. Chemical analyses

Pooled soft tissues from approximately 100 individuals per sitewere subjected to MS analysis to measure the concentrations ofPCDDs/Fs (Method EPA 1613B, 1994), polychlorinated biphenylsand dioxin-like PCBs (PCBs and PCBs-DL; Method EPA 1668B, 2008),polybrominated biphenyls ethers (PBDEs; Method EPA 1614, 2007)and hexachlorobenzene (HCB; MPI003, Internal method). Thechemical analyses are described in detail in File S1.

2.3. R. philippinarum oligonucleotide microarray

Gene transcription analyses of 15 digestive glands pools (fivepools for each sampling sites composed by 10 digestive glands)were performed using a 8 � 15 K Agilent oligo-DNA microarrayplatform deposited in the GEO database (http://www.ncbi.nlm.nih.gov/geo/) under accession numbers GPL15723. For details aboutR. philippinarum DNA microarray platform used in the presentstudy see Milan et al. (2013a). In order to further increase thenumber of putatively annotated contigs represented onto theR. philippinarum DNA microarray platform, similarity searchesagainst the recently published (Zhang et al., 2012) Pacific oyster

(Crassostrea gigas) protein database (http://gigadb.org/dataset/100030) were carried out. Microarrays were synthesized in situusing the Agilent non-contact ink-jet technology with a 8 � 15 Kformat and including default positive and negative controls.

RNA extraction, samples labelling and hybridisation were per-formed according to the Agilent Gene Expression Analysis protocolsdescribed by Milan et al. (2011). Details are reported inSupplementary File S1.

2.4. Gene expression analyses

Two-class unpaired SAM tests (Tusher et al., 2001) were carriedout to identify differentially expressed genes among the threegroups of samples (CH, CO and PM) with the following parameters:False Discovery Rate (FDR) ¼ 5%; Fold Change (FC) > 1.5; number ofpermutations ¼ 200.

A SAM quantitative correlation analysis (FDR ¼ 1%) was per-formed to identify genes correlated either with the concentration ofthe different classes of pollutants (PCDDs/PCDFs, PCBs, PBDEs, HCB)or with environmental parameters (water temperature andsalinity).

A more systematic functional interpretation of differentiallyexpressed genes was obtained through an enrichment analysisperformed by DAVID software (Huang et al., 2009) as described inFile S1.

Sample class prediction was carried out using PAM (PredictionAnalysis for Microarrays), available online (http://www.stat.stanford.edu/~tibs/PAM). PAM analyses have been carried outconsidering three different gene sets (see Results section). PAM firstperforms a discriminant analysis on “known” samples (TrainingSample Set). The training set consisted of 10 samples collected inCH and PM representing two areas that are associated with a“good” and “bad” ecological status, respectively, in relationwith thecorresponding level of chemical contamination. The accuracy ofclass prediction on the Training Sample Set has been estimatedthrough cross-validation (10% of samples randomly extracted andclassified based on the discriminant function calculated on theremaining cases). Then the same set of probes has been used toclassify CO samples (Test Sample Set).

2.5. Real-time RT-PCR

Real-time RT-PCR was conducted in single digestive glands ofclams collected in January (n ¼ 60) and February (n ¼ 26) in thethree sampling sites (Table S1). A total of 6 target genes (File S1)were selected for real-time RT-PCR analysis. Due to technicalproblems, it has been possible to perform qRT-PCR on 6 of 7 tran-scripts able to discriminate PM samples in the previous study(Milan et al., 2013b; sulfotransferase phenol-member 1 (SULT1),glutathione S-transferase theta-1 (GST-q1), acetylcholinesterase(AChE) and three not annotated transcripts). The details of the real-time qPCR experiments (e.g. run settings, primer pair efficiency) aredescribed in File S1.

Gene expression values of 6 transcripts obtained by qRT-PCRanalyses have been implemented in PAM for the classification ofthe three investigated areas as done for DNA microarray data (seeabove). The training set consisted of 57 individuals collected in CHand PM. The accuracy of class prediction on the Training Sample Sethas been estimated through cross-validation. Then the same tran-scripts were used to classify 29 CO individuals (Test Sample Set)that were not included in the training set.

Spearman's correlation test was performed to measure the de-gree of association between real-time and microarray data. Thisstatistical test was carried out using SPSS ver. 12.0.

Table 1Organic pollutant concentrations measured in R. philippinarum from soft tissue pools.

PM CO CH

January February January February January February

SPCDDs/PCDFs (pg-WHO-TE/g) 0.6 0.83 0.1 0.092 <d.l. m.d.SPCB-DL (pg-WHO-TE/g) 0.032 0.27 0.0163 0.016 0.012 m.d.SPCDDs/PCDFs þ SPCBs-DL 0.632 1.52 0.116 0.108 0.012 m.d.SPCBs (pg/g) 2873 6179 1429 1327 838 m.d.SPBDEs (pg/Kg) 80 110 24 11 21 m.d.HCB (mg/Kg) 0.484 0.759 0.127 0.137 0.032 m.d.

The concentrations of dioxins and furans (PCDDs/PCDFs) and dioxin-like PCB (PCB-DL) congeners are reported as ‘lower bound’ (seeMethods). PCB: polychlorinated biphenyls;PBDE: polybrominated biphenyls ethers; HCB: hexachlorobenzene; m.d. ¼ missing data; d.l. ¼ detection limit.

