Abstract
Fish farming can have a negative impact on water quality and aquatic organisms due to emerging blooms of Cyanobacteria and the production of cyanotoxins. In this study, the effect of aquaculture in hydroelectric reservoirs in Brazil was evaluated in six fish farms and in upstream and downstream water through analysis of the microbiome, Cyanobacteria and microcystin concentrations. Synechococcus and Microcystis were observed at all six locations, while Limnothrix was also observed abundantly at two locations. An increase in the relative abundance of Cyanobacteria inside the fish farms was observed at two locations, while an increase of Cyanobacteria was observed in downstream at five of the six locations. Microcystins were detected in significant and high values in all locations, with concentrations up to 1.59 μg/L. The trend in microcystin concentrations was mirrored in copy numbers of the mcyE gene (encodes microcystin synthetase) and presence of Microcystis, but not in any of the other observed cyanobacterial groups. In summary, the study shows that aquaculture production influenced the water microbiome inside and downstream the fish farms, and a direct correlation was found between mcyE gene copies, microcystin production and abundance of Microcystis, but not for the total abundance of Cyanobacteria.
HIGHLIGHTS
Fish breeding affected the microbial community structure in Brazilian tilapia farms.
Synechococcus and Microcystis were the dominant cyanobacterial genera.
Microcystins and the mcyE gene were detected at all locations.
Aquaculture production stimulated the relative abundance of Cyanobacteria downstream the tilapia farms.
INTRODUCTION
Cyanobacteria constitute a diverse group of photosynthetic bacteria that are found in most environments exposed to light (Whitton 2012). In recent years, harmful blooms of Cyanobacteria (CyanoHABs) have become a concern due to a global increase in frequency, duration and magnitude in response to accelerated eutrophication and climate change (Paerl & Paul 2012). Aquaculture production is one of the contributing factors for eutrophication of freshwaters, mainly due to undigested and excess fish feed, faeces and other fish wastes. This leads to elevated levels of nitrogen and phosphorous compounds and of organic matter in aquaculture areas (Martínez-Córdova et al. 2009). These nutrients, in combination with oxygen production from cyanobacterial activity, provide excellent growth conditions for many heterotrophic bacteria. Populations of heterotrophic bacteria are the main players in biomass formation and biodegradation during CyanoHABs (Cai et al. 2014).
Analysis of microbial populations represents a powerful approach in characterizing the factors that select for specifically involved microbiota and the roles of these microorganisms in ecosystems. This knowledge is highly relevant for understanding the ecosystem dynamics, but it is also useful for developing effective strategies to manage and manipulate microbial communities (Ghanbari et al. 2015). Therefore, this microbiome study aims at improving our knowledge on bacterial community profiles in hydroelectric reservoirs and how these communities respond to aquaculture production.
Growth of heterotrophic bacteria during CyanoHABs can cause depletion of oxygen when bacteria degrade organic matter from the dense blooms. Furthermore, increased turbidity and reduced light penetration due to the CyanoHABs may also negatively affect the abundance of submerged plants, eradicate aquatic animals and alter food web dynamics in aquatic ecosystems (Paerl & Paul 2012). Moreover, some Cyanobacteria produce secondary metabolites, some of which are harmful due to their toxicity (Drobac et al. 2016). The toxic metabolites (cyanotoxins) can cause human and animal health problems and in some cases even lead to mortalities (Carmichael 2001). In fish, exposure to cyanotoxins may result in accumulation of the toxins in tissues and cause histopathological changes of vital organs, affecting both growth, development, reproduction and survival (Drobac et al. 2016). The toxins may also threaten the quality and safety of fish intended for human consumption. Cyanotoxins have diverse chemical structures and toxicological properties and are usually organized into three classes according to their target organ: hepatotoxins, neurotoxins and dermatotoxins (Merel et al. 2013).
The most well-known cyanotoxin family is the monocyclic heptapeptide microcystins (MCs), which consist of more than 310 variants (Jones et al. 2020). The primary effect of MCs in eukaryotic organisms is inhibition of protein phosphatases in the liver (Campos & Vasconcelos 2010), leading to hepatocyte deformation, collapse of tissue organization and necrosis (Falconer & Yeung 1992). Recent research suggests that MCs may also target other organs, such as kidneys and lungs, and affect intestines, reproductive systems and the brain (Dias et al. 2014). Exposure to low but chronic concentrations of MC-LR has been found to induce liver and skin tumors (Nishiwaki-Matsushima et al. 1992). The MC synthetase (mcy) gene cluster encodes an enzyme complex responsible for MC biosynthesis (Nishizawa et al. 2000). Ten genes (mcyA to mcyJ) form this cluster (55 kbp total) in Microcystis, Anabaena and Planktothrix (Nishizawa et al. 2000; Tillett et al. 2000).
