Effects of finfish aquaculture on biogeochemistry and bacterial communities associated with sulfur cycles in highly sulfidic sediments

A combination of biogeochemical analyses and molecular microbiological analyses were conducted to assess the environmental impact of finfish aquaculture and to elucidate the major microbial assemblages responsible for the production and removal of reduced sulfur compounds in fish-farm sediments. The average concentrations of H2S (123 μM) and NH4 (1310 μM) and the dissimilatory sulfite reductase (dsr) gene copy number (1.9 × 109 copies cm−3) in the sediments at the farm site were 15-, 1.5and 2-fold higher, respectively, than those measured at the less-impacted reference site. Accordingly, the sulfate reduction rate (SRR) at the farm site (118 mmol m−2 d−1) was 19-fold higher than that measured at the reference site (6.2 mmol m−2 d−1). Analyses of dsrA and 16S rRNA gene sequences revealed that the Syntrophobacteraceae and Desulfobulbaceae groups are the major sulfate-reducing bacteria around the fish-farm sediment. Interestingly, despite the high SRR (12.2−19.6 mmol m−2 d−1), the H2S concentration was low (<8 μM) in the top 0−2 cm of the fish-farm sediments. In this sulfide-mismatched zone, sulfur-oxidizing bacteria associated with Gammaand Epsilonproteobacteria were abundant. Especially at the 1−2 cm depth, bacteria related to Sulfurovum in the Epsilonproteobacteria showed the highest relative abundance, comprising 62% of the 16S rDNA sequences. The results strongly suggest that Sulfurovum-like bacteria play a significant ecological and biogeochemical role in oxidation and reduction of reduced sulfur compounds from the organic-rich, highly sulfidic fish-farm sediments.


INTRODUCTION
The global production of aquaculture, including finfishes, crustaceans, mollusks and other aquatic animals, increased from 2.6 million t in 1970 to 66.6 million t in 2012 (FAO 2010(FAO , 2014)).Accordingly, the contribution of aquaculture to total global fisheries production doubled from 22.2% in 1996 to 42.2% in 2012 (FAO 2010(FAO , 2014)).Although aquaculture provides a stable, high-quality food source for humans, large-scale industrial aquaculture activities inevitably cause severe environmental concerns by releasing large quantities of organic materials, such as uneaten fish feed and fecal pellets, to the surrounding coastal areas (Gray et al. 2002, Brooks & Mahnken 2003, Holmer et al. 2005, Kutti et al. 2007).
In organic-rich coastal sediments, oxygen is depleted quickly within a few mm of the surface sediment, and organic carbon (C org ) oxidation is dominated by anaerobic microorganisms relying on different terminal electron-accepting processes, such as denitrification and reduction of Mn(IV), Fe(III) and sulfate (Canfield et al. 2005).Because of the abundance of sulfate in seawater, sulfate reduction is regarded as the predominant C org oxidation pathway in coastal sediments, especially in organic-enriched fish-farm sediments (Holmer et al. 2005, Hyun et al. 2013).Major environmental concerns regarding the dominance of C org oxidation resulting from sulfate reduction in fish-farm sediments include the accumulation of highly toxic, reactive hydrogen sulfide (H 2 S), which affects the diversity and community structure of the benthic macrofaunal and microbial communities (Lee et al. 2004, Bissett et al. 2006, Heilskov et al. 2006, Kutti et al. 2007, Castine et al. 2009, Kawahara et al. 2009, Yoon et al. 2009, Hargrave 2010, Valdemarsen et al. 2015, Jung et al. 2016).
Because large-scale aquaculture has such a large effect on sediment biogeochemistry and coastal ecosystems via benthic−pelagic coupling (Holmer et al. 2005, Strain & Hargrave 2005, Lee et al. 2011, Hyun et al. 2013), it is important to develop a tool to assess the environmental conditions of fish-farm sediments (MOF 2016).Total organic carbon content and concentrations of acid volatile sulfur compounds in the sediments have been proposed as chemical indicators for monitoring the environmental changes caused by fish farms and for evaluating farming capacity (Yokoyama 2003, Cho et al. 2013, Jung et al. 2016).However, the environmental assessment of fish-farm sediments relying solely on sulfur compounds is not useful for identifying major biogeochemical processes because these compounds are quickly recycled within the sediments (Canfield et al. 2005, Hyun et al. 2013, NIFS 2013).Additionally, biological indicators have been proposed for monitoring the environmental changes resulting from aquaculture.For example, changes of benthic macrofauna assemblages are generally recommended as a proxy for monitoring the impacts of cage farming (Macleod et al. 2008, Yoon et al. 2009, Jung et al. 2016).However, investigation of the benthic community using video surveillance and the identification of macrofauna and meiofauna to the level of individual species is expensive and requires professional skill to make an accurate assessment of the impact (Castine et al. 2009).
Alternatively, because bacterial communities respond quickly to environmental changes (Bissett et al. 2006, Castine et al. 2009, Kawahara et al. 2009), quantitative and qualitative analyses of the spatial distribution and metabolic activities of major microbial indicators may provide crucial information on the major C org oxidation pathways and subsequent biogeochemical conversion of specific compounds that would be veiled by geochemical analysis alone (Vezzulli et al. 2002, Jørgensen 2006, Kondo et al. 2012a,b).However, little is known about the microorganisms directly responsible for the biogeochemical sulfur cycles in fish-farm sediments, where rapid recycling of sulfur compounds occurs (Asami et al. 2005, Bissett et al. 2006, Kondo et al. 2012a,b).The main objectives of this study were (1) to assess the environmental impact of finfish aquaculture in the sediments and (2) to elucidate microbial assemblages closely related to the production and removal of reduced sulfur from the surface sediments of a fish farm.