M. Milan et al. / Environmental Pollution 197 (2015) 90e98 93

3. Results

3.1. Chemistry

The concentrations of organic pollutants measured in clamwhole bodies are reported in Table 1. PCDDs/PCDFs and PCBs-DLconcentrations detected in January were five times higher in Mar-ghera compared to Colmata, while at Chioggia, the PCDDs/PCDFsconcentration was under the detection limit. Chemical analysis ofMarghera clams sampled in February showed increased PCDDs/Fsand PCBs concentrations compared to clams sampled in January,while in Colmata there were no significant temporal changes.

Only Marghera clams showed PCDDs/PCDFs concentrationshigher than the limit of 0.5 pg/gWHO-TEQ imposed by the regionaldirective (DGR 3366/2004). However, the limits imposed by theEuropean Regulation (Commission Regulation 1259/2011) for themarketing of fish products were never reached. The clam harvestban for human consumption in the Venice lagoon includes also theColmata area, where PCDDs/Fs concentration was significantlylower than the regional limit. Finally, HCBs and PBDEs concentra-tions were four times higher in Marghera bivalves compared toColmata, while Chioggia clams showed the lowest chemical con-centrations for all analysed pollutant classes.

3.2. Transcriptional changes between Manila clams collected indifferent Venice lagoon areas

Raw and normalised fluorescence data from all microarray ex-periments have been deposited in the GEO database under acces-sion number GSE50062.

Pairwise comparisons among the three groups (PM, CH and CO)were performed through a two-class unpaired SAM analysis (Foldchange > 1.5; FDR ¼ 5%). The highest number of differentiallyexpressed genes (DEGs) was found comparing Chioggia (CH) andMarghera (PM) clams (Table S2), with a total of 580 transcripts (377over- and 203 under-expressed in PM). The comparison betweenColmata (CO) and PM revealed a total of 234 DEGs, while fewerdifferences were found between CH and CO samples (45 DEGs).Lists of the significant transcripts identified by each SAM analysisare reported in Table S3.

A ‘core’ of 142 transcripts (Table S3) was found significantly up-or down-regulated in both comparisons with Marghera clams (CHvs PM and CO vs PM), which represents a highly significant overlap(Fisher's exact test p < 0.001). Among these, several DEGs up-regulated in PM play a role in drug metabolism and detoxificationprocesses, such as four glutathione S-transferase transcripts, fivesulfotransferases, and three different cytochrome p450. Up-regulation of superoxide dismutase (SOD), which is involved inscavenging reactive oxygen species, suggests increased levels ofoxidative stress in PM clams. Moreover, PM samples showed up-regulation of two acetylcholinesterase-coding transcripts

compared to CH and CO clams.Among the six DEGs found in common when comparing Col-

mata either with PM or CH, temptin and E74-like factor 5 were up-regulated in CO clams. Temptin has been characterized in Aplysia asa highly expressed peptide hormone in the albumen gland, releasedwith eggs into the environment (Cummins et al., 2006). E74-likefactor 5 is a transcriptionally activator that regulates a number ofepithelium-specific genes found in tissues containing glandularepithelium.

DEGs obtained in each SAM pairwise comparison were used todefine a gene list for functional enrichment analysis performed inDAVID (http://david.abcc.ncifcrf.gov/) considering GO term classi-fication, and KEGG (Kyoto Encyclopedia of Genes and Genomes)pathways (Table S4). SAM comparison between CH and PMrevealed two significantly enriched KEGG-pathways, ‘drug meta-bolism’ and ‘metabolism of xenobiotics by cytochrome P450’, andthe enriched MF terms ‘sulfotransferase activity’, highlightingenhanced detoxification processes in samples collected in PortoMarghera. The most significantly enriched KEGG-pathway term inthe comparison between CH and PM was ‘aminoacyl-tRNAbiosynthesis’ (fold enrichment ¼ 7.2), represented by fouraminoacyl-tRNA synthetases up-regulated in clams collected inMarghera. Among the enriched KEGG pathways, ‘focal adhesion’and ‘ECM-receptor interaction’ were also significant, representedrespectively by 9 and 4 DEGs.

The comparison between Colmata and Porto Marghera pooledsamples showed 27 GO_BP, 3 GO_CC, 2 GO_MF and 3 KEGG-pathways significantly over-represented (Table S4). Despite bothareas are classified as forbidden sites, the inclusion of transcriptsup-regulated in PM clams in the enriched terms ‘sulfotransferaseactivity’ and ‘xenobiotic metabolic process’, likely reflects thehigher xenobiotic exposures in clams collected close to the indus-trial area of Marghera. Both PM pairwise comparisons (CH vs PMand CO vs PM) also showed over-representation of genes involvedin ‘PPAR (peroxisome proliferator-activated receptors) signallingpathway’ and ‘pyruvate metabolism’. No enriched term was foundin the comparison between CO and CH clams.

3.3. Transcriptional changes associated to environmental variablesand chemical exposures

Environmental factors (e.g. temperature, salinity, chlorophyllcontent) may induce physiological differences in clams collected indifferent areas (Milan et al., 2013a,b). To understand whether DEGsobserved between different Venice lagoon areas were related toxenobiotic exposure or to environmentally-related physiologicalchanges, the entire gene expression dataset was correlated to in-dividual pollutant concentrations (PCDDs/PCDFs/PCBs-DL, PCBs,PBDEs, HCB) and to two environmental parameters (temperatureand salinity). A total of 307 probes were found directly or inverselycorrelated to pollutants classes (Table S5), while a total of 58 and 32

Fig. 2. Venn diagrams and charts reporting for each SAM pairwise comparison among sampling areas (CHvsPM; COvsPM; CHvsCO) the number and percentage of DEGs correlatedto chemical concentrations (at least with one chemicals class), and/or with environmental variables (salinity and temperature).