In Brazil, a growing population and an increased fish consumption have introduced fish farming in cages as a feasible technology for tilapia production in large hydroelectric reservoirs. However, since intensive fish stocks and overcrowding in the cages risk implicating water pollution (David et al. 2015), aquaculture development should be planned and designed in a responsible manner to minimize negative impacts on water quality, since hydroelectric reservoirs have multiple users (Hambrey & Senior 2007).
The present study aims at assessing the effects of aquaculture production on microbial water quality in selected Brazilian hydroelectric reservoirs. The potential effects were addressed by analysis of the water microbiome, composition and abundance of Cyanobacteria, and the composition, production and concentrations of MCs in six fish farms and in your upstream and downstream waters.
MATERIALS AND METHODS
Sampling
In October 2015, six fish farms located in hydroelectric reservoirs of São Paulo state, Brazil, with the production of Nile tilapia (Oreochromis niloticus) in cages, were studied. The reservoir were Chavantes (farm F1), Nova Avanhandava (farms F2, F3 and F4) and Ilha Solteira (farms F5 and F6) (Figure 1). Water samples were collected in triplicate in 2 L sterile bottles at 0.6 m depth within the fish farms and downstream and upstream of the farms (a total of 17 points).
Sampling sites in six fish farms in three hydroelectric reservoirs in São Paulo state, Brazil. The reservoirs were Chavantes (farm F1), Nova Avanhandava (farms F2, F3 and F4) and Ilha Solteira (farms F5 and F6). Numbers 1–3 represent reservoirs; U, upstream; F, inside fish farm; D, downstream. Fish farms 3 and 4 share upstream and downstream water.
Sampling sites in six fish farms in three hydroelectric reservoirs in São Paulo state, Brazil. The reservoirs were Chavantes (farm F1), Nova Avanhandava (farms F2, F3 and F4) and Ilha Solteira (farms F5 and F6). Numbers 1–3 represent reservoirs; U, upstream; F, inside fish farm; D, downstream. Fish farms 3 and 4 share upstream and downstream water.
DNA extraction
DNA extraction was performed on 200 mL of water using the Wizard® SV Genomic DNA Purification System Kit (Promega®, Southampton, UK) according to the manufacturer's instructions. Genomic DNA was eluted in 20 μL of nuclease-free water and stored at −20 °C until use. Purity of the extracted DNA was validated using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, MA, USA) and quantified using a Qubit® 2.0 spectrofluorometer using the dsDNA BR Assay Kit (Invitrogen™, Carlsbad, CA, USA).
16S rRNA gene amplicon sequencing
The DNA extracts were subjected to PCR amplification of the V6 to V8 hypervariable region of the 16S rRNA gene using the previously described Cyanobacteria targeting primer pair CC/CD (Rudi et al. 1997). The reaction mixture consisted of 10 μL of GoTaq® colourless Mastermix (Promega®, Southampton, UK), 0.4 μM of each primer (CC and CD), 10 ng of template DNA, and nuclease-free water to a total volume of 20 μL. The PCR program consisted of an initial denaturing step at 94 °C for 5 min, followed by 35 cycles of denaturation at 94 °C for 20 s, annealing at 53 °C for 20 s and extension at 72 °C for 30 s, followed by a final extension at 72 °C for 5 min. The obtained amplicons were visualized using agarose gel electrophoresis (1.5%) stained with SYBR Safe (Thermo Fischer Scientific®, Wilmington, MA, USA). Bands with the proper size were cut out using a sterile scalpel, and purified with Illustra GFX PCR DNA and Gel Band Purification Kit (GE Healthcare®, Cardiff, UK), according to the manufacturer's instructions.
The purified amplicons were quantified by a Qubit® 2.0 spectrofluorometer using the dsDNA BR Assay Kit (Invitrogen™, Carlsbad, CA, USA). Subsequently, 1 ng of purified amplicons was indexed for sequencing using the Nextera XT® Index Kit (Illumina®, San Diego, CA, USA) according to the manufacturer's instructions. Finally, the libraries were quantified by Kapa Library Quantification Kit Illumina® Platforms (Kapa Biosystems®, Boston, MA, USA). All samples were pooled at equimolar concentrations and validated by Kapa Library Quantification Kit Illumina® Platforms (Kapa Biosystems®, Boston, MA, USA) to a final concentration of 1.8 pM. The library pool was sequenced using the NextSeq 500 V2 reagent Kit (2 × 150 bp paired-end) on an Illumina NextSeq 500 platform.
Data processing and analysis
Obtained forward read sequences were quality checked using Trimmomatic (v0.32) (Bolger et al. 2014). Subsequently, the reads were formatted for use with the UPARSE pipeline and screened for chimeric sequences (Edgar 2013). USEARCH7 was used to de-replicate and cluster the reads into operational taxonomic unit (OTU) at 97% sequence similarity. Taxonomy was assigned using RDP as implemented in QIIME, while applying MiDAS (version 2.1.3) (McIlroy et al. 2015) as the reference database.
