A high-throughput sequencing ecotoxicology study of freshwater bacterial communities and their responses to tebuconazole

Abstract

The pollution of lakes and rivers by pesticides is a growing problem worldwide. However, the impacts of these substances on microbial communities are still poorly understood, partly because next-generation sequencing (NGS) has rarely been used in an ecotoxicology context to study bacterial communities despite its interest for accessing rare taxa. Microcosm experiments were carried out to evaluate the effects of tebuconazole (TBZ) on the structure and composition of bacterial communities from two types of freshwater ecosystem (lakes and rivers) with differing histories of pollutant contamination (pristine vs. previously exposed sites). Pyrosequencing revealed that bacterial diversity was higher in the river than in the lakes and in previously exposed sites than in pristine sites. Lakes and river stations shared very few OTUs, and differences at the phylum level were identified between these ecosystems (i.e. the relative importance of Actinobacteria and Gammaproteobacteria). Despite differences between these ecosystems and their contamination history, no significant effect of TBZ on bacterial community structure or composition was observed. Compared to functional parameters that displayed variable responses, we demonstrated that a combination of classical methods and NGS is necessary to investigate the ecotoxicological responses of microbial communities to pollutants.

The combination of classical methods and NGS approach is necessary when tackling the ecotoxicological responses of microbial communities to pollutants.

Open in new tabDownload slide

Introduction

So far, very few ecotoxicological studies have explored the potential of next-generation sequencing technologies (NGS) for assessing the impact of pollutants on microbial communities (dos Santos et al., 2011; Ge et al., 2012; Chariton et al., 2014), even this could offer a wealth of information about the effects of pollutants effects at the community level (Pesce et al., 2011; Zinger et al., 2012). Indeed, most studies of the effects of pollutants on microbial community structure and composition have been based on fingerprinting or cloning–sequencing approaches (e.g. Tadonléké et al., 2009; Villeneuve et al., 2011), accessing only the dominant species within the communities (Pedrós-Alió, 2006; Fuhrman, 2009). As shown by Sjöstedt et al. (2012) on marine bacterioplankton communities exposed to different salinity concentrations, pyrosequencing makes it possible to access the composition of the whole community by means of thousands of reads and thus to reveal minor or major changes in the relative abundances of various species.

Moreover, despite major concerns about pesticides as priority environmental contaminants in the context of European Water Framework Directive, their effects on microbial community structure and composition remain unknown as they have seldom been assessed by means of these new methods (Richardson, 2009). The azole fungicide pesticides are one of the most important groups of systemic site-specific fungicides and have been intensively used since the 1980s against a broad spectrum of plant pathogenic fungi (Becher et al., 2010). One of these pesticides, tebuconazole (C₁₆H₂₂CIN₃O) (TBZ), belongs to the triazole class and inhibits the biosynthesis of ergosterol (a component of yeast and fungal cell membranes; Copping & Hewitt, 1998). This fungicide is essentially used to target plant-infecting fungi, but displays other complex effects on ecosystems such as affecting the endocrine system of several organisms (e.g. Richardson, 2009) or impacting nontargeted aquatic fungi, such as the hyphomycetes (Dijksterhuis et al., 2011; Artigas et al., 2012). When looking to the effects of TBZ on bacterial communities, Jackson et al. (2000) have demonstrated the bactericidal activity of TBZ on the species Mycobacterium smegmatis and Candida albicans, but no data are available on the ecophysiological effects of TBZ on bacteria. Some other authors have highlighted the ability of bacteria to degrade TBZ under controlled conditions, suggesting that this molecule can be used as carbon source by microorganisms (Wallace & Dickinson, 2006; Obanda & Shupe, 2009; Sehnem et al., 2010). Despite increasing evidence of the widespread contamination of surface waters by TBZ (Knäbel et al., 2014), widely contrasting results have been obtained in studies of its impact on natural microbial communities in soils (Strickland et al., 2004; Cycon et al., 2006; Muñoz-Leoz et al., 2011).