M. Milan et al. / Environmental Pollution 197 (2015) 90e9894

probes were found significantly correlated to salinity and watertemperature, respectively (Table S5). Functional analysis of tran-scripts significantly correlated to chemicals revealed several GOterms and KEGG pathways that were found to be enriched also inPM pairwise comparisons, such as ‘PPAR signalling pathway’,‘aminoacyl-tRNA biosynthesis’, ‘fatty acid metabolism’, ‘sulfo-transferase activity’, ‘xenobiotic metabolic process’ and ‘gluta-thione metabolism’ (Table S6).

Venn diagrams (see Fig. 2) were then constructed considering i)DEGs in each SAM pairwise comparison (Table S3); ii) genescorrelated to chemicals concentrations (Table S5); iii) genescorrelated to salinity and temperature (Table S5). Comparisons ofMarghera with Chioggia and Colmata showed high percentages ofDEGs that correlated also with pollutant concentrations (40% and55%, respectively; Fig. 2). Conversely, only 6% of DEGs in both PMcomparisons correlated with environmental parameters.

A completely different scenario was observed in the comparisonbetween CH and CO, with 44% of DEGs that correlated also withtemperature or salinity, and only two transcripts correlated to

chemical concentrations. These results suggest that CO clams werenot affected by significant pollutants exposure and that differencesrevealed in the comparison with CH clams might be mainly relatedto environmental variables.

3.4. Classification of Venice lagoon areas

The entire set of probes (307) that correlate with chemicalconcentrations was employed for the classification of the threeinvestigated areas using PAM (Prediction Analyses for Microarray),a software for class prediction with multiple gene expression data.As reported in Table 2, PAM was able to define a discriminantfunction that achieved 100% accuracy in cross-validation. Whenapplied to five Marghera samples and five Chioggia ones (theTraining Set), all were correctly assigned to the site of origin, whichrepresent a “polluted” (PM) and a “clean” area (CH), respectively.Colmata samples (Test Sample Set) were then classified through thesame set of probes (classifier) and the discriminant function ob-tained using the Training set (CH and PM). In agreement with data

Table 2Number of pools/individuals assigned to “bad” or “good” ecological status by PAM(Prediction Analysis for Microarrays).

Marghera Chioggia Colmata

Microarray data (pools) N� ofgenes

Bad Good Bad Good Bad Good

Genes correlated with pollutantsconcentration

307 5 0 0 5 0 5

Genes confirmed correlated withpollutant concentrations in thepresent and in the previous study(Milan et al., 2013a)

37 5 0 0 5 0 5

Genes discriminating Porto Margherasamples

7 5 0 0 5 0 5

qRT-PCR data Bad Good Bad Good Bad Good

Genes discriminating Porto Margherasamples (single individuals)

6 23 5 1 28 4 25

For each areas has been reported the number of pools/individuals assigned to “bad”or “good” ecological status, represented by PM and CH respectively. PAM analyseshave been performed considering DNAmicroarray and qRT-PCR data of different setof genes.

M. Milan et al. / Environmental Pollution 197 (2015) 90e98 95

from chemical analysis, all Colmata pooled samples were assignedto the “CH” class (Table 2).

In a previous study, gene expression analyses of Manila clamcollected during the year 2009 in different seasons at four locationsincluding Chioggia andMarghera, revealed a total of 162 transcriptscorrelated with pollutant concentrations (Milan et al., 2013b).Comparing the results of SAM correlation analysis of the presentwork (307 significant probes) with the 162 transcripts identified inthe previous study, a total of 37 probes were found commonlycorrelated with at least one chemical class with a consistentregulation through time (i.e. always down- or up-regulated in bothstudies). Considering that samples have been collected in differentyears and seasons (2009 and 2011), those 37 probes may beconsidered a temporally stable set of genes reflecting conditions ofchemical stress independently from sampling season. Geneexpression profiles of this set of transcripts have thus been used forthe classification of the three investigated areas. Although thenumber of probes was smaller, the discrimination of PM and CHachieved 100% accuracy after cross-validation, as already observedfor the larger set of gene classifiers. Likewise, using such a reducedset of probes, Colmata samples were assigned to the “CH group”with 100% probability.

In the previous study, it was also identified a minimum set ofseven genes able to discriminate samples collected in the restrictedarea of PM from clams collected in farming areas of the Venicelagoon (Milan et al., 2013b). This set of genes was composed bysulfotransferase phenol-member 1 (SULT1), glutathione S-trans-ferase theta-1 (GST-q1), acetylcholinesterase (AChE) and four notannotated transcripts (probes 2_c2341, 2_c1189, c20179, c14595).In the present study, these seven transcripts were found to bepositively correlated with the concentration of at least one chem-ical class (Table S5). When using gene expression values for this setof seven genes, a correct assignment of Marghera and Chioggiasamples to the corresponding site of origin was obtained with anaccuracy of 100%, while Colmata was assigned to the “CH” classwith probability close to 100% (Table 2).

3.5. Real-time and low-cost: a prototypical assay to assess chemicalexposures

Six of the seven transcripts able to discriminate clams collectedin Marghera, were then analysed in real-time RT-PCR, applied todigestive gland samples collected from single individuals in January

(n¼ 60) and February 2011 (n¼ 26) (Table S1). The purpose of real-time RT-PCR analysis was twofold. First, an independent validationof microarray data is generally required when using a novel arrayplatform. To this end, a significant positive correlation in a Spear-man's correlation test between real-time and microarray data (p-value < 0.01; Spearman's rho correlation coefficients ranging from0.664 to 0.789) has been obtained, confirming that the novelmicroarray format (15 K probes) provides reliable expression data.