The processed amplicon data was analysed using R (3.5.3) (R Development Core Team 2020) in RStudio (1.1.463) (http://www.rstudio.com) using the R package ampvis2 (Andersen et al. 2018).
Data availability
The datasets generated for this study can be found in the European Nucleotide Archive (Accession Number: PRJEB28160).
Analysis of MCs composition in freshwater and biomass by HPLC-PDA
For each sample, particulate matter in 1 L of water was collected by filtration through individual Whatman GF/C filters (Whatman™, Maidstone, UK). Each filter was frozen and thawed three times and extracted in 2.0 mL 75% methanol (MeOH). The extracts were incubated in an ultrasonic bath for 15 min, sonicated for 2 min and centrifuged at 10,000 × g for 10 min. Subsequently, the supernatant (1 mL) was evaporated at 50 °C and dissolved in 0.3 mL 75% MeOH. Before the HPLC analysis, the samples were re-filtered through 0.2 μm Teflon syringe filters into HPLC vials with inserts (Meriluoto & Codd 2005).
Dissolved MCs in the filtrate from the GF/C filtration were extracted by passing 500 mL of water through solid-phase extraction cartridges (Speed SPE Cartridges C18 Octadecyl/22, 12 mL, Applied Separations, Allentown, PA, USA). MCs were eluted from the column using 4 mL of acetonitrile with 0.05% trifluoroacetic acid (TFA). The eluate was evaporated at 50 °C and suspended in 500 μL of 75% MeOH. The extract was centrifuged at 10,000 × g for 10 min, and the supernatant was re-filtered through 0.2 μm Teflon syringe filters into vials before HPLC analysis (Meriluoto & Codd 2005).
The HPLC system consisted of a Shimadzu (Japan) Nexera-XR system equipped with a photodiode array (PDA) detector and a column (Shimadzu, 5 μm, C18, 250*4.6 mm) with an attached pre-column. The mobile phases were 0.05% aqueous TFA (solvent A) and 0.05% TFA in acetonitrile (solvent B). The gradient program was 0 min: 30% B; 10 min: 35% B; 40 min: 70% B; 42 min: 100% B; 46 min: 30% B; 60 min: end of analysis. The flow rate was 0.75 mL/min and the injection volume was 10 μL per sample (Meriluoto & Codd 2005). The HPLC analysis was calibrated with individual MCs standards of the variants MC-LR, MC-RR and MC-YR (DHI Lab Products, Hørsholm, Denmark), and toxins were quantified in biomass and in dissolved water at the 238-nm peak. MCs were detected by their retention times and absorption spectra by photodiode array detection operated between 200 and 300 nm. The detection limit of the analysis was approximately 1 ng of MC per injection.
Real-time PCR for detection of the mcyE gene
The mcyE gene was detected by a quantitative PCR (qPCR) protocol that included a reaction mixture (25 μL) with 4 μL of template DNA, 0.5 μM of each primer (10 pmol/μL), 12.5 μL of GoTaq qPCR Mastermix 2X (Promega®, Southampton, UK) and 7.5 μL of nuclease-free water. For relative quantification of the mcyE gene, the forward primer 5′-AAGCAAACTGCTCCCGGTATC-3′ and the reverse primer 5′-CAATGGGAGCATAACGAGTCAA-3′ were expected to yield a 120-bp product (Sipari et al. 2010). The PCR amplification was conducted in a 7500 Real-Time PCR System (Applied Biosystems®, Foster City, CA, USA) using the thermal cycler settings described elsewhere (Qiu et al. 2013).
For confirmation of the expected PCR products, four amplified samples were sequenced by the Sanger method. The 120-bp amplicons were purified using an Illustra Microspin™ S-400 HR Columns Kit (GE Healthcare®, Cardiff, UK) according to the manufacturer's instructions. For this purpose, the purified amplicon was sequenced in both directions using the BigDye™ Terminator Cycle Sequencing Kit (Applied Biosystems®, Foster City, CA, USA) on an Applied Biosystems capillary 3500 Genetic Analyzer. The quality of the electropherograms was assessed using Sequencing Analysis version 5.4 (Applied Biosystems®, Foster City, CA, USA). Finally, sequences were identified by the Basic Local Alignment Search Tool (BLAST) algorithm.
RESULTS
Microbial community composition
Sequencing of the V6 to V8 region of the 16S rRNA gene produced a total of 916,351 high-quality sequencing reads across 46 samples, representing 2645 OTUs. The average number of reads was 19,921 ± 13,411 per sample. A minimum of 9,000 sequences were selected from a rarefaction curve (Supplementary Figure S1) for statistical analysis.