Study area
The study area was located in Gamak Bay, a semienclosed bay near Yeosu on the southern coast of the Korean Peninsula (Fig. 1), where the aquaculture industry is heavily developed.In the study area, 2 main fish species, rockfish Sebastes schlegeli and sea bream Pagrus major, are cultured for 1.5 to 2 yr before they are harvested after October.To improve the commercial value of the farmed fish before shipment, a large quantity of fish food is distributed just before harvest (NIFS 2007).Sediment samples were collected on 15 September 2010, when the sediments received a high organic load from uneaten feed and fecal pellets.Two sampling sites were chosen: one directly underneath the fish cages (hereafter the farm site), and another frontward of the water flow located ~50 m in front of the fish cages (hereafter the reference site) to reduce direct impact by uneaten feed or fecal pellets (Fig. 1).Water depth was approximately 4 and 8.5 m at the farm and reference sites, respectively.The bottom water temperature and salinity were 24°C and 30.1 psu, respectively, at both sites.Dissolved oxygen (DO) concentration in the bottom water was 6.16 mg l −1 at both sites (NIFS 2010).Surface sediments at both sites were composed mostly of silt-and clay-sized sediment (NIFS 2010).Capitella capitata, a macrofaunal indicator of highly sulfidic conditions, was observed in the surface sediments of the farm site (NIFS 2010).

Sampling and handling
Sediment samples were taken before feeding to avoid disturbing the sediment condition.Triplicate sub-samples were collected for geochemical analyses using polycarbonate cores (6.5 cm i.d., 25 cm long) by SCUBA divers to minimize sediment distur-bance during sampling activities.In the N 2 -filled plastic glove bag, the sediment core was sectioned into 1 cm intervals from the top of the core to a depth of 6 cm and at 2 cm intervals from 6 cm to 10 cm in depth.The sediments were then transferred to sterile centrifuge tubes (BD).The tubes were tightly capped and centrifuged for 10 min at 2700 × g.After reintroduction into the glove bag, porewater was sampled and filtered through 0.2 µm cellulose ester syringe filters (ADVANTEC).Duplicate samples of surface sediment (0−2 cm) for measuring particulate organic carbon (POC) and total nitrogen (TN) were frozen at −20°C until processed in the laboratory.Surface sediments for DNA extraction were collected using plastic cores that were pre-washed with sterilized water, rinsed with 70% ethanol and then stored at −80°C until analysis.