The second goal was to develop a prototype for a rapid and cost-effective assay to detect changes in expression of gene markersrevealing chemical risks. In fact, a quick assay, cheaper thanmicroarray analysis, might be more easily employed in routineenvironmental monitoring. Such a prototype assay was also vali-dated on additional samples, collected independently from thoseanalysed with microarrays, in the same sites, but one month afterthe first sampling (February 2011). Moreover, to evaluate theimpact of individual variability on gene expression variations, real-time RT-PCR analysis was carried out on single samples, instead ofpooled ones.

Accordingly, PAM-based classification was carried out on a totalof 86 samples, 60 of them already analysed by DNA microarraysafter being pooled, while 26 individuals represented a second, in-dependent sampling (February 2011). A discriminant function wasestimated in PAM using RT-PCR data from the 20 individuals fromPM and the 20 ones from CH (Training Set), defining two classes ofsamples, associated respectively with high- and low-level ofchemical contamination. The obtained discriminant function wasthen used to classify all individuals from Colmata (29 individuals),as well as samples from PM and CH (9 individuals) collected onemonth after the first sampling. These results confirmed that clas-sification accuracy was extremely good for PM and CH, even onindependently collected samples and using a set of six transcripts.The same analysis assigned most CO samples to the class of low-level chemical contamination (25/29). The use of individual geneexpression data, however, revealed some degree of inter-individualvariability, which might be due to divergent exposure and/orresponse to pollutants, but also to biotic and abiotic variables un-related to xenobiotics. Pooling samples might therefore be themostadvisable strategy for reliable routine testing.

4. Discussion

4.1. Chemical analysis and classification of Venice lagoon areas

Chemical analysis confirmed the highest concentrations ofpollutants in clams collected at Porto Marghera (PM). PM remains acritical area, although chemical contamination appears to bereduced if compared to previous studies (Matozzo et al., 2010;Milan et al., 2013b). Such reduction is likely correlated with thesharp decrease of industrial activities in PM, together with morestringent regulations and controls on production standards. Theeffects on aquatic animals of such changes in production and reg-ulations should be constantly monitored in the future to confirmthe positive trend observed in the present study.

For clams sampled in the restricted area of Colmata, reportedpollutant concentrations were similar to those measured inChioggia, and significantly lower than those found in PM clams. Inaddition, PCDDs/Fs concentration was five times lower than limitsimposed by local regulations. Although additional data arerequired, the chemical evidences obtained in the present studysuggest that the classification of Colmata as a restricted area shouldprobably be re-evaluated.

M. Milan et al. / Environmental Pollution 197 (2015) 90e9896

4.2. Identification of molecular signature reflecting a “badecological status”

To assess the ecological status of marine environments, newinternational guidelines require the development and integrationof approaches and knowledge from different disciplines able toreflect changes in ecosystem functioning due to human-relatedactivities (e.g. WFD, 2000/60/EC). To date, it has been widelydemonstrated that high-throughput transcriptomic analysis canrepresent a new tool that might provide detailed molecular infor-mation on the interactions between living organisms and theenvironment where they live (e.g. Chapman et al., 2009; Leaveret al., 2010; Pujolar et al., 2012).

In agreement with data from chemical analyses, the major dif-ferences in gene expression profiles were observed in clamscollected in PM, with a high percentage of DEGs that also correlatedto xenobiotics concentrations (Fig. 2). Further evidence comes fromfunctional analysis of up-regulated DEGs in PM samples, whichshowed significant enrichment of genes involved in detoxificationand drug metabolism (Tables S3 and S4), and from the observedover-expression of Cu/Zn-SOD, a result that likely reflects increasedlevels of oxidative stress induced by exposure to chemical con-taminants (Bou�etard et al., 2012; Valavanidis et al., 2006). Inaddition, two transcripts coding for putative AChE, a serine prote-ase that hydrolyses the neurotransmitter acetylcholine, were up-regulated in PM. AChE activity is commonly used as a biomarkerof exposure in bivalve species (Binelli et al., 2006; Danellakis et al.,2011; Matozzo et al., 2005). It has been widely demonstrated thatAChE is inhibited by several pesticides used for agricultural activ-ities (Führer et al., 2012; Greco et al., 2011; Mora et al., 1999).Matozzo et al. (2005) reported a significant reduction in AChE ac-tivity in Manila clams sampled inMarghera, suggesting exposure toneurotoxic compounds released from the nearby industrial area.Our analysis showed over-expression of AChEmRNA in PM samplesas already observed in previous studies (Milan et al., 2011, 2013b;Somnuek et al., 2009). Such apparent contrast between AChEenzyme activity and its expression at the mRNA level might bereconciled under the hypothesis that increased AChE transcriptionis a compensatory modification to counteract the inhibition ofenzyme activity due to xenobiotic exposure.

Moreover, functional analysis seems to suggest an increasedprotein turnover in the hepatopancreas of PM animals, as shown bythe up-regulation of several ribosomal proteins and the significantenrichment of the KEGG term “aminoacyl-tRNA biosynthesis” (withfour aminoacyl-tRNA synthetases up-regulated in clams collectedin Marghera; Table S4). Aminoacyl-tRNA synthetases catalyse theesterification of amino acids to their cognate tRNAs, playing a keyrole in protein synthesis (Cusak, 1997). Concurrently, up-regulationof several lysosomal proteolytic enzymes (i.e. 13 transcripts codingfor cathepsin B, cathepsin L and cathepsin K) suggests increasedprotein degradation.