The most abundant members of the microbial community identified at the sampled locations are visualized in Figure 2. The highest overall estimated microbial richness (Chao1 index) was observed at location 1 with up to 1,300 OTUs per sample (Figure 2(a)). A lower richness was found inside the fish farms relative to upstream and downstream samples at locations 2, 4 and 5. Significant differences in measured richness were observed between location 1 and 2 (p = 0.035) and locations 1 and 6 (p = 0.035), as tested by the Wilcoxon rank-sum test (testing performed on all samples containing >9,000 reads). Overall, no significant difference (p > 0.05) was observed among the upstream, inside and downstream samples across locations. Similarly, no significant changes were found in the total amount of DNA in samples collected upstream, within the fish farm and downstream of the fish farm (p > 0.05), and we, therefore, compare the relative sequence abundance as an indication of real biomass changes.
Abundant microbial community composition in water upstream, inside and downstream of six fish farms. The estimated microbial richness (a) is shown as boxplots, and the distribution of the 25 most abundant genera (b) is shown as stacked bar plots. The samples from 3U and 6F had <9,000 reads but were included to show the overall trend in microbial community composition. U, upstream; F, inside fish farm; D, downstream. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wh.2020.089.
Abundant microbial community composition in water upstream, inside and downstream of six fish farms. The estimated microbial richness (a) is shown as boxplots, and the distribution of the 25 most abundant genera (b) is shown as stacked bar plots. The samples from 3U and 6F had <9,000 reads but were included to show the overall trend in microbial community composition. U, upstream; F, inside fish farm; D, downstream. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wh.2020.089.
The majority of the identified microorganisms could be assigned to the phyla Proteobacteria, Bacteroidetes, Cyanobacteria and Actinobacteria (Supplementary Figure S2). Acidobacteria, Chloroflexi and Fusobacteria had minor contributions to the microbial communities. The most abundantly observed organisms across all farms included the genera Synechococcus and Microcystis, uncharacterized Actinobacteria (the hgcl clade and the CL500-29 marine group), and the families Acetobacteraceae and Sporichthyaceae (Figure 2(b)). The other abundant genera were Flavobacterium at locations 2 and 3 and Acinetobacter inside fish farms 4, 5, 6 and downstream 6.
The phylum Cyanobacteria comprised 0.3–54.1% of the obtained sequences at the sampled locations and made up 23.6% of the total reads. A more abundant and diverse representation of Cyanobacteria was observed at locations 4 and 5, compared to the other sampled locations (Supplementary Figure S3). An increase in the relative abundance of Cyanobacteria inside the fish farms was only observed in farms 3 and 4. At the other locations (1, 2, 5 and 6), a lower abundance of Cyanobacteria was found inside the farms, as compared to in the up- and downstream water. The most abundant genus was Synechococcus, which was the single most common cyanobacterial genus at locations 1, 5 and 6. Unclassified Cyanobacteria were observed abundantly at locations 3 and 4 and downstream location 5, and were generally most abundant downstream than upstream. The genera Microcystis and Cyanobacteria Family I had small but ubiquitous contributions to the cyanobacterial diversity. At all locations (except 1), a higher abundance of Cyanobacteria was found downstream than inside the farms.
Upon ordination of the data by Principal Component Analysis (PCA), three distinct clusters of sampling points were found (Figure 3). The cluster including farms and downstream water at locations 3, 4 and 5 clustered together with OTUs representative of various Cyanobacteria, such as the filamentous species Limnothrix. The largest cluster, including samples from locations 1, 5 and 6, grouped together with an OTU representative of Synechococcus. OTUs representing the family Flavobacteraceae and the genus Rheinheimera clustered together with the remaining points, primarily stemming from location 2.
Principal component analysis on square-root transformed OTU abundance counts. Symbols are shaped and coloured by sampling location (upstream = green triangles, farm = blue squares and downstream = red circles) and labelled with their location of origin. The size of the symbols represents the total concentration of microcystins. The 10 most loaded OTUs are shown on the plot with genus level or otherwise highest possible taxonomic classification possible. Organisms that are not bacteria are indicated with an asterisk (*). U, upstream; F, inside fish farm; D, downstream. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wh.2020.089.
Principal component analysis on square-root transformed OTU abundance counts. Symbols are shaped and coloured by sampling location (upstream = green triangles, farm = blue squares and downstream = red circles) and labelled with their location of origin. The size of the symbols represents the total concentration of microcystins. The 10 most loaded OTUs are shown on the plot with genus level or otherwise highest possible taxonomic classification possible. Organisms that are not bacteria are indicated with an asterisk (*). U, upstream; F, inside fish farm; D, downstream. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wh.2020.089.
Microcystin abundance and correlation to cyanobacterial populations
Dissolved MCs were not detected in the water, indicating that the concentrations were below the analytical detection limit of 0.02 μg/L. In contrast, MCs were detected in particulate biomass from all the fish farms and upstream and downstream the reservoirs, except inside fish farm 6 (Table 1 and Figure 4). The mean concentrations of individual MC variants ranged from 0.036 to 0.931 μg/L. The MC-LR was the most common type of MC, except at location 2, where MC-YR was dominant. The highest total concentrations were measured at locations in the Nova Avanhandava reservoir upstream farm 3 (1.59 μg/L), inside farm 3 (1.38 μg/L) and in farm 4 (0.90 μg/L). At all locations, the MCs concentrations were higher upstream than inside and downstream of the fish farms. At location 3, 4 and 5, the MCs concentrations were lower at the downstream sampling points than inside the farms.