In the context of a paucity of knowledge about the impact of TBZ on indigenous aquatic microbial communities, we investigated the impact of TBZ on freshwater bacterial communities as part of a larger scientific project intended to reveal the responses of microbial communities to this pollutant, as well as its potential degradation by biological processes. We therefore carried out a 21-day microcosm experiment with two planktonic bacterial communities isolated from two subalpine lakes and two benthic bacterial communities isolated from two stations on a French river; these communities had all experienced different levels of previous exposure to pollutants. Different TBZ treatments were applied to communities from each lake and river in order to simulate the contamination levels encountered in natural aquatic ecosystems. Previous findings had highlighted some moderate structural and functional effects of TBZ on algal and bacterial communities (Artigas et al., 2014). Nevertheless, assessment of the effect was based on general structural parameters, such as bacterial densities and algal biomass (chlorophyll-a concentrations) and on highly redundant activities in microbial communities (i.e. bacterial production, substrate-induced respiration and photosynthesis). To complete this study, we investigated the impact of TBZ at a deeper level, by focusing on the structure and composition of freshwater bacterial communities, as well as on abundant and rare species, using NGS approaches. In view of previous results (Artigas et al., 2014), our main working hypotheses were that (1) depending on the lifestyle (planktonic vs. benthic) and the contamination history of these bacterial communities, different responses in terms of richness, diversity and composition could be observed after exposure to the pollutant and (2) TBZ might lead to the replacement of some dominant species in the communities by species that were less abundant or even rare before the pollutant was added.

Materials and methods

Field sampling strategy

Bacterial communities were collected from two lakes and one river characterized by differing histories of pollutant contamination. Plankton samples were collected from two lakes located in the subalpine area of Western Europe. Lake Geneva (altitude 311 m, 46.378750°N, 6.442867°E) is located between France and Switzerland and is surrounded by vineyards, and agricultural and urbanized areas. The average TBZ concentration in this lake prior to the present study was 0.001 μg L⁻¹, and in this study, it was considered to be a previously exposed lake (Lake-PE). Lake Aiguebelette (altitude 390 m, 45.55537°N, 5.80135°E) is surrounded by grassland and woods and is less exposed to pollutants than Lake Geneva, which is characterized by a larger catchment area and higher anthropogenic pressures (Thevenon et al., 2011). Previous TBZ concentrations in its waters were < 0.001 μg L⁻¹ and so it was considered to be our pristine lake (Lake-PI). Four hundred litres of freshwater was collected from each lake on 16 August 2010 at a depth of 50 cm and immediately filtered through a 50-μm pore size net to remove the metazooplankton. The prefiltered freshwater was stored overnight at 4 °C until the microcosm experiment was set up (Artigas et al., 2014).

Benthic bacterial communities were collected from the River Morcille. This ecosystem is located in the Beaujolais wine-producing area of eastern France (46.150°N, 4.600°E) and presents a pesticide contamination gradient along the river, because of the increasing (1) area of vineyards, (2) length of ditches connected to the stream and (3) housing surface area (Montuelle et al., 2010). The upstream segment of the river (St Joseph) was considered to be a pristine river station (River-PI), with TBZ concentrations ranging from 0.001 to 0.1 μg L⁻¹, and the downstream segment of the river (St Ennemond) was considered to be the previously exposed river station (River-PE) with TBZ concentrations ranging from 0.01 to 4 μg L⁻¹. Moreover, the overall exposition to pollutants is much higher at the River-PE station than at the River-PI station (Kim Tiam et al., 2014). Artificial glass substrata (20-cm² tiles) were left to be naturally colonized by biofilms at both river stations for a period of 1 month in late summer 2010. The resulting pristine and previously exposed biofilms were collected and transported to the laboratory for further experiments in indoor channels (Artigas et al., 2014).