Rate of sulfate reduction
The microbial sulfate reduction rates (SRR) were determined in triplicate from intact cores (3 cm i.d., 35 cm long) according to the radiotracer ( 35 S-SO 4 2− , 15 kBq µl −1 , Institute of Isotopes) method of Jørgensen (1978).Five microliters of the 35 S-SO4 2− working stock solution were injected into the sediment cores at 1 cm intervals.After 2 h of incubation at in situ temperature, the sediment was sliced into sections and then fixed in Zn acetate (20%) before being frozen immediately.Reduced 35 S extraction was performed using the 1-step distillation method (Fossing & Jørgensen 1989).

Nucleic acid extraction and quantification of the dsr gene
Total DNA was extracted using a PowerSoil DNA Isolation Kit (Mo Bio Laboratories) following the manufacturer's instructions.A quantitative PCR (qPCR) was used to determine copy numbers of the alpha subunit of the dissimilatory (bi)sulfite reductase gene (dsrA).The copy number of dsrA was used to quantify sulfate-reducing prokaryotes and was determined using SYBR Green I assays, as described previously (Kondo et al. 2004, 2008, Liu et al. 2009).
The primers DSR 1F (5'-ACS CAC TGG AAR CAC G-3') (Wagner et al. 1998) and DSR R (5'-GTG GMR CCG TGC AKR TTG G-3') (Kondo et al. 2004) were used to amplify the dsrA gene.The amplicon size was ~220 bp.The SYBR Green I assay was always performed with a melting curve analysis to check PCR specificity.qPCR was performed on an ABI 7500 Real Time PCR system (Applied Biosystems) as follows: an initial incubation step for 2 min at 50°C (activation of the polymerase) followed by a 10 min pre-denaturation step at 95°C, and 40 cycles of denaturation for 30 s at 95°C and annealing for 1 min at 60°C.The PCR products from the envi ronmental dsrA gene clones were ligated into pGEM T-Easy vector (Promega) and then used as the standard DNA for the dsr gene quantification (Cho et al. 2017).

Diversity of 16S rRNA genes and the dsr gene
The bacterial 16S rRNA genes obtained from the sediment samples collected at the 0−1 cm and 1−2 cm depths of the farm and reference sites were amplified using the primers 27F (5'-AGA GTT TGA TCC TGG CTC AG-3') (Rho et al. 2005) and 1518R (5'-AAG GAG GTG ATC CAN CCR CA-3') (Campbell et al. 2001) as described by Cho et al. (2017).
PCR products were purified using the MEGAquick-spin™ fragment DNA Purification kit (iNtRON Biotechnology).The purified PCR amplicons were ligated into the pGEM-T Easy vector (Promega) and transformed into Escherichia coli DH5α cells.The insert DNA in the pGEM-T Easy vector was amplified using the primers M13F (-20) (5'-GTA AAA CGA CGG CCA G-3') and M13R (-20) (5'-CAG GAA ACA GCT ATG AC-3').All sequences obtained from SolGent (Daejeon, Korea) were checked for chimeric artifacts using the Pintail program of the Bioinformatics Toolkit (Ashelford et al. 2005).These sequences were aligned with closely related sequences obtained from GenBank (www.ncbi.nlm.nih.gov/ BLAST) using ClustalW in MEGA 4.0.The filtered 16S rRNA and dsrA sequences that shared > 97% similarity were grouped in the same operational taxonomic unit (OTU).Phylogenetic trees were constructed using the neighbor-joining method with the Jukes and Cantor distance model, implemented within MEGA 4.0.Node support was assessed by bootstrapping using 1000 bootstrap replicates.

