Aquaponics using a fish farm effluent shifts bacterial communities profile in halophytes rhizosphere and endosphere

The intensification of marine aquaculture raises multiple sustainability issues, namely the handling of nutrient-rich effluents that can adversely impact ecosystems. As integrated multi-trophic aquaculture (IMTA) gains momentum, the use of halophyte plants to phytoremediate aquaculture effluents has received growing attention, particularly in aquaponics. It is, therefore, important to obtain a more in-depth knowledge of the microbial communities present in the root systems of these plants, both in their natural environment (sediment) and in aquaponics, in order to understand their nutrient removal potential. The present study used denaturing gradient gel electrophoresis (DGGE) and barcoded pyrosequencing to assess the bacterial community present in the endosphere and rhizosphere of three halophyte plants: Halimione portulacoides, Salicornia ramosissima and Sarcocornia perennis. Species-specific effects were recorded in the profile and diversity of the bacterial communities present in halophyte roots, with significant differences also recorded for the same halophyte species grown in contrasting environments (sediment vs. aquaponics). In aquaponics the most abundant groups belonged to the orders Rhodocyclales, Campylobacterales, Rhodobacterales and Desulfobacterales, while in the natural environment (sediment) the most abundant groups belonged to the orders Rhizobiales, Sphingomonadales and Alteromonadales. An overall enrichment in bacterial taxa involved in nutrient cycling was recorded in the roots of halophytes grown in aquaponics (such as Denitromonas, Mesorhizobium, Colwellia, Dokdonella and Arcobacter), thereby highlighting their potential to reduce the nutrient loads from aquaculture effluents.

www.nature.com/scientificreports www.nature.com/scientificreports/ However, in recent years, a group of macrophytes, namely, salt-tolerant or halophyte plants, has received growing attention in the bioremediation of aquaculture effluents generated by facilities operating in brackish water. The potential use of halophytes as biofilters for marine and brackish water aquaculture has already been highlighted [13][14][15][16] , namely within an IMTA framework 17 . Moreover, the ability to successfully bioremediate brackish water aquaculture effluents through the production of halophytes in aquaponics and add value to these crops has been recently demonstrated 18 .
Apart from incorporating inorganic dissolved nutrients, plant roots act as a trap for suspended solids and are important for microbial nutrient transformation processes 19 . Nonetheless, knowledge of the overall diversity and profile of bacterial communities present in the rhizosphere and endosphere of halophytes used for the bioremediation of aquaculture effluents has only recently started to be unraveled, although so far has been limited to the genus Salicornia 20 . According to Xiaona et al. 20 , in a study acessing the microbial communities in pilot-scale constructed wetlands with Salicornia for treatment of marine aquaculture effluents, the most common groups present in the root systems are: Proteobacteria, Bacteriodetes, Cyanobacteria and Firmicutes.
In the present study we compared the bacterial communities in the rhizosphere and endosphere of three halophyte species grown in aquaponics and in their natural environment. The halophyte species selected were sea purslane Halimione portulacoides (L.) Aellen (also known as Atriplex portulacoides L.), perennial marsh samphire Sarcocornia perennis (Mill.) A. J. Scott and purple glasswort or marsh samphire Salicornia ramosissima J. Woods. The rationale supporting the selection of these three species is their recognized potential to grow in aquaponics and bioremediate the effluents from marine fish farms 17,18 .
The following two null hypotheses were tested: (1) bacterial communities present in the rhizosphere or endosphere do not differ between halophyte species grown in the same environment (sediment or aquaponics); and (2) bacterial communities present in the rhizosphere or endosphere do not differ if the same halophyte species is grown in contrasting environments (sediment vs. aquaponics).