Similar evidence comes from correlation analyses of geneexpression profiles with chemical concentrations (Tables S4 andS6), revealing several genes involved in detoxification processes(e.g. glutathione metabolism, xenobiotic metabolic process, sulfo-transferase activity), protein turnover (e.g. aminoacyl-tRNAbiosynthesis, protein dimerization activity), and energy meta-bolism (e.g. fatty acid metabolism, PPAR signalling pathway). Theseresults confirm the evidence observed in Crassostrea virginica byChapman et al. (2009), where shifts in gene expression of proteinsinvolved in energy production, protein synthesis and basic meta-bolism were reported in response to land-use practices.

4.3. Transcriptomics in water bodies classification

The use of transcriptomic data in environmental monitoring canbenefit from the broad experience accumulated in cancer geno-mics, where a multitude of methodological approaches have beendeveloped to allow for data meta-analysis and to provide accuratesample classification (Phan et al., 2012). The comparison of genesets across studies and the use of classification tools for expressiondata (PAM) presented here are just a preliminary exampledemonstrating how methodologies derived from other disciplinesmight be effectively employed and adapted to the specific needs ofenvironmental genomics.

Nonetheless, the ability of classifying an area that is not reliablyclassified yet such as Colmata, already provides clear evidence onthe potential for such an approach. Samples from two areas that arewell characterized in terms of chemical risk were used to train aclassification tool, which was based on a set of genes that areconsistent across experiments. The trained classifier, when appliedto Colmata samples, assigned all of them to CH group regardless ofwhich gene expression dataset was employed, i.e. suggested a lowrisk of exposure to chemical pollutants. In agreement with chem-ical analyses, these results indicate that, at variance with whathighlighted in PM animals, there are no detectable biological effectsin CO clams referable to chemical exposures. Although further in-vestigations are required, this result suggests that a “positive” re-classification of Colmata might be taken into account, despite thegeographical proximity to Porto Marghera.

It should be noted, however, that transcriptomic data are ex-pected to provide an even greater contribution if calibrated withinan expert system together with other types of analysis (lines ofevidence, LOEs), such as sediment chemistry, bioavailability,cellular biomarkers and ecotoxicological bioassays. Within a modelable to integrate various LOEs giving them different weights duringenvironmental risk assessment, (as recently proposed by Piva et al.(2011); Benedetti et al. (2012); Regoli et al. (2014)), genomic ana-lyses of sentinel species could add significant biological value tochemical analyses, supporting site-orientedmanagement decisionsin chronically polluted areas.

4.4. Hurdles, achievements, and future perspectives for genomictools in the environmental risk assessment

Overall, the present study serves as a proof of concept for the useof genomic tools in the monitoring of water bodies, suggesting thatgenomic analysis on sentinel species targeted to molecular path-ways disrupted by chemical exposures or associated to bad qualityhabitat provides reliable and quantitative data, which may beplaced in a context of integrated environmental risk assessment.

Transcriptomics has a great potential for innovation in envi-ronmental assessment and it could be employed in extensivemonitoring programmes of areas not yet characterized in terms ofecological quality, through the comparison of unknown sites withareas of known environmental risk. However, to provide a usefulcontribution to environmental risk assessment, transcriptomictools should have specific characteristics, which have beenaddressed here through the case study of the Venice lagoon.

First, the observed transcriptomic response to chemical expo-sure should be reproducible across sampling areas and temporalreplicates. By comparing previously obtained results (Milan et al.,2013b), it was demonstrated that a core response was conservedboth across and within different studies, with a set of responsivegenes that significantly correlated with concentrations of chemicalspecies. Functional annotation of significant genes also appeared tobe biologically meaningful, suggesting that the link betweenchemical pollution and transcriptome response might not be the

M. Milan et al. / Environmental Pollution 197 (2015) 90e98 97

result of random association.Secondly, transcriptomic markers should mainly reflect the

specific environmental risk under scrutiny, without being exces-sively affected by other variables. The set of genes identified in thepresent study seems to conform to such requirement, as it allowedprecise classification of samples in agreement with the potentialrisk of chemical exposure measured by chemical analysis, despiteexisting differences in water temperature and salinity across sites.

Third, it should be possible to use transcriptomic data as apractical tool for routine monitoring. In this respect, gene expres-sion analysis should be cost- and time-effective. The possibility toimplement qRT-PCR assays for a limited, yet meaningful set ofmarkers is extremely interesting, for real-time monitoring as wellas for increasing the frequency of controls.

Last, but not less important, transcriptomic analysis is able togenerate quantitative data, which are relatively objective andcomparable across studies and laboratories, since internal stan-dards and controls are included. This is much important in order toensure that environmental monitoring could assess temporaltrends, using quantitative thresholds on the determination of goodenvironmental status and in environmental risk assessment.

While technical hurdles (technology cost, limited genomic in-formation for non-model species) are being progressively over-come, lack of communication between researchers and end-usersremains a major problem, partially due to limited opportunities tointeract, but also because of a gap in technical knowledge (Bluntet al., 2007). Accordingly, effective communication and collabora-tion for the establishment of a work-flow able to interpret andintegrate toxicogenomics in quantitative risk assessment, will becritical in giving it a larger role in the regulatory decision-making.

Author information

Notes

The authors declare no competing financial interest.