Microcystin concentration and composition in water upstream, inside and downstream of six fish farms
Location . | Microcystins mean ± standard deviation (μg/L) . | |||
---|---|---|---|---|
MC-LR . | MC-RR . | MC-YR . | MCs total . | |
1U | 0.036 ± 0.002 | 0.100 ± 0.017 | – | 0.137 |
1F1 | 0.037 ± 0.020 | 0.065 ± 0.006 | – | 0.102 |
1F2 | 0.078 ± 0.000 | – | – | 0.078 |
1D | 0.041 ± 0.003 | 0.083 ± 0.001 | – | 0.125 |
2U | 0.070 ± 0.002 | 0.092 ± 0.008 | 0.205 ± 0.009 | 0.368 |
2F | 0.069 ± 0.003 | 0.059 ± 0.020 | 0.235 ± 0.007 | 0.362 |
2D | 0.064 ± 0.002 | 0.088 ± 0.051 | 0.214 ± 0.004 | 0.365 |
3U | 0.931 ± 0.029 | 0.452 ± 0.093 | 0.209 ± 0.052 | 1,591 |
3F | 0.776 ± 0.002 | 0.422 ± 0.020 | 0.183 ± 0.020 | 1,381 |
4F | 0.491 ± 0.072 | 0.289 ± 0.031 | 0.123 ± 0.020 | 0.903 |
4D | 0.265 ± 0.020 | 0.144 ± 0.007 | 0.091 ± 0.002 | 0.5 |
5U | 0.224 ± 0.084 | 0.102 ± 0.026 | – | 0.326 |
5F | 0.184 ± 0.007 | – | – | 0.184 |
5D | 0.057 ± 0.004 | 0.051 ± 0.005 | – | 0.108 |
6U | 0.180 ± 0.020 | – | – | 0.18 |
6F | – | – | – | – |
6D | 0.066 ± 0.015 | – | – | 0.066 |
Location . | Microcystins mean ± standard deviation (μg/L) . | |||
---|---|---|---|---|
MC-LR . | MC-RR . | MC-YR . | MCs total . | |
1U | 0.036 ± 0.002 | 0.100 ± 0.017 | – | 0.137 |
1F1 | 0.037 ± 0.020 | 0.065 ± 0.006 | – | 0.102 |
1F2 | 0.078 ± 0.000 | – | – | 0.078 |
1D | 0.041 ± 0.003 | 0.083 ± 0.001 | – | 0.125 |
2U | 0.070 ± 0.002 | 0.092 ± 0.008 | 0.205 ± 0.009 | 0.368 |
2F | 0.069 ± 0.003 | 0.059 ± 0.020 | 0.235 ± 0.007 | 0.362 |
2D | 0.064 ± 0.002 | 0.088 ± 0.051 | 0.214 ± 0.004 | 0.365 |
3U | 0.931 ± 0.029 | 0.452 ± 0.093 | 0.209 ± 0.052 | 1,591 |
3F | 0.776 ± 0.002 | 0.422 ± 0.020 | 0.183 ± 0.020 | 1,381 |
4F | 0.491 ± 0.072 | 0.289 ± 0.031 | 0.123 ± 0.020 | 0.903 |
4D | 0.265 ± 0.020 | 0.144 ± 0.007 | 0.091 ± 0.002 | 0.5 |
5U | 0.224 ± 0.084 | 0.102 ± 0.026 | – | 0.326 |
5F | 0.184 ± 0.007 | – | – | 0.184 |
5D | 0.057 ± 0.004 | 0.051 ± 0.005 | – | 0.108 |
6U | 0.180 ± 0.020 | – | – | 0.18 |
6F | – | – | – | – |
6D | 0.066 ± 0.015 | – | – | 0.066 |
− denotes the below detection limit.
U, upstream; F, inside fish farm; D, downstream.
Microcystin concentration and composition in water upstream, inside and downstream of six fish farms. Symbols are ordered by location, coloured by type of microcystin and shaped by sampling location. Upstream = circles, Farm = triangles, Downstream = squares. Fish farms 3 and 4 share upstream and downstream water. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wh.2020.089.
Microcystin concentration and composition in water upstream, inside and downstream of six fish farms. Symbols are ordered by location, coloured by type of microcystin and shaped by sampling location. Upstream = circles, Farm = triangles, Downstream = squares. Fish farms 3 and 4 share upstream and downstream water. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wh.2020.089.