Nucleotide sequence accession number
The bacterial 16S rRNA and dsrA sequences obtained in this study were deposited in the NCBI Gen-Bank database under accession numbers KC631436 to KC631612 for the bacterial 16S rRNA gene related to Epsilonproteobacteria and MH071538 to MH071597 for the dsrA gene.

Statistical analysis
The rarefraction analyses of partial sequences of 16S rRNA and dsr were calculated using the MOTHUR program (Fig. S1 in the Supplement at www.int-res.com/articles/suppl/q010p413_ supp.pdf; Schloss et al. 2009).To compare the difference of means for geochemical properties, POC, TN, SRR and the abundance of dsrA gene copy number, Student's t-test was used.Before analysis, the homogeneity of variance was checked using Levene's test.

Geochemical parameters
In the sediment trap, the suspended particulate matter (SPM), POC and TN at the farm site were ~2-fold higher than those of the reference site (Table 1).The POC and TN contents in the sediments and the acetic acid concentration in the porewater at the farm site were higher than those measured at the reference site, although the difference between the 2 sites was not significant (p = 0.076, p = 0.096 and p = 0.149, respectively, Table 1).The vertical profile of NH 4 + in porewater increased with depth to 1424 µM at the farm site and to 1133 µM at the reference site (Fig. 2A).The average concentration of NH 4 + down to 10 cm depth was 1.5-fold higher at the farm (1310 µM) than at the reference site (890 µM) (Table 2).NO 3 − concentrations were 5.2 µM and 3.8 µM at the sediment -water interface of the farm and reference sites, respectively, and decreased with depth to < 2.4 µM at both sites (Fig. 2B).PO 4 3− concentration ranged from 7.70 to 17.56 µM at the farm site and from 0.87 to 6.89 µM at the ref erence site (Fig. 2C).The average concentration of PO 4 3− was 3-fold higher at the farm site (10.74 µM) than at the reference site (3.74 µM).H 2 S concentration in porewater was < 8 µM down to 2 cm at the farm site but then rapidly increased with depth, reaching 394 µM at 10 cm, whereas the accumulation of H 2 S was relatively indiscernible at the reference site (Fig. 2D).The average concentration of H 2 S at the farm site (123.6 µM) was 15-fold higher than that at the reference site (8.20 µM) (Table 2).SO 4 2− concentrations were similar at both sites (~24 mM) (Fig. 2E).Fe 2+ concentration ranged from 1.4 to 10.4 µM at the farm site and from 0.8 to 36.0 µM at the reference site (Fig. 2F).The average concentration of Fe 2+ at the farm site (3.38 µM) was 6-fold lower than that at the reference site (20.31µM).The average concentrations of NH 4 + , PO 4 3− , HS − and Fe 2+ at the farm site were significantly different from those measured at the reference site (p < 0.0001, p = 0.006, p = 0.007 and p < 0.0001, respectively), whereas the difference in concentrations of NO 3 − and SO 4 2− between the 2 sites was not significant (p = 0.509 and p = 0.077, respectively).

SRR and dsr gene copy number
SRR ranged from 753 to 1957 nmol cm −3 d −1 at the farm site and from 22.1 to 77.0 nmol cm −3 d −1 at the reference site (Fig. 3A).Average SRR at the farm site (1209 nmol cm −3 d −1 ) was 19-fold higher than that measured at the reference site (62.1 nmol cm −3 d −1 ).The abundance of dsr gene copy number ranged from 1.1 × 10 9 to 3.4 × 10 9 copies cm −3 at the farm site and from 4.3 × 10 8 to 1.3 × 10 9 copies cm −3 at the ref- erence site (Fig. 3B).The peak of the dsr gene copy number (3.4 × 10 9 copies cm −3 ) was observed at a depth of 1 to 2 cm where SRR was maximal (1958 nmol cm −3 d −1 ).The SRR and the abundance of dsr gene copy number were significantly different between the farm and reference sites (p < 0.0001 and p = 0.0001, respectively).