Author contributions

LBA, MM and TP conceived and designed the project. MM, MPand LP performed Manila clam sampling. MM and MP performedmicroarray experiments. MP performed qRT-PCR analyses. CC, FSand GF performed chemical analyses. MM executed all statisticalanalyses and functional annotation. LBA, MM and LBO wrote themanuscript. All listed authors edited the manuscript. All authorshave given approval to the final version of the manuscript.

Data accessibility

Gene expression analyses were performed using a 8 � 15 KAgilent oligo-DNA microarray platform deposited in the GEOdatabase under accession number GPL15723 (http://www.ncbi.nlm.nih.gov/geo/). Microarray raw and normalised fluorescencevalues were deposited in the GEO database (http://www.ncbi.nlm.nih.gov/geo) under accession number GSE50062.

Acknowledgements

This studywas supported by funding fromMagistrato alle Acquedi Venezia and fromUniversity of Padova, Progetto Giovani Studiosie Bando 2013 e Scienze Agrarie e Veterinarie (Prot. GRIC13XVFI).

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://

dx.doi.org/10.1016/j.envpol.2014.12.005.

References

Apitz, S.E., Barbanti, A., Bocci, M., Carlin, A., Montobbio, L., Berstein, A.G., 2007.Sediments of the Venice lagoon evaluated in a screening risk assessmentapproach to inform risk management strategies: part Idapplication of inter-national sediment quality guidelines (SQGs). Integr. Environ. Assess. Manag. 3,393e414.

Benedetti, M., Ciaprini, F., Piva, F., Onorati, F., Fattorini, D., Notti, A., Ausili, A.,Regoli, F., 2012. A multidisciplinary weight of evidence approach for classifyingpolluted sediments: integrating sediment chemistry, bioavailability, biomarkersresponses and bioassays. Environ. Int. 38, 17e28. http://dx.doi.org/10.1016/j.envint.2011.08.003.

Binelli, A., Ricciardi, F., Riva, C., Provini, A., 2006. Integrated use of biomarkers andbioaccumulation data in Zebra mussel (Dreissena polymorpha) for site-specificquality assessment. Biomarkers 11, 428e448. http://dx.doi.org/10.1080/13547500600733788.

Blunt, R.E., Walsh, K.A., Ashton, D.K., Viant, M.R., Chipman, J.K., 2007. Knowledgetransfer initiative between molecular biologists and environmental researchersand regulators. Environ. Sci. Pollut. Res. Int. 14, 293e296. http://dx.doi.org/10.1065/espr2007.06.431.

Bou�etard, A., Besnard, A.L., Vassaux, D., Lagadic, L., Coutellec, M.A., 2012. Impact ofthe redox-cycling herbicide diquat on transcript expression and antioxidantenzymatic activities of the freshwater snail Lymnaea stagnalis. Aquat. Toxicol.126, 256e265. http://dx.doi.org/10.1016/j.aquatox.2012.11.013.

Bourlat, S.J., Borja, A., Gilbert, J., Taylor, M.I., Davies, N., Weisberg, S.B., Griffith, J.F.,Lettieri, T., Field, D., Benzie, J., Gl€ockner, F.O., Rodríguez-Ezpeleta, N., Faith, D.P.,Bean, T.P., Obst, M., 2013. Genomics in marine monitoring: new opportunitiesfor assessing marine health status. Mar. Pollut. Bull. http://dx.doi.org/10.1016/j.marpolbul.2013.05.042.

Chapman, R.W., Mancia, A., Beal, M., Veloso, A., Rathburn, C., Blair, A., Sanger, D.,Holland, A.F., Warr, G.W., Didonato, G., 2009. A transcriptomic analysis of land-use impacts on the oyster, Crassostrea virginica, in the South Atlantic bight. Mol.Ecol. 18, 2415e2425. http://dx.doi.org/10.1111/j.1365-294X.2009.04194.x.

Cummins, S.F., Nichols, A.E., Schein, C.H., Nagle, G.T., 2006. Newly identified water-borne protein pheromones interact with attractin to stimulate mate attraction Iaplysia. Peptides 27, 597e606. http://dx.doi.org/10.1016/j.peptides.2005.08.026.

Cusak, S., 1997. Aminoacyl-tRNA synthetases. Curr. Opin. Struct. Biol. 7, 881e889.http://dx.doi.org/10.1016/S0959-440X(97)80161-3.

Danellakis, D., Ntaikou, I., Kornaros, M., Dailianis, S., 2011. Olive oil mill wastewatertoxicity in the marine environment: alterations of stress indices in tissues ofmussel Mytilus galloprovincialis. Aquat. Toxicol. 101, 358e366. http://dx.doi.org/10.1016/j.aquatox.2010.11.015, 25.

Denslow, N.D., Garcia-Reyero, N., Barber, D.S., 2007. Fish ‘n’ chips: the use ofmicroarrays for aquatic toxicology. Mol. BioSyst. 3, 172e177. http://dx.doi.org/10.1039/b612802p.

DGR 3366/2004. Delibera della Giunta n� 3366 del 29 ottobre 2004. Regione delVeneto.

Directive 2008/56/EC. Marine Strategy Framework Directive. Directive 2008/56/ecof the European Parliament and of the Council of 17 June 2008 Establishing aFramework for Community Action in the Field of Marine Environmental Policy(Marine Strategy Framework Directive). http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri¼OJ:L:2008:164:0019:0040:EN:PDF.