Material and Methods
Experimental design. A total of 12 tanks (6 m long × 1 m wide × 0.3 m deep) were used to perform phytoremediation trials of the effluent water generated by a super-intensive marine fish farm producing Senegalese sole (Solea senegalensis) at Torreira, Aveiro, Portugal (40°47′40.59"N, 8°41′47.69"W). Tanks were assembled in parallel, with each one being supplied effluent water previously filtered by a polychaete assisted sand filter (1 m long × 1 m wide × 0.7 m deep) stocked with ragworms (Hediste diversicolor) to remove excess organic particulate matter from the water. All sets of polychaete assisted sand filters and phytoremediation tanks were gravity fed by two interconnected header tanks (2 m long × 1 m wide × 0.7 m deep). The header tanks were stocked with bio-blocs where effluent water pumped from the fish farm settling basin was trickled and strongly aerated before being distributed to the 18 sets of polychaete assisted sand filters and phytoremediation tanks. Water inflow to each set was controlled manually through the use of ball-valves and set at 180 L h −1 of effluent water (thus allowing 2 full water renewals per day for each phytoremediation tank). A schematic representation of the culture system and further technical details can be found in Marques et al. 18 . The physico-chemical parameters of effluent water were as follows: salinity: 20.0 ± 1.0, water temperature: 19.7 ± 1.8 °C, pH: 7.9 ± 0.2, dissolved oxygen: 8.8 ± 0.7 mg L −1 , dissolved inorganic nitrogen (DIN = NO x -N + NH 4 -N): 8.9 ± 1.3 mg L −1 and dissolved inorganic phosphorus (DIP): 0.32 ± 0.11 mg L −1 . Each phytoremediation tank was stocked with 12 floating Styrofoam seed trays (0.9 m long and 0.6 m wide) with 100 pyramid-shaped cells, thus operating as a raft hydroponic tank. Every other cell of the tray was stocked with a halophyte plant transplanted from the wild, for a total of 800 halophytes per tank grown on aquaponics (for further details on plants and stocking conditions please see 18 ).
Halophytes sampling. For this study, due to tank dimensions and inter-variability regarding nutrient dynamics 18 , one single tank stocked with each of the halophyte species referred above was selected and composite samples covering the 6 m long × 1 m wide tank were collected. Halophytes were sampled from different floating Styrofoam trays within the selected H. portulacoides, S. perennis and S. ramosissima tank. After 5 months of culture in aquaponics, four composite samples per halophyte species were collected (4 composite samples × 3 halophyte species in aquaponics = 12 composite samples). Each composite sample was assembled by pooling 4 plants per tray within the selected phytoremediation tank. Additionally, 4 composite samples per halophyte species were also collected at the same location where halophytes were initially sampled to stock the phytoremediation tanks (4 composite samples × 3 halophyte species from their natural habitat = 12 composite samples). Each composite sample (for a total of 24 composite samples) was stored separately in sterile plastic bags and processed on the same day of sampling.
Rhizosphere and endosphere sampling. Roots were manually shaken to remove loosely bound sediment, which was discarded. The endophytic community was recovered from roots. Root material was vigorously washed in distilled water (5 min) and the root surface was disinfected by sequential washing with 95% ethanol and 1% sodium hypochlorite supplemented with one droplet of Tween 80 ® (Sigma-Aldrich) and rinsed three times using sterile distilled water.
16S rRNA gene denaturing gradient gel electrophoresis (DGGE) profiling of bacterial communities. A nested PCR approach was used to amplify 16S rRNA gene sequences from rhizosphere and endosphere samples 21 , as this approach is more efficient for amplification of 16S rRNA gene fragments suitable for DGGE analyses. Briefly, in the first PCR the universal bacterial primers 27 F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1494R 22 (5′-CTACGGRTACCTTGTTACGAC-3′) were used. www.nature.com/scientificreports www.nature.com/scientificreports/  PCR products containing approximately equal amounts of DNA (estimated based on band intensity detected in the agarose gel electrophoresis) were loaded onto 6-10% (w/v) polyacrylamide gel in 1 × TAE buffer (0.04 M Tris-Acetate, 0.001 M EDTA; pH 8.0). The 6-10% polyacrylamide gel was made with a denaturing gradient ranging from 40 to 58%. A DGGE marker was used for internal normalization and as an indication of the quality of the analysis. Electrophoresis was performed for 16 h at 80 V at 60 °C in 1 × TAE buffer. Following electrophoresis, gels were silver-stained according to Heuer et al. 23 . Scanned DGGE gels (Figs. S1 and S2) were processed using the Bionumerics software 6.6 (Applied Maths, Sint-Martens-Latem, Belgium). The matrix constructed by the program incorporated both band position and intensity of each band that was processed in a spreadsheet. The intensity of each DGGE band was then normalized by total sample intensity to obtain relative abundances.
Barcoded pyrosequencing. A barcoded pyrosequencing approach was used for compositional analysis of bacterial communities. Before pyrosequencing, TC-DNA of all four replicates per experimental treatment was combined, forming one DNA library for each. The V3-V4 region was amplified using barcoded fusion primer V3 Forward (5′ -ACTCCTACGGGAGGCAG-3′) and V4 Reverse (5′ -TACNVRRGTHTCTAATYC-3′) with Roche 454 titanium sequencing adapters (see Oliveira et al. 24 , for a detailed description). Sequences generated in this study can be downloaded from the NCBI SRA: SRP155695. A detailed description of barcoded pyrosequencing analysis using QIIME and UPARSE (http://qiime.org/; https://www.drive5.com/uparse/) can be found in Cleary www.nature.com/scientificreports www.nature.com/scientificreports/ et al. 25,26 The taxonomic affiliation of all bacterial OTUs was determined using Ribosomal Database Project (RDP) classifier.
Data analysis. Two square matrices were imported into R (version 3.1.2; http//www.r-project.org/) using the read.table() function: (1) containing band 'abundance' based on band intensity and position on the DGGE gel and (2) containing the presence and raw abundance of all operational taxonomic units (OTUs) per sample generated in the pyrosequencing analysis. The distribution of OTUs in samples was assessed using a Venn diagram with the venn() function in the gplots package 27 in R. In the OTU abundance matrix, sequences not classified as bacteria or classified as chloroplasts or mitochondria were removed prior to statistical analysis. Both matrices were log e (x + 1) transformed and a distance matrix was constructed using the Bray-Curtis index with the vegdist() function, in the VEGAN package 28 in R. Variation in composition was visualized with principal coordinates analysis (PCO) using the cmdscale() function in R with the Bray-Curtis distance matrix as input. Selected sequences from dominant OTUs (≥50) and their closest relatives were downloaded using the NCBI Basic Local Alignment Search Tool (BLAST) command line 'blastn' tool with the _db argument set to nt 29 . A heatmap was constructed to visualize the distribution of the dominant OTUs (≥50 sequence reads). The heatmap was generated using the heatmap2() function in the R package gplots. A two-way permutational multivariate analysis of variance (PERMANOVA) was performed to test for the existence of statistically significant differences between DGGE profiles performed for the rhizosphere and the endosphere of the halophytes grown in the different environments using PRIMER v6 with the PERMANOVA + add-on software. Halophyte species (H. portulacoides vs. Sar. perennis vs. Sal. ramosissima) and environment (sediment vs. aquaponics) were used as categorical factors.