Locations with an elevated occurrence of MCs (Figure 4) appeared to coincide with the more diverse representation and higher abundance of Cyanobacteria observed at location 3 and 4. To test a possible correlation between measured MCs concentrations and cyanobacterial representation, a Spearman's correlation matrix was generated (Supplementary Figure S4). Microcystis showed a weak to moderate correlation to the measured MCs concentrations (ρ = 0.25–0.44), while a strong negative correlation was observed between Synechococcus and the detected MCs (correlation coefficient ρ of −0.66 for the total MC concentrations). In general, positive correlations were observed among most of the individual cyanobacterial populations, except for Synechococcus, which had a negative correlation to both MC concentrations and most other groups of Cyanobacteria. Comparison of the downstream values for cyanobacterial abundance, mcyE gene copies and microcystin concentrations, relative to the upstream values (Figure 5), showed that the abundance of Microcystis co-varied with the increase in mcyE gene copies and microcystin concentrations, except for MC concentrations at location 6. None of the other detected cyanobacterial groups approximated this trend, and in many cases, they had an opposite relationship, especially in farms 3, 4 and 5.
Relationship between Cyanobacteria abundance (a), mcyE gene detection (b) and microcystin concentrations (c). Abundances are shown as heatmaps of values measured in downstream samples and related to upstream values. The samples from 3U and 6F had <9,000 reads but were included to show the overall trend in microbial community composition. U, upstream; F, inside fish farm; D, downstream.
Relationship between Cyanobacteria abundance (a), mcyE gene detection (b) and microcystin concentrations (c). Abundances are shown as heatmaps of values measured in downstream samples and related to upstream values. The samples from 3U and 6F had <9,000 reads but were included to show the overall trend in microbial community composition. U, upstream; F, inside fish farm; D, downstream.
Detection of the mcyE gene
The relative distribution of the MC-specific mcyE gene was detected by quantitative PCR at all locations by estimating the cycle threshold (Ct) values (Supplementary Table S1). The lowest Ct values ranged between 21.42 and 31.72 indicate that the highest abundance of the mcyE gene copies was found in fish farms and downstream of the farms at locations 3 and 4, as well as at location 6. Agarose gel electrophoresis of the amplified products showed a single band of the expected size and sequencing of this showed a 120-bp amplicon with 100% identity to the mcyE gene (Sequence ID: HM854743.1). No primer dimers or non-specific amplification products were observed.
DISCUSSION
This study of microbial communities with a focus on Cyanobacteria, microcystin content and mcyE gene copy levels across six Brazilian fish farms, as well as their surrounding waters, provides new insight into the relationship between these variables and their relevance for understanding biological effects of CyanoHABs.
Water microbiome
The microbial richness in this study was found to vary between 472 and 1248 OTUs, which is in agreement with studies from similar ecosystems. In marine aquaculture facilities with recirculation, studies showed microbial richness ranging from 229 to 911 OTUs (Huang et al. 2016) and 343 to 972 OTUs within a marine fish farm in Norway (Karlsen et al. 2017). Species richness is important for a given microbial ecosystem since the richness has been shown to have a direct impact on functioning, stability and resilience of the ecosystem (Downing & Leibold 2010). Maintaining a high microbial community richness has also been suggested as a measure to improve pathogen control in aquaculture systems (De Schryver & Vadstein 2014). In the present fish farms, the microbial richness showed no significantly different tendencies among water samples collected before and after passage through the fish farms. However, a significantly lower richness was found inside the fish farms, relative to upstream and downstream the farms, at locations 2, 4 and 5. This may indicate an in-farm impact from the fish production on the microbial community structure as has been observed in other aquaculture studies (De Schryver & Vadstein 2014). Likely drivers in reduced richness in fish farms are increased nutrient levels in the water due to undigested fish feed, faeces and fish debris (Martínez-Córdova et al. 2009). The higher nutrients levels may favour the growth of heterotrophic bacteria that are well-adapted to enriched nutrient levels.
The microbial community analysis showed that the most abundant phyla were represented by Proteobacteria, Bacteroidetes, Cyanobacteria and Actinobacteria. Previous studies have shown a similar dominance of these phyla in a freshwater aquaculture system (Gomes et al. 2019). In our study, the phylum Cyanobacteria was prevalent and comprised 0.3–54.1% of the obtained sequences across the sampled locations. This is a more variable abundance than observed in a marine aquaculture facility in which Cyanobacteria comprised 18–26% of total reads (Porchas-Cornejo et al. 2017). A high abundance of Cyanobacteria may not always be characteristic for fish farms, as shown for high-productive, integrated fish ponds in China, where Cyanobacteria in some ponds only made up 5% of the total reads (Klase et al. 2019).