Bacterial community composition
A total of 376 bacterial 16S rRNA gene sequences were analyzed to elucidate the major bacterial groups inhabiting the surface sediments at both sites.These sequences were assigned to 302 OTUs based on a 3% cutoff (Table S1 in the Supplement).The coverage rates of 16S rRNA gene libraries were 11 and 9% at a depth of 0−1 cm at the farm and reference sites, respectively, and 59 and 8% at a depth of 1−2 cm at the farm and reference sites, respectively (Table S1).Rarefaction curves based on the 16S rRNA gene sequences were obtained by plotting with the observed OTUs for each library (Fig. S1), yet none reached the curvilinear or plateau phase at the species level (3% difference) (Fig. S1).However, the underestimation of diversity at the family and order levels (10 and 16% difference) was less significant since the curves came close to reaching a plateau.Most sequences (> 50% of total 16S rRNA gene sequences) were affiliated with the Alpha-, Gamma-, Delta-and Epsilon proteobacteria, Acidobacteria and several minor groups including Actinobacteria, Armatimonadetes, Bacteroidetes, Betaproteobacteria, Chloroflexi, Cyano bacteria, Deferribacteres, Deinococcus, Firmicutes, Fusobacteria, Parcubacteria, Planctomycetes and Verrucomicrobia (see Fig. 6, Table S1).Major groups occupying >15% of total 16S rRNA gene sequences were related to Gamma-, Delta-and Epsilonproteo bacteria (Table 3).
In total, 114 dsrA gene sequences were obtained from the 1−2 cm depth interval and were sorted into 70 OTUs using our definition of > 97% sequence identity (Table S2).All dsrA gene sequences were affiliated with bacteria, and no archeal dsrA genes were detected.The coverage of the dsrA gene libraries showed 35 and 43% at the farm and reference sites, respectively.Rarefaction curves based on dsrA gene sequences at the farm and reference sites approached near saturation at the family level (10% difference) (Fig. S1).

Bacterial communities associated with sulfate reduction
Sulfate-reducing bacteria (SRB) revealed by 16S rRNA gene sequences were closely associated with Desulfarculaceae, Desulfobulbaceae, Desulfuromonadaceae and Syntrophobacteraceae in Deltaproteobacteria at both sites (Table 3).The relative abundance of SRB in Deltaproteobacteria comprised 11% (0−1 cm depth) and 6% (1−2 cm depth) of total 16S Fig. 2. Vertical distributions of porewater and solid-phase constituents at the farm and reference sites.TRS: total reduced sulfur rRNA gene sequences at the farm site and 10% (0−1 cm depth) and 28% (1−2 cm depth) of total 16S rRNA gene sequences at the reference site (Table 3).
The next 2 minor groups, Group I and Group II, were closely related to ated with Group I accounted for 11 and 6% of total dsrA gene sequences at the farm and reference sites, respectively (Table S2).S2).The sequences of Group II accounted for 8% of total dsrA gene sequences at the reference site (Table S2).Most sequences of Group II were closely related to those of uncultured SRB found in environmental samples from sites such as Victoria Harbor in Hong Kong (DQ112190; Zhang et al. 2008) (similarity > 96%) and a salt marsh on the east coast of the USA (KP992730; Angermeyer et al. 2016) (similarity: 90%); both sites are affected by high organic material input via anthropogenic activities (Fig. 4).

Bacterial communities associated with sulfur oxidation
As shown by the 16S rRNA gene sequencing, sulfur-oxidizing bacteria (SOB) were closely related to Gamma-and Epsilonproteobacteria (Table 3).The relative abundance of SOB affiliated with Gammaproteobacteria was 19% (0−1 cm depth) and 3% (1− 2 cm depth) of total sequences at the farm site, whereas they accounted for 12% (0−1 cm depth) and 8% (1−2 cm depth) of total sequences at the reference site (Table 3).Members such as Thioalbus, Thioalkalivibrio and Thiohalomonas belonging to the family Ectothiorhodospiraceae (similarity > 90%) appeared as the dominant SOBs (Table 3).Known as denitrifying chemolithoautotrophic SOB, these bacteria have been isolated from various environments such as the East Sea, hydrothermal vents and hypersaline environments (Sorokin et al. 2001, Park et al. 2011, Nunoura et al. 2014).