Setting up the microcosms

For the lakes, the microcosm experiment was performed in eighteen 20-L polycarbonate bottles (Nalgene, Dominique Dutscher, France) filled with freshwater collected from the two lakes (two lakes × three treatments × three replicates). The prepared bottles were incubated in a growth chamber at 20 °C with a 12 h/12 h light/dark cycle and air circulated through the bottles (SECOH pump SLL 50). For the river water, the biofilm glass tiles were incubated in 18 indoor glass channels (two river stations × three treatments × three replicates), each of them connected by means of an aquarium pump (NEWA MJ 750) to a separate 20-L reservoir (containing dechlorinated tap water). Water temperature was regulated at 20 °C, the water flow rate was set at 1.2 L min⁻¹, and there was a 13 h/11 h light/dark cycle. More details about the channels are given in Artigas et al. (2014). TBZ (C₆H₂₂CIN₃O; Sigma Aldrich) was added to the microcosms and the channels on day 0 (the start day of the study) at 2 or 20 μg L⁻¹, whereas the control treatments were not contaminated. The physico-chemical parameters measured during the experiments are presented elsewhere (Artigas et al., 2014).

Sampling of the microbial communities was performed directly on bottles for the plankton communities, and glass tiles were taken for biofilm community survey. For the lake communities, 100 mL of freshwater was sampled after gently shaking each microcosm on days 0, 3, 6, 9, 13, 17 and 21 after adding TBZ, which produced a total of 114 samples. First, plankton samples were filtered through a 1.2-μm polycarbonate filter (Isopore Membrane Millipore, Dustcher, France) to separate the eukaryotes. Second, the filtrate was passed through a 0.2-μm polycarbonate filter (Nucleopore Polycarbonate Whatman; Fischer, France) to concentrate the prokaryote fraction on the filter. For the river communities, sampling was performed on days 0, 3, 9, 13, 17 and 21 after adding the TBZ, producing 96 samples; one colonized glass tile was suspended in 45 mL of water from the corresponding channel, sonicated for 90 s (40W power, 40 kHz frequency) to detach the microorganisms (Romani et al., 2004) and then filtered as described above for the lake communities. All filters were stored at −20 °C until DNA extraction was performed.

DNA extraction

DNA was extracted from the 0.2-μm filters following the procedure described in Massana et al. (1997) with some modifications. Briefly, each filter was mixed with 750 μL of lysis buffer (40 mM EDTA, 50 mM Tris–HCl, 0.75 M sucrose), re-frozen at −80 °C for 15 min, and then thawed by putting the tubes into a water bath at 55 °C for 2 min, before being vortexed and placed in a sonication bath for 2 min. Lysozyme (0.6 mg mL⁻¹) was added to the filters, which were then incubated at 37 °C for 45 min with gentle stirring. Subsequently, sodium dodecyl sulfate (1% final concentration) and proteinase K (0.2 mg mL⁻¹ final concentration; Eurobio, Courtaboeuf, France) were added, and the filters were incubated at 55 °C for at least 90 min. After quick centrifugation step, the lysates were collected and purified twice with phenol–chloroform–isoamyl alcohol. After precipitating with 0.1 volume of sodium acetate and 0.6 volume of isopropanol, the nucleic acids were washed with 80% ethanol and then dissolved in 100 μL milliQ water. The extracted DNA was stored at −20 °C until analysis.