In the present study, the highest abundance of Cyanobacteria was found at downstream locations 4 and 5 but also at upstream location 5. A greater abundance of Cyanobacteria in downstream than inside the farms was observed in all locations (except location 1). The downstream effect may reflect the accumulation of N and P from degradation of organic matter or direct input of inorganic nutrients from the fish (Martínez-Córdova et al. 2009). At the other locations (1, 2, 5 and 6), a lower abundance of Cyanobacteria was found inside the farms, as compared to the up- and downstream water. The lower fraction of Cyanobacteria inside the fish farms could potentially be due to grazing by tilapia (Oreochromis niloticus) in the cages since tilapia is capable of filtering phytoplankton from the water (Salazar Torres et al. 2016). However, it appears more likely that water currents through the farms contributed nutrients to the downstream water, leading to a downstream production of Cyanobacteria.
Across the sampled fish farms, the most abundant non-cyanobacterial genera included Flavobacterium, and representatives of the uncharacterized Actinobacteria groups (the clades hgcl and CL500-29), as well as the family Sporichthyaceae. The presence of these microbial groups is in line with findings from previous studies of tilapia farms (Lukassen et al. 2019). The genus Flavobacterium, which was found in relative abundant numbers in fish farms 2 and 3, contains potentially pathogenic species, e.g. the opportunistic pathogen Flavobacterium columnare that is responsible for columnaris disease in freshwater fish (Declercq et al. 2013).
Diversity of Cyanobacteria and toxin production
Within the Cyanobacteria populations, the genera Synechococcus and Microcystis were the most abundant and ubiquitous across all locations. These two genera have been widely observed as common colonizers of freshwater aquaculture systems (Klase et al. 2019; Lukassen et al. 2019). Synechococcus is a single-celled or microcolony-forming genus found ubiquitously in freshwater (Callieri et al. 2016), and environmental concern regarding this genus includes production of MCs (Carmichael & Li 2006). Species of Microcystis have previously been shown to form pervasive blooms and can produce MCs (Harke et al. 2016). Studies based on molecular biological methods showed that Synechococcus and Microcystis could coexist ubiquitously in many water bodies (Huo et al. 2018; Tan et al. 2019). However, Synechococcus has a cellular advantage by virtue of smaller size and larger specific surface area for nutrient acquisition compared to Microcystis (Tan et al. 2019).
Other identified Cyanobacteria were the potentially toxin-producing genera Cylindrospermopsis, Snowella, Nostoc, Limnothrix and Planktothrix. M. aeruginosa and C. raciborskii are the most widespread toxic Cyanobacteria in Brazil, occurring in both tropical and subtropical areas (Sant'Anna et al. 2008). Species within the genera Snowella and Limnothrix are also known as MCs producers (Falconer 2005; Furtado et al. 2009), while Nostoc and Planktothrix are able to produce both MCs and other cyanotoxins (Nishizawa et al. 2000; Gaget et al. 2017). Although an increase in abundance and diversity of Cyanobacteria was observed in two of the fish farms (farms 4 and 5), a rather similar community structure was found among Cyanobacteria across the six locations. Thus, changes in the toxin production in the farms most likely reflect different physical and environmental conditions in specific farms (Beversdorf et al. 2015), rather than induced changes of the cyanobacterial communities. Nutrient data of the water in the fish farms, measured during 2014–2016, have revealed that total N varied seasonally were 380–1,440 μg/L upstream, 700–2,340 μg/L within farms and 780–1,920 μg/L downstream of the farms (data not published, Gianmarco S. David). For total P, concentrations were 17–27 μg/L upstream, 24–41 μg/L in farms and 19–40 μg/L downstream of the farms. Thus, the most pronounced effect from the fish production was seen in total N, while a minor increase in total P occurred. Possibly, the stimulated biomass of Cyanobacteria in some samples might be linked to higher concentrations of N, but other and not-accounted for factors might also have contributed to the enhanced cyanobacterial biomass.
Microcystins studied in the farm locations included the variants MC-LR, MC-RR and MC-YR, which are the most commonly detected (Spoof et al. 2010). The total MC concentrations ranged from 0.066 to 1.59 μg/L and were highest (>1 μg/L) in upstream water and inside fish farm 3, while farm 4 had a concentration just below 1 μg/L. The concentration range agrees with concentrations of 0.06–1.68 μg/L of total MCs measured in shallow eutrophic lakes in eastern Canada (Monchamp et al. 2014) but was lower than concentrations exceeding 5 μg/L in hypertrophic ponds in India (Singh et al. 2015) and in a Turkish lake (Gurbuz et al. 2016).
The main concern from the presence of MCs in freshwaters with fish production is the accumulation in various organs of the fish, causing mortality and/or risk to consumers eating the fish, as it was presented in the ‘Introduction’ section. MCs remain inside actively growing cells of Cyanobacteria and are only released into the water upon senescence or induced lysis (Edwards & Lawton 2009). Low levels of MCs in the water may not only reflect a reduced production but may indicate bacterial degradation, which appears to be the main fate for most cyanotoxins in aquatic systems (Edwards & Lawton 2009). Yet, relatively high levels may persist for several weeks, posing a serious health hazard to fish and consumers eating the fish, and challenge the public water utility management.