Impact of the fish farm on sediment geochemistry and sulfur cycles
The fish-farm sediments were characterized by highly reduced conditions with increased accumulation of NH 4 + , PO 4 3− and H 2 S (Fig. 2, Table 2) and extremely high SRR compared to measurements at the reference site (Fig. 3).The SRRs reported here were within the range (92 to 310 mmol m −2 d −1 ) reported from other organic-rich fish-farm sediments (Holmer & Kristensen 1992, Holmer et al. 2003, 2005) but were markedly higher than those measured in shellfish farms (30−61 mmol m −2 d −1 ), where no artificial fish feed is used (Dahlbäck & Gunnarsson 1981, Holmer et al. 2003, Kim et al. 2011, Hyun et al. 2013).
The most intriguing geochemical property observed in the fish-farm sediment was the sulfide depletion at the 0 to 2 cm depth (Fig. 2D), where the highest SRR and dsr gene abundance were also observed (Fig. 3).A combination of microbial C org oxidation by Fe(III) reduction and abiotic reduction of Fe(III) coupled with S oxidation is responsible for this mismatch of sulfide (Thamdrup et al. 1993, Canfield et al. 2005).First, the amounts of Fe(III) (oxal) (28−37 µmol cm −3 ) appeared to be a major solid form of Fe (oxal) , comprising 60 to 95% of total Fe (oxal) at the 0−2 cm depth in the fish-farm sediments (Fig. 2H,I).The availability of Fe(III) (oxal) ultimately stimulates microbial C org oxidation coupled with Fe(III) reduction, thereby resulting in an accumulation of Fe 2+ (Eq.1; Canfield et al. 2005).Second, the Fe(III) (oxal) in the sulfidic sediments is readily reduced by the sulfide to form Fe 2+ and S 0 in the sediment (Eq.2; Canfield & Thamdrup 1996).Both biotic and abiotic reduction of Fe(III) via Eqs.( 1) and ( 2), respectively, should produce substantial amounts of Fe 2+ in the surface sediment (Fig. 2F).Finally, at the farm site where sulfate reduction was high (Fig. 3A), the Fe 2+ removed the sulfide to form FeS (Eq.3; Canfield et al. 2005).Because the dissolved Fe 2+ was highly depleted at the farm site (Fig. 2F), H 2 S oxidation coupled with Fe 2+ oxidation (Eq. 3) is likely to be higher at the farm site than at the reference site.Accordingly, average concentrations of TRS (H 2 S, S 0 , FeS and FeS 2 ) at the 0−2 cm depth interval were 3-fold higher at the farm site (80 ± 8.3 µmol cm −3 ) than at the reference site (26 ± 17 µmol cm −3 ) (Fig. 2J).