Pyrosequencing

The diversity of the bacterial communities was assessed by pyrosequencing of the 16S rRNA gene. Primers 563F (Claesson et al., 2010) and 907rM (Schauer et al., 2003) were used to amplify a 16S rRNA gene fragment covering the V4–V5 regions (hypervariable regions), corresponding to a size of c.350 bp entirely pyrosequenced by Titanium FLX technology. Ten base pair tags in the 5′ position were added to the primers in order to identify each sample (as recommended by the manufacturer, Roche) together with the following adaptors: A Adaptor – TAGx – 563F and B Adaptor – 907rM. For each of the 210 samples (114 from the lakes and 96 from the river stations), three replicated PCR reactions of 50 μL (following manufacturer's instructions of TaKaRa LA Taq, Lonza) were conducted under the following conditions: 94 °C for 3 min, 25 cycles at 94 °C for 50 s (denaturing), 52 °C for 30 s (annealing) and 72 °C for 30 s (extension), followed by 5 min at 72 °C. The pool of amplicons for each sample was purified using a Multiscreen PCR_μ96 Plate (Millipore, St Quentin en Yvelines, France) according to manufacturer's instructions. Pyrosequencing was carried out on a Genome Sequencer Flx 454 (plateforme GINA; Centre Jean Perrin, Clermont-Ferrand, France), using 30 ng of pooled amplicons per sample. A total of 156 PCR products were sequenced from the 210 DNA samples, with at least one sample per microcosm triplicate.

Cleaning, clustering and phylogenetic affiliation of the 16S rRNA gene fragments

A total of 379 747 reads (half-plate run) for the lake samples and 262 877 reads (half-plate run) for the river samples were obtained. These raw reads were processed through different cleaning steps using pangea software (Giongo et al., 2010), which removed short reads (i.e. reads not including both forward and reverse primers) and reads with low-quality scores. Reads that perfectly matched these primers (100% identity) were selected with Fuznuc (Rice et al., 2000).

The quality-controlled reads were then analysed with PANAM 16S: a tool for the Phylogenetic Analysis and Taxonomic Affiliation of SSU rRNA Amplicons (Taib et al., 2013; https://code.google.com/p/panam-phylogenetic-annotation/downloads/list). Reads were de-multiplexed according to the tag in order to recover reads from each sample and then clustered into Operational Taxonomic Units (OTUs) with UCLUST (Edgar, 2010) with a cut-off of 98% similarity (equivalent for the V4–V5 region to 97% of similarity on the complete 16S rRNA gene). The OTUs were then compared to a database of phylum-specific alignments (based on SSUref SILVA, Taib et al., 2013) and affiliated to the different phylogenetic groups. Ad hoc profile alignments (hmmer, Eddy, 1998) were generated and used to build maximum likelihood phylogenetic trees (FastTree, Price et al., 2009). Based on these trees, an accurate taxonomy was assessed for each OTU, and putative environmental clades were highlighted.

In this study, the thresholds used to define abundant and rare OTUs are the same as those used in Galand et al. (2009). Abundant OTUs correspond to OTUs with a proportion (relative to total reads) of > 1% within a sample, whereas rare OTUs correspond to OTUs with a proportion of < 0.01%.

Analyses and statistics

The number of reads was normalized by randomly selecting the same number of reads per environment/treatment, based upon the sample with the lowest number of reads (n = 14 000 reads for control assessment and n = 12 000 reads for TBZ treatments). Three randomized selections were performed and averaged before the construction of Venn diagrams. In the same way, diversity indices were also calculated on the three normalized selections.

Venn diagrams for graphical descriptions of unshared and shared OTUs between environments or treatments were designed using the library VennDiagram under r software version 2.14.2 (R Development Core Team, 2013). Principal component analysis (based on the relative abundance of all bacterial OTUs), which provided an ordination of bacterial communities in a factorial map based on the scores of the first two principal components, was performed with the ade4 package under r software (R Development Core Team, 2013). Diversity indices were assessed with package fossil for Chao1, vegan for Shannon and ADE-4 (Dray et al., 2007) for Sorensen under r software (R Development Core Team, 2013). Finally, UniFrac was used to compare all the microbial communities using phylogenetic information (Lozupone & Knight, 2005).

Nucleotide sequence accession numbers

16S sequences data are available on Mg-Rast Web server (http://metagenomics.anl.gov) under projects 4556432.3, 4556433.3, 4556434.3 and 4556435.3.