The present Brazilian reservoirs with tilapia cage breeding also serve as a source of public drinking water. The international guideline value for concentrations of MC-LR (dissolved and cell-bound) in drinking water is 1 μg/L, as recommended by WHO (WHO 1998). In Brazil, a similar guideline value for drinking water of 1 μg/L is standard, but it includes the total MCs concentration, not individual MCs (BRASIL 2017). This threshold means that water upstream and inside farm 3 had unacceptable concentrations of MCS for drinking water purposes.
Concentrations of MCs were higher in the upstream waters at all locations, except at location 2, where similar MC concentrations occurred up- and downstream and inside the farm. The higher concentrations could potentially be explained by a greater relative abundance of Synechococcus or Microcystis in upstream samples, as both genera are known MC producers (Callieri et al. 2016; Harke et al. 2016). Moderate to strong positive correlations were observed between all individually measured MCs types and the cyanobacterial groupings Caenarcaniphilales, Cylindrospermosis, Nostoc and Planktothrix (Supplementary Figure S4), although only the latter two genera are known MCs producers (Nishizawa et al. 2000; Gaget et al. 2017). Interestingly, Microcystis showed a weak to moderate correlation to the measured MC concentrations, while a strong, negative correlation was observed between Synechococcus and the detected MCs. If cyanotoxin production largely is controlled by different physical and environmental parameters, as mentioned above (Beversdorf et al. 2015), it may not be possible to find a direct universal link between cyanobacterial taxa and microcystin concentrations in fish farms.
Relationship between Cyanobacteria, mcyE gene and microcystin concentrations
Detection of the mcyE gene has been used as a proxy for the presence of toxic Cyanobacteria in freshwater, e.g. in lakes in Finland (Rantala et al. 2006) and in lakes in the United States (Francy et al. 2016). In addition to mcyE, the mcyD gene has been found to correlate positively with concentrations of MCs and serve as a better indicator of blooms of toxic Cyanobacteria than chlorophyll content and cell counts (Davis et al. 2009). In the present study, although the mcyE gene was detected at all sampled locations, no immediate relationships between Cyanobacteria, mcyE gene copies and microcystin concentrations were evident. However, an exception was the presence of the genus Microcystis, which followed the trend of mcyE and microcystins in 5 of 6 farms (Figure 5). The usage of mcyE and mcyA genes as proxies for microcystin production has been contradicted in a study of lakes in Wisconsin, USA, which concluded that the presence of MC-specific genes is not a useful indicator of toxins in the environment since the production of microcystins likely is governed by multiple environmental factors (Beversdorf et al. 2015). The present study showed that abundance of Microcystis correlated well with the detected mcyE gene copies and the observed microcystin concentrations, while all other detected cyanobacterial groups did not, and were in many cases inversely correlated to the microcystin and qPCR results. This contradictory observation, as well as the study by Beversdorf et al. (2015), shows that more studies are needed to provide better information on the relationship between the presence of Cyanobacteria and cyanotoxin production in the environment. The data analysed in this study represent a snapshot in time and does not provide insight into seasonal variations. However, given the nature of sample strategy, we believe that our observations reflect general impacts of fish production on water quality and growth of Cyanobacteria.
CONCLUSION
The present study aimed at studying possible environmental effects from tilapia aquaculture production in cages through studies of the water microbiome, Cyanobacteria diversity and microcystin concentrations upstream, inside and downstream six fish farms. Fish breeding reduced the microbial richness in two of the farms, relative to upstream and downstream samples, indicating an impact of fish farming on the microbial community structure. An increase in abundance and diversity of Cyanobacteria inside the fish farms was only observed in two of the farms, suggesting that the fish breeding did not have a universal effect on the cyanobacterial community structure in the tilapia farms. However, the fish breeding contributed to a stimulated production of Cyanobacteria downstream the farms at five of the six locations. The increased abundance of Cyanobacteria in two farms coincided with higher MCs concentrations. The potential MC producer Microcystis showed similar trends in abundance to mcyE gene copies and MCs concentrations at five of the six locations, while Synechococcus and other cyanobacterial groups correlated negatively to occurrence of the MCs. The present study supports the hypothesis that molecular approaches can potentially serve as tools for detecting and studying CyanoHABs and cyanotoxin production in aquaculture systems. However, the mixed results also highlight that many aspects of biology of Cyanobacteria, their toxin production and dynamics in aquaculture systems require further studies.
ACKNOWLEDGEMENTS
This work was supported by the São Paulo Research Foundation (FAPESP) grant numbers 2015/13025-7, 2014/13718-0 and 2013/50504-5 and the Danish Council for Strategic Research grant number 3050-00008A.
CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.
COMPLIANCE WITH ETHICAL STANDARDS
This article does not contain any studies with animals performed by any of the authors.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.