Results and Discussion
Bacterial communities profiles. PCO ordination analysis of bacterial DGGE profiles showed distinct bacterial communities in halophyte growth environments (sediment vs. aquaponics) (Fig. 1). The first two PCO axes explained 48.9% and 40.0% of the compositional variation in rhizosphere data (PCO axis 1: 26.8%, PCO axis 2: 22.0%) and endosphere data (PCO axis 1: 23.4%, PCO axis 2: 16.7%), respectively. PERMANOVA analysis of bacterial DGGE profiles revealed a significant interaction (P = 0.0001) between plant species and the type of growth environment (sediment vs. aquaponics) (Table S1 and S2). Moreover plant species specific effects were observed on the profile of root bacterial communities (P = 0.0001). These results were somehow expected, as oxygen and nutrient availability, along with other factors, play a relevant role in wetland performance 13 and consequently may affect the composition of bacterial communities 30 . Salicornia ramosissima grown in its natural environment (sediment) displayed the most distinct microbial community in the endosphere and rhizosphere. Salicornia ramosissima plants were collected close to the area where effluent from the super-intensive marine fish farm is discharged into the environment, whereas samples of the other two species were collected in Ria de Aveiro channels solely subject to tidal nutrient inputs. This may explain the distinct composition differences among plant species. Our findings are in line with those described by several authors 30-33 that revealed that composition www.nature.com/scientificreports www.nature.com/scientificreports/ and temporal variation of root-inhabiting bacterial communities were associated with habitat characteristics and vegetation type, such as the influence of root exudates.
OTU compositional analysis. Barcoded pyrosequencing analysis yielded 12,924 sequences clustered into 1859 bacterial OTUs after quality control, OTU picking and removal of chimeras and sequences not assigned to the domain Bacteria (Table S3).
In line with DGGE profiles, the PCO ordination revealed compositional differences among rhizosphere and endosphere bacterial communities recorded in aquaponics and sediment (Fig. 2). The first PCO axis separated aquaponics samples from sediment samples. There were three dominant bacterial OTUs (OTUs 18, 29 and 61), which were almost equally distributed in both growth environments ( Fig. 2B and D) and were classified as belonging to the order Rhizobiales (Table S3). The distribution of most abundant OTUs (≥50 sequences) among halophyte species in the same growth environment showed that the largest component consisted of abundant OTUs that were found in all three plants (Fig. 3). The halophyte H. portulacoides was the plant where the highest number of exclusive OTUs was recorded, while in the S. perennis rhizosphere sediment there was only one restricted OTU (OTU 27, Fig. 3). The most abundant OTUs (≥50 sequences) were only recorded in a certain growth environment (Fig. 4), with exception of H. portulacoides rhizosphere and endosphere where the largest component consisted of abundant OTUs that were shared between sediment and aquaponics. The highest number of restricted OTUs was recorded in the rhizosphere and endosphere of plants grown in sediment (Table S4 to S9). OTUs were assigned to 33 phyla, 82 classes and 117 orders and only 13 OTUs remained unclassified at the phylum level (Table S3). The two most abundant phyla were Bacteroidetes and Proteobacteria. The majority of bacterial OTUs in all samples were assigned to Proteobacteria and all five major classes were detected, representing on average more than 82% of all sequences. The phylum Proteobacteria consists of a diverse group of bacteria recovered from different hosts and environments, and is an important player in nutrient cycling 34 . Bacteroidetes was the second most abundant www.nature.com/scientificreports www.nature.com/scientificreports/ phyla, with relative abundance ranging from 0.7% to 25%. Members of the Bacteroidetes were more abundant in sediment and are known degraders of plant polysaccharides 35 .