Results

From the total of 642 624 reads obtained from all samples, a total of 276 806 high-quality reads were selected after the cleaning step and were clustered into 13 783 OTUs using a cut-off of 98% sequence similarity.

Structure and composition of planktonic and benthic bacterial communities from the lake and river stations

We first compared the structure and composition of planktonic and benthic bacterial communities of the lakes and river samples, respectively, and depending on their different histories of exposure to pollutants (PI: pristine vs. PE: previously exposed). By considering all the reads from control collected during our experiment, a total of 31 phyla and candidate divisions (cds) were identified from the two lakes and river stations. Among them, only 10 phyla contained more than 1% of the reads, the other phyla being poorly represented (< 1%) (Fig. 1). Proteobacteria was the dominant phylum at every site (from 46% to 69% of the reads), and within this phylum, the Alphaproteobacteria (33–58%) were the most abundant. The main difference between communities from the lakes and river stations was the relative importance of Actinobacteria, which were highly represented in the lakes (c.30% of the reads), but accounted for < 1% of the reads for the river stations. The opposite pattern was found for Gammaproteobacteria, which accounted for from 10% to 16% of the reads in the river samples, but < 1% in the lake samples. To a lesser extent, Bacteroidetes were also dominant, being found mainly at both river stations (13–17%) but only in the Lake-PE sample (12%). The phylogenetic affiliation of the most abundant OTUs is presented in Supporting Information, Table S1. Briefly, the two main clades detected in the lake samples are the LD12 clade (Alphaproteobacteria), which contained 27% and 54% of the reads for the Lake-PE and the Lake-PI, respectively, followed by the hgcI clade (Actinobacteria), which accounted for about 22% of the reads in lake samples. At the river stations, in contrast, the GOBB3-C201 cluster (Alphaproteobacteria) was the dominant group (River-PI: 15%; River-PE: 6.5%), followed by another Alphaproteobacteria, clade wr0007 (river-PI: 9.1%; river-PE: 4%) and a clade belonging to Rickettsiales (River-PI: 6.9%; River-PE: 8.8%). Finally, the Legionella genus (Gammaproteobacteria) was well represented in the river samples (River-PI: 11.4%; River-PE: 6.5%). These taxonomic observations highlighted differences between the lakes and river stations at the different level of taxonomic affiliation (from phylum to OTU), these differences being greater when considering the OTU level.

At the OTU level, after normalizing the number of reads (14 000 reads per environment, i.e. a total of 56 000 reads), 1577 OTUs were detected in the Lake-PI, 1911 in the Lake-PE, 2278 in the River-PI and 3095 in the River-PE. Less than a quarter of the OTUs (22%) from the lakes were shared by both lakes (635 OTUs), and among the OTUs from the river stations, only 29% were found at both river stations (1206 OTUs, Fig. 2). The OTUs shared by both lakes plus those shared by the river stations represented around 80% of the reads (c.40% for each ecosystem). When the four environments (lakes and river stations) were compared, we found that they shared only 26 OTUs, whereas 251 OTUs (3.7%) were shared by at least one lake and one river station (Fig. 2). All these OTUs that were found in both kinds of ecosystems (lake and river) accounted for around 10% of the reads. Finally, the numerous OTUs detected in only one environment (4866 OTUs; 72% of the OTUs) generally contained a limited number of reads (0.004% in average of the reads per OTU; percentage estimated vs. the whole number of reads) accounting for 21% of the total number of reads (Fig. 2), only some of them containing a significant number of reads (up to 0.4%, percentage estimated on the whole of reads). Hence, it appears that most of the limited number of OTUs detected in several environments were abundant ones, whereas the numerous OTUs detected in only one environment were generally less abundant (0.01% < abundance of reads < 1.0%) or rare (< 0.01% of reads).