Falciani, F., Diab, A.M., Sabine, V., et al., 2008. Hepatic transcriptomic profiles ofEuropean flounder (Platichthys flesus) from field sites and computational ap-proaches to predict site from stress gene responses following exposure tomodel toxicants. Aquat. Toxicol. 90, 92e101. http://dx.doi.org/10.1016/j.aquatox.2008.07.020.

FAO Data. FAO. http://www.fao.org/fishery/culturedspecies/Ruditapes_philippinarum/en.

Führer, E., Rudolph, A., Espinoza, C., Díaz, R., Gajardo, M., Cama~no, N., 2012. Inte-grated use of biomarkers (O:N ratio and acetylcholinesterase inhibition) onAulacomya ater (Molina, 1782) (Bivalvia: Mytilidae) as a criteria for effects oforganophosphate pesticide exposition. J. Toxicol. 951568. http://dx.doi.org/10.1155/2012/951568.

Greco, L., Pellerin, J., Capri, E., Garnerot, F., Louis, S., Fournier, M., Sacchi, A., Fusi, M.,Lapointe, D., Couture, P., 2011. Physiological effects of temperature and a her-bicide mixture on the soft-shell clamMya arenaria (Mollusca, Bivalvia). Environ.Toxicol. Chem. 30, 132e141. http://dx.doi.org/10.1002/etc.359.

Hagger, J.A., Jones, M.B., Leonard, D.R., Owen, R., Galloway, T.S., 2006. Biomarkersand integrated environmental risk assessment: are there more questions thananswers? Integr. Environ. Assess. Manag. 2, 312e329. http://dx.doi.org/10.1002/ieam.5630020403.

Huang, D.W., Sherman, B.T., Lempicki, R.A., 2009. Systematic and integrative anal-ysis of large gene lists using DAVID bioinformatics resources. Nat. Prot. 4,44e57. http://dx.doi.org/10.1038/nprot.2008.211.

Leaver, M.J., Diab, A., Boukouvala, E., Williams, T.D., et al., 2010. Hepatic geneexpression in flounder chronically exposed to multiply polluted estuarinesediment: absence of classical exposure 'biomarker' signals and induction ofinflammatory, innate immune and apoptotic pathways. Aquat. Toxicol. 96,234e245. http://dx.doi.org/10.1016/j.aquatox.2009.10.025.

Matozzo, V., Tomei, A., Marin, M.G., 2005. Acetylcholinesterase as a biomarker ofexposure to neurotoxic compounds in the clam Tapes philippinarum from the

M. Milan et al. / Environmental Pollution 197 (2015) 90e9898

lagoon of Venice. Mar. Pollut. Bull. 50, 1686e1693. http://dx.doi.org/10.1016/j.marpolbul.2005.07.011.

Matozzo, V., Binelli, A., Parolini, M., Locatello, L., Marin, M.G., 2010. Biomarker re-sponses and contamination levels in the clam Ruditapes philippinarum forbiomonitoring the lagoon of Venice (Italy). J. Environ. Monit. 12, 776e786.http://dx.doi.org/10.1039/b920536e.

Meneghetti, F., Moschino, V., Da Ros, L.D., 2004. Gametogenic cycle and variations inoocyte size of Tapes philippinarum from the lagoon of Venice. Aquaculture 240,473e488. http://dx.doi.org/10.1016/j.aquaculture.2004.04.011.

Micheletti, S., Gottardo, A., Critto, S., Chiarato, A., Marcomini, A., 2011. Environ-mental quality of transitional waters: the lagoon of Venice case study. Environ.Int. 37, 31e41. http://dx.doi.org/10.1016/j.envint.2010.06.009.

Milan, M., Ferraresso, S., Ciofi, C., Chelazzi, G., Carrer, C., Ferrari, G., Pavan, L.,Patarnello, T., Bargelloni, L., 2013a. Exploring the effects of seasonality andchemical pollution on the hepatopancreas transcriptome of the Manila clam.Mol. Ecol. 22, 2157e2172.

Milan, M., Pauletto, M., Patarnello, T., Bargelloni, L., Marin, M.G., Matozzo, V., 2013b.Gene transcription and biomarker responses in the clam Ruditapes philip-pinarum after exposure to ibuprofen. Aquat. Toxicol. 126, 17e29. http://dx.doi.org/10.1016/j.aquatox.2012.10.007.

Milan, M., Coppe, A., Reinhardt, R., Cancela, L., Leite, R.B., Saavedra, C., Ciofi, C.,Chelazzi, G., Patarnello, T., Bortoluzzi, S., Bargelloni, L., 2011. Transcriptomesequencing and microarray development for the Manila clam, Ruditapes phil-ippinarum: genomic tools for environmental monitoring. BMC Genomics 12,234. http://dx.doi.org/10.1186/1471-2164-12-234.

Mora, P., Michel, X., Narbonne, J.F., 1999. Cholinesterase activity as potentialbiomarker in two bivalves. Environ. Toxicol. Pharmacol. 7, 253e260. http://dx.doi.org/10.1016/S1382-6689(99)00019-8.

Moschino, V., Delaney, E., Meneghetti, F., Ros, L.D., 2010. Biomonitoring approachwith mussel Mytilus galloprovincialis (Lmk) and clam Ruditapes philippinarum(Adams and Reeve, 1850) in the lagoon of Venice. Environ. Monit. Assess. 177,649e663. http://dx.doi.org/10.1007/s10661-010-1663-5.

Phan, J.H., Young, A.N., Wang, M.D., 2012. Robust microarray meta-analysis iden-tifies differentially expressed genes for clinical prediction. Sci. World J. 2012,989637. http://dx.doi.org/10.1100/2012/989637.