Bacterial communities associated with production and removal of reduced sulfur
Phylogenetic analysis of the dsrA functional gene and the 16S rRNA gene revealed that the sequences closely affiliated with Syntrophobacteraceae and Desulfobulbaceae predominated at both sites (Fig. 4, Table 3).Most members (e.g.Desulfacinum, Desulforhabdus and Thermodesulforhadus) in family Syn-trophobacteraceae perform a complete oxidation of organic substrates, whereas the members of Syntrophobacter are known to be associated with an incomplete oxidation to produce acetate (Rosenberg et al. 2014).Due to the low similarity (57−75%) with the cultured isolates in Fig. 4, it remains to be resolved if the uncultured Syntrophobacteraceae groups obtained in our dsrA gene analysis (Fig. 4) are directly responsible for the complete or incomplete oxidation of C org .However, the clones from the 16S rRNA gene sequence showed high similarity (> 95%) with Desulfobulbus (Table 3), which is known to oxidize a broad range of substrates (e.g.propionate, alcohols and lactate) resulting from various fermentation pathways to produce acetate in anoxic freshwater and marine sediments (Lien et al. 1998, Pagani et al. 2011, Sorokin et al. 2012, Rosenberg et al. 2014).Because incomplete oxidizers can grow faster than complete oxidizers in natural sediments receiving pulses of rich organic material (Widdel 1988, also see Canfield et al. 2005), the Desulfobulbus-like bacterial communities in this study seem to be able to proliferate readily around the fish-farm sediments.
Despite the high SRR and dsrA abundance, the depletion of sulfide at the 0−2 cm depth at the fish farm (Fig. 3A,B) remains to be explained in terms of which microorganisms are associated with the removal of sulfide.Interestingly, at the 1−2 cm depth interval of this sulfide mismatch layer, Epsilonproteobacteria closely affiliated with Sulfurovum lithotrophicum (similarity > 95%), S. riftiae (similarity > 94%) and S. aggregans (similarity > 95%) composed the major fraction (62.0%) of total sequences (Figs. 5 & 6, Table 3).Similar results were reported in the sediments of salmon farms where Epsi lonproteobacteria is a major bacterial group (Aranda et al. 2015), although the percentage (39% of total clones) was lower than that reported in the present study.Both S. lithotrophicum and S. riftiae, isolated from the oxic-anoxic interface where sulfides meet oxygenated sea water, and the vent polychaete Riftia pachptila, respectively, are known to be chemo lithoautotrophs using S 0 or S 2 O 3 2− as an electron donor and O 2 or NO 3 − as an electron acceptor (Inagaki et al. 2004, Giovannelli et al. 2016) (Fig. 5).In this study, it is likely that the S 0 or S 2 O 3 2− was produced by the oxidation of H 2 S coupled with the reduction of Fe(III) via Eq.( 2) and served as electron donors for the Sulfurovum-like SOB that flourished in the sulfidemismatched zone where either O 2 or NO 3 − is available as an electron acceptor.Unlike S. litho trophicum and S. riftiae, S. aggregans uses H 2 as an electron donor and S 0 , S 2 O 3 2− and NO 3 − as electron acceptors under more anoxic conditions (Mino et al. 2014).Thus, bacteria closely related to S. aggregans may play a significant role in the reduction of S 0 and S 2 O 3 2− in the 1−2 cm depth of fish-farm sediments.In contrast to the Sulfurovum-like bacteria that thrived at the 1−2 cm depth, the clones that had high similarity (> 95%) with Thioalbus, Thioalkalivibrio, Thioalomonas, Thiohalobacter, Thiolapillus and Thioprofundum in Gammaproteobacteria occupied a major fraction (12−19% of total clones) mostly at the 0 to 1 cm depth interval of both farm and reference sites (Fig. 6, Table 3).Both Sulfurovum and the members of Gammaproteobacteria are microaerophilic chemolitho autotrophic sulfur oxidizers (Sorokin et al. 2001, Park et al. 2011, Nunoura et al. 2014).Gamma proteobacteria have a kinetically advantageous energyproducing pathway when oxygen and reduced sulfur compounds are steadily supplied, whereas Epsilonproteobacteria have the metabolic versatility to adapt to transient environmental conditions where the shift from aerobic to anaerobic microbial communities oc-Fig.6.Relative abundance of bacterial communities based on the 16S rRNA gene in the surface sediments (0−2 cm) associated with the vertical distribution of H 2 S, NO 3 − , Fe(III) (oxal) and sulfate reduction rates (SRR).Bar represents the percentage of clone library composition represented by each group at the farm and reference sites curs (Yamamoto & Takai 2011, Ihara et al. 2017).Consequently, our results