Comparison of the four aquatic environments, based on Sørensen and the Bray-Curtis similarity indices (Table 1), showed that the intra-ecosystem diversity (Lake-PI vs. Lake-PE; River-PI vs. River-PE) displayed higher similarity values (0.413/0.472 and 0.464/0.380, respectively) than the interecosystem diversity (Lake vs. River) (< 0.077/< 0.031). Based on the OTU richness, the Chao1 values (estimated from the entire data set of the control experiments) were greater at the river stations (3213 for the River-PI and 4161 for the River-PE) than in the lakes (1878 for the Lake-PI and 2198 for the Lake-PE; Table 1). In the same way, the Shannon diversity index values were also greater for the river stations than for the lakes, which was consistent with the rarefaction curves (Fig. S1). Interestingly, the Shannon index values were very similar in the control samples at all dates and at T0 (the start date of the experiment, just after the communities had been sampled from their respective environment), showing that the diversity of bacterial communities in the untreated samples did not decrease during the 21 days of the experiment. In addition, our data also showed that bacterial communities from the previously exposed sites (River-PE and Lake-PE) were richer and more diverse (Shannon index) than those from the pristine environments (River-PI and Lake-PI).

Effect of TBZ on planktonic and benthic bacterial communities

To test the effect of TBZ, the bacterial community structure and composition were determined during a 21-day incubation period under three experimental conditions (control, 2 μg L⁻¹ TBZ and 20 μg L⁻¹ TBZ) for each lake and river station. The principal component analysis (Fig. 3) based on bacterial OTUs showed on the first axis, which accounted for 43% of the variability, that plankton communities from the two lakes and benthic communities from the two river stations could be distinguished. The second axis, which accounted for only 12% of variability, discriminated between the two lakes, and the early and late incubation dates for the two river stations. No differences were found between the changes occurring over time in the bacterial communities in relation to the TBZ contamination, whatever the treatment or the type of ecosystem (i.e. lake or river) (Fig. S2). Similarly, the UniFrac analysis did not reveal any impact of TBZ on the composition of the bacterial communities.

We investigated also whether some abundant (> 1%) OTUs had become rare (< 0.01%), and vice versa, when bacterial communities from control and TBZ treatments were compared. None of the four environments studied displayed any evidence of such a shift in OTU abundance (see an example of a rank–abundance curve, Fig. S3). Moreover, as illustrated by the Venn diagrams (Fig. 4), the OTU distribution in each environment for the three treatments (control, 2 and 20 μg L⁻¹ TBZ) showed that the specific OTUs (found in only one treatment) corresponded to a very restricted number of reads and most of them were rare OTUs (< 0.01% of reads for one environment, Table 2). Finally, when comparing the OTUs affiliation at four different taxonomic levels (phylum, class, order and family) in each environment, no major differences were observed depending on the TBZ exposure (Fig. S4). Taken together, these data strongly suggest that TBZ did not have any significant impact on either abundant or rare OTUs of the bacterial communities in the lakes and river tested.

Discussion

In a previous joint paper based on the same microcosm experiments (Artigas et al., 2014), we pointed out that TBZ has moderate but differing effects on several structural and functional parameters of the microbial communities, suggesting that these effects tended to be modulated by the lifestyle of these communities (benthic vs. planktonic) and whether there had been a previous history of exposure to various pollutants, including TBZ. In particular, it was shown in the presence of TBZ that pristine biofilm communities (River-PI) display a decrease in their heterotrophic respiration and photosynthetic activities and an increase in bacterial cell mortality. On the other hand, plankton communities only displayed a slight increase in bacterial densities and were weakly affected by TBZ. Moreover, the decrease in TBZ concentrations in water was steeper in biofilm experiments (river stations) than in the planktonic ones (lakes), and secondary metabolites of TBZ were only detected in the biofilm experiments. We therefore expected that the TBZ-induced changes observed in the viability and the metabolism of microbial communities could be accompanied by changes in the community composition (i.e. richness and diversity descriptors), in particular in the heterotrophic compartment (mainly composed of bacteria and fungi).