Phylotype-level 16S rRNA gene analysis. The in-depth bacterial composition analysis detected fifty
abundant OTUs (≥50 sequences, Figs. 5 and 6) that were assigned to three phyla: Bacteroidetes, Cyanobacteria and Proteobacteria. The taxonomic assignment of these abundant OTUs and their known ecophysiological traits are given in Table 1. The most abundant OTU overall was OTU 6, which was enriched in aquaponic samples, and was assigned to the family Rhodocyclaceae (Betaproteobacteria) and closely related to the genus Denitromonas that was isolated from a marine aquaculture (GI: 115334100, Table S10). Members of this genus have been associated with denitrifying processes in a biofilm developed in a recirculating marine aquaculture system aerated filter 36 . Only one abundant OTU (OTU 38), recorded in S. ramosissima aquaponic biotopes, was assigned to the phylum Cyanobacteria; this OTU was closely related to an Acaryochloris strain isolated from a red alga 37 . The fact that this OTU was only detected in Salicornia tank may be explained by higher water temperature recorded in this tank (23 °C) compared to the other tanks (21 °C). The availability of nutrient-rich effluents from aquaculture combined with higher water temperatures may have promoted the growth of cyanobacteria in Salicornia rhizosphere and endosphere 38,39 . Of the three abundant OTUs assigned to phylum Bacteroidetes, two were restricted to sediment samples from H. portulacoides and S. ramossisima (OTU 42 and 79) and were related to a Lewinella sp. strain isolated from seawater (GI: 530549991, Table S12). Members of this genus are aerobic chemo-organotrophic that require NaCl for growth 40 . The restricted presence of members of the phylum Bacteroidetes in the sediments samples could be explain by the fact that they are specialized in degrading complex carbon sources common found in this type of sediments 41 . Furthermore, they have shown to have a clear preference for growth attached to surfaces or particles (e.g. sediments) 41 . All the main proteobacterial classes were well represented in the present study (Alphaproteobacteria, Deltaproteobacteria, Epsilonproteobacteria, Gammaproteobacteria and Betaproteobacteria;

Saprospirales Lewinella
The Saprospiraceae family includes stains isolated from aquatic environments and members of this family have the ability for the hydrolysis and utilization of complex carbon sources. The genus Lewinella includes a few isolates retrieved from marine sources which were aerobic chemoorganotrophic that require NaCl to growth.

38/50
Acaryochloridaceae Acaryochloris Members of the Acaryochloris genus (Cyanobacteria) contain chlorophyll d as a major photopigment and are found widely distributed in both aquatic and terrestrial ecological systems.

Sphingomonadales Erythrobacteraceae
Members of Erythrobacteraceae family (Sphingomonadales) consist of aerobic chemoorganotrophs that are mainly isolated from aquatic environments, but there are also isolation reports from sediment, sand, and rice.