Piva, F., Ciaprini, F., Onorati, F., Benedetti, M., Fattorini, D., Ausili, A., Regoli, F., 2011.Assessing sediment hazard through a weight of evidence approach with bio-indicator organisms: a practical model to elaborate data from sedimentchemistry, bioavailability, biomarkers and ecotoxicological bioassays. Chemo-sphere 83, 475e485. http://dx.doi.org/10.1016/j.chemosphere.2010.12.064.

Pujolar, J.M., Marino, I.A., Milan, M., et al., 2012. Surviving in a toxic world: tran-scriptomics and gene expression profiling in response to environmentalpollution in the critically endangered European eel. BMC Genomics 13, 507.http://dx.doi.org/10.1186/1471-2164-13-507.

Pujolar, J.M., Milan, M., Marino, I.A., Capoccioni, F., Ciccotti, E., Belpaire, C., Covaci, A.,Malarvannan, G., Patarnello, T., Bargelloni, L., Zane, L., Maes, G.E., 2013, May 15.Detecting genome-wide gene transcription profiles associated with highpollution burden in the critically endangered European eel. Aquat. Toxicol.132e133, 157e164. http://dx.doi.org/10.1016/j.aquatox.2013.02.012.

Regoli, F., Pellegrini, D., Cicero, A.M., et al., 2014. A multidisciplinary weight of

evidence approach for environmental risk assessment at the Costa Concordiawreck: integrative indices from Mussel Watch. Mar. Environ. Res. 96, 92e104.http://dx.doi.org/10.1016/j.envint.2011.08.003.

Regulation No 1259/2011. Commission Regulation (EU) No 1259/2011 of 2December 2011 Amending Regulation (EC) No 1881/2006 as Regards MaximumLevels for Dioxins, Dioxin-like PCBs and Non Dioxin-like PCBs in Foodstuffs.Regulation No 1259/2011. https://www.fsai.ie/uploadedFiles/Reg1259_2011.pdf.

Schirmer, K., Fischer, B.B., Madureira, D.J., Pillai, S., 2010. Transcriptomics in eco-toxicology. Anal. Bioanal. Chem. 397, 917e923. http://dx.doi.org/10.1007/s00216-010-3662-3.

Solidoro, C., Canu, D.M., Rossi, R., 2003. Ecological and economic considerations onfishing and rearing of Tapes phillipinarum in the lagoon of Venice. Ecol. Model.170, 303e318. http://dx.doi.org/10.1016/S0304-3800(03)00235-7.

Solimini, A., Cardoso, A., Heiskanen, A., 2006. Indicators and Methods for theEcological Status Assessment under the Water Framework Directive. Linkagebetween Chemical and Biological Quality of Surface Waters. Office for OfficialPublication of the European Communities, Luxembourg.

Solimini, A.G., Ptacnik, R., Cardoso, A.C., 2009. Towards holistic assessment of thefunctioning of ecosystems under the Water Framework Directive. Trac-Trend.Anal. Chem. 28 (2) http://dx.doi.org/10.1016/j.trac.2008.10.015.

Somnuek, C., Boonphakdee, C., Cheevaporn, V., Tanaka, K., 2009. Gene expression ofacetylcholinesterase in hybrid catfish (Clarias gariepinus X Clarias macro-cephalus) exposed to chlorpyrifos and carbaryl. J. Environ. Biol. 30, 83e88.

Turolla, F., 2007. La venericoltura in Italia. In: FAO Actas de Pesca y Acuicultura. FAO,Roma, pp. 177e188. No. 12.

Tusher, V.G., Tibshirani, R., Chu, G., 2001. Significance analysis of microarraysapplied to the ionizing radiation response. Proc. Natl. Acad. Sci. U. S. A. 98,5116e5121.

Valavanidis, A., Vlahogianni, T., Dassenakis, M., Scoullos, M., 2006. Molecular bio-markers of oxidative stress in aquatic organisms in relation to toxic environ-mental pollutants. Ecotoxicol. Environ. Saf. 64, 178e189. http://dx.doi.org/10.1016/j.ecoenv.2005.03.013.

Van Aggelen, G., Ankley, G.T., Baldwin, W.S., et al., 2010. Integrating omic technol-ogies into aquatic ecological risk assessment and environmental monitoring:hurdles, achievements, and future outlook. Environ. Health Perspect. 118, 1e5.http://dx.doi.org/10.1289/ehp.0900985.

Van Straalen, N.M., Feder, M.E., 2012. Ecological and evolutionary functionalgenomicsdhow can it contribute to the risk assessment of chemicals? Environ.Sci. Technol. 46, 3e9. http://dx.doi.org/10.1021/es2034153.

Vitousek, P.M., Mooney, H.A., Lubchenco, J., Melillo, J.M., 1997. Human dominationof Earth's ecosystems. Science 277, 494e499. http://dx.doi.org/10.1126/science.277.5325.494.

WFD 2000/60/EC. Water Framework Directive. Directive 2000/60/EC of the Euro-pean Parliament and of the Council of 23 October 2000 Establishing a Frame-work for Community Action in the Field of Water Policy. http://eur-lex.europa.eu/resource.html?uri¼cellar:5c835afb-2ec6-4577-bdf8-756d3d694eeb.0004.02/DOC_1&format¼PDF.

Zhang, G., Fang, X., Guo, X., et al., 2012. The oyster genome reveals stress adaptationand complexity of shell formation. Nature 490, 49e54. http://dx.doi.org/10.1038/nature11413.