Our results revealed that TBZ did not have a significant impact on the structure or composition of the bacterial communities, whether from plankton or from biofilm samples. Despite large differences at the OTU level in the initial composition of both benthic and planktonic communities from river and lakes, only temporal changes have been found in the bacterial communities during the course of the experiments. In this sense, no differences were observed between the trajectories of control and TBZ-exposed communities by multivariate analysis. The occurrence of temporal changes in bacterial community composition (BCC) is common during the development of freshwater biofilms (Jackson et al., 2001; Lyautey et al., 2005) and at seasonal scale in planktonic bacterial communities (Eiler et al., 2012; Paver et al., 2013).

Although there was no significant difference in the overall change of community composition in TBZ-exposed and control bacterial communities during the experiments, a question open remained concerning the potential impact of the pollutant on rare species. Indeed, our analysis was based on a multivariate analysis that mainly takes into account changes occurring in the relative abundance of dominant species. Our hypothesis was that TBZ could lead to the replacement of some dominant species in the bacterial communities by species that were less abundant or even rare before the exposure to the pollutant. In our study, no significant difference emerged in the temporal evolution of the richness and diversity estimators between control and TBZ treatment experiments. Moreover, the NGS approach, which provides a detailed description of the BCC, did not reveal any specific selection of rare species in the samples exposed to TBZ during the experiment. Taken together, our findings suggested that under our experimental conditions, the direct impact of TBZ on the structure and composition of bacterial communities living in freshwater ecosystems is probably very limited, at least over a short timescale and with the concentrations of TBZ tested in the present study. These results are consistent with those of Bending et al. (2007) showing that TBZ did not affect the soil bacterial community structure (assessed by PCR-DGGE), but did reduce dehydrogenase activity, knowing that dehydrogenases play a major role in the biological degradation of the organic matter and are an indicator of overall soil microbial activity (Wolinska & Stepniewska, 2012). Although Jackson et al. (2000) showed that TBZ displayed bactericidal activity against some bacterial species (e.g. M. smegmatis and C. albicans), their findings were obtained using cultures exposed to very high TBZ concentrations (between 4 and 64 mg L⁻¹ depending on the species). In our study, the TBZ doses tested were chosen to mimic realistic concentrations found in freshwater environments. Similarly, our experimental protocols were designed to minimize the stress resulting from culture conditions undergone by the bacterial communities. The absence of any decrease in the richness and diversity estimators during the experiment (the Shannon values were the same at the beginning of the experiment and at the other sampling dates) suggested that our experimental conditions were not too stressful for bacterial communities and that bottle confinement did not significantly affect the microbial communities studied. The lack of any effect of TBZ on the BCC suggested that the TBZ-induced functional changes previously described in Artigas et al. (2014) were mainly due to effects on eukaryotic communities (i.e. algae for photosynthesis and fungi for glucose-based respiration) knowing that TBZ, like all triazoles, is known to inhibit sterol synthesis in plants and fungi.

The present work extends current knowledge about the structure and community composition of bacterial communities in lake and rivers, as the knowledge available about these communities in freshwater biofilm is very limited (Besemer et al., 2012; Hall et al., 2012; Bricheux et al., 2013). First, we have shown that river and lake bacterial communities displayed some common features in their overall structure (like the dominance of Proteobacteria, in particular of Alphaproteobacteria), but also differed by the relative contribution of Actinobacteria and Gammaproteobacteria to these communities. The higher nitrogen and phosphorus concentrations at the river stations compared to the two lakes (Artigas et al., 2014) could explain the differences in the relative importance of Actinobacteria and Gammaproeobacteria in the lakes and river. Indeed, in a recent study of stream biofilms in a gradient of nutrient resources, variations of the relative abundance of Actinobacteria and Gammaproteobacteria were also observed, with an increase of Gammaproteobacteria at the highest levels of enrichment (Van Horn et al., 2011). Finally, Newton & McMahon (2011) have also positively correlated Bacteroidetes levels with the addition of nutrients, notably with labile carbon (glucose). The oligotrophic status of the Lake-PI (Lake Aiguebelette) can thus explain the absence of Bacteroidetes in the bacterial community of this lake.