Devosia Hyphomicrobium
The family Hyphomicrobiaceae is phenotypically, metabolically, and ecologically diverse. Devosia genus comprises non-photosynthetic bacteria which have been mainly isolated from terrestrial soil sources. Members of Hyphomicrobium genus are anerobic facultative methylotrophs, reproduce by budding and are ubiquitous in water and soils, but can also be found in sewage treatment plants. Some strains are characterized by their denitrification capacities. Rhodobacteraceae are, fundamentally, aquatic bacteria that frequently thrive in marine environments, and are deeply involved in sulfur and carbon biogeochemical cycling. The Jannaschia genus are isolated from marine environments (seawater, seashore sand, tidal flat sediment and solar saltern water). The genus Loktanella have been isolated from Antarctic lakes and marine environments.
Marivita genus consist in six species, one of theme isolated from estuarine water, that contain genes for bacteriochlorophyll a synthesis (although some of them do not produce the pigment). Only two species of Octadecabacter are described and both isolated from polar regions. The genus Phaeobacter includes a few isolates, and are candidates to be used as probiotics in marine aquaculture because their ability to efficiently adhere and grow on surfaces and to secrete inhibitors compounds that enable them to antagonize invertebrate settlement and algal or microbial growth

Betaproteobacteria 6/283 Rhodocyclaceae
Members of Rhodocyclaceae family have been isolated from diverse environments: soil; sewage treatment plants; polluted and unpolluted waters of ponds, rivers, aquifers and plant roots. They are capable to degrade a wide range of carbon sources including many aromatic compounds and fix nitrogen in association with plants.

52/55
Methylophilaceae Methylotenera The family Methylophilaceae includes obligate or restricted facultative methylotrophs such as Methylotenera spp. This family belong to a group of methylobacteria, which play an important role for the aerobic conversion of C1 compounds in different ecological niches.  www.nature.com/scientificreports www.nature.com/scientificreports/  42 . Within the class Alphaproetobacteria, several dominant OTUs were closely related to members assigned to the Hyphomicrobiaceae family, which includes strains known to be involved in denitrification 43 . Another group of dominant OTUs (18,29,74, 248 and 418), assigned to family Phyllobacteriaceae, were enriched in the root endophytic community, and were closely related to Mesorhizobium sp. that was isolated from a root endophytic microbiome (GI: 725096748, Table S12) and an endophytic bacterium isolated from the roots of the salt marsh plant Spartina alterniflora 44 . The genus Mesorhizobium is known to include root-nodule bacteria that can establish nitrogen-fixing symbiosis with plants 45 . Gammaproteobacteria were the second most abundant proteobacterial class. The phylogenetic analysis showed that these OTUs were closely related to uncultured Gammaproteobacteria or species within the genera Colwellia or Dokdonella (SI Table S12). These genera include bacterial guilds often associated with nitrate reduction 46 (Table S12). In aquaponics, the order Campylobacterales was the most abundant detected, mainly in H. portulacoides and S. perennis samples and assigned to the genus Arcobacter. Members of this genus isolated from the roots of Spartina alterniflora, a salt marsh plant, have been shown to be capable of nitrogen fixation 48 . Rhodobacterales was the second most abundant order of Proteobacteria recorded in samples from aquaponics. Within this order, family Rhodobacteraceae was the most prevalent in this study and is characterized by highly diverse metabolisms that include sulfur-oxidizing species in freshwater and marine environments 49 . Desulfobacterales were markedly more abundant in the endophytic community of S. perennis in aquaponics. This order includes sulfate-reducing bacteria which are important players in anoxic mineralization of organic matter 50 . However, the reasons why members of this order was markedly more abundant in S. perennis endophytic community need further investigation.
Results from the present study showed that specific proteobacterial groups involved in nutrient cycling, namely, nitrogen fixation, denitrification and mineralization of organic matter, were enriched in root associated bacterial communities in aquaponics. The more prevalent bacterial groups in the roots of plants grown in sediment included the Rhizobiales and the Sphingomonadales. While the first is well-known for its ability to fix atmospheric nitrogen in association with plants, the second includes aerobic chemeorganotrophs isolated from diverse environments, such as seawater, tidal flats and marine sediment 51 . The endophytic community of S. ramosissima in the sediment was enriched with Alteromonadales, which are a relatively common marine group of chemoheterotrophs that are known to be important player in carbon and sulfur cycling 52 .

Concluding Remarks
In this study we applied a combined DGGE and pyrosequencing approach to assess the bacterial community present in the root system of three halophyte plants: Halimione portulacoides, Salicornia ramosissima and Sarcocornia perennis.
Overall, our study revealed the existence of a plant species-specific effect on the profile and diversity of bacterial communities present in the rhizosphere and endosphere. Moreover, significant differences were also revealed in the bacterial composition present in the rhizosphere and endosphere of conspecific plants grown in contrasting environments (aquaponics vs. sediment). The composition analysis performed showed an enrichment in the bacterial taxa of halophytes produced in aquaponics, namely those involved in nutrient cycling, such as sulfur, carbon cycles nitrogen fixation and denitrification processes.