At a fine phylogenetic scale, our study reveals dissimilar OTU composition in lakes and rivers, which is in agreement with previous data reported by Tamames et al. (2010) showing that the environmental specificity emerges at lower taxonomic levels, such as the genus and species levels. A typical example is the Alphaproteobacteria phylum, which was recovered in similar proportions from lakes and rivers, but was dominated by different clusters: the LD12 cluster and GOBB3-C201 for lakes and rivers, respectively (Glöckner et al., 2000; Zwart et al., 2002; Debroas et al., 2009). Similarly, the other abundant Alphaproteobacteria clade in two river stations, clade wr0007, was not found in the lakes. This clade is assigned to an uncultured rape rhizosphere bacterium, which confirms the close relationships between the bacterial community from the River Morcille and that from the soil of the river bank, as had already been suggested by Pesce et al. (2010). Finally, the occurrence of Legionella species in the river is not surprising, because aquatic biofilms are good ecological niches for such species (Atlas, 1999; Parthuisot et al., 2010).

At the OTU level, we observed that bacterial species richness was greater in biofilms than in plankton communities. They shared very few OTUs, which contained a significant proportion of the total reads number, suggesting that their distribution in these two ecosystems was not due to dispersion between the different environments. More generally, the OTUs found in all four environments studied (two lakes and two river stations) seemed to contain a large number of reads, whereas OTUs found in only one environment contained few reads. As already observed by Humbert et al. (2009) in temperate and tropical lake bacterial communities, there was a positive correlation between the mean local abundance of each OTU and the number of locations occupied.

Another interesting finding of the present study is that the richness and diversity of benthic bacterial communities from river stations were higher than those of planktonic bacterial communities from the two lakes. It is known that lifestyle, free suspension (planktonic communities) vs. biofilm (benthic communities), has a major influence on the diversity of bacterial communities (Verhagen et al., 2011; Besemer et al., 2012). In particular, three major processes have been proposed to explain the higher diversity found in benthic microbial communities than in planktonic ones. First, the biofilm habitat offers more variable microhabitats and thus a greater number of ecological micro-niches (for example, with regard to redox and nutrient gradients) than the water column, which allows these ecosystems to support more species. Second, it is known that biofilms display considerable autogenic variability during their development, and this could contribute to explain their higher diversity (Jackson et al., 2001; Lyautey et al., 2005). This hypothesis is in agreement with the fact that the Shannon values for the lakes and river stations were similar at the beginning of the experiment (T0), but differed (data not shown) by the end. Third, river conditions (such as the rate of flow, discharge variations, terrestrial organic matter inputs) are more variable than lake conditions (during the sampling period, the water column was stratified), which probably promotes greater diversity in the microbial communities of the river than in the lakes.

Finally, it appeared that the diversity of bacterial communities in the previously exposed (PE) environments was greater than in those of the pristine environments (PI). However, this result must be interpreted with caution, because it is difficult to distinguish between the direct or indirect impacts of previous exposure to pesticides, from the impacts of confounding factors such as nutrient levels. It is well known that pollution of aquatic ecosystems by pesticides is often associated with nutrient pollution from fertilizers, especially in agricultural areas. Both types of pollutants (pesticides and nutrients) can have direct and indirect impacts on the biodiversity of microbial communities (Pesce et al., 2008; Tlili et al., 2010; Villeneuve et al., 2011).

Acknowledgements

This study was funded by the Sendefo project in the Contaminants Ecosystems Health (CES-2009) programme of the French ANR (Agence Nationale de la Recherche). M. Ghosh is acknowledged for revising the English version of the manuscript, and the two anonymous reviewers are thanked for helping us to improve the manuscript.

References

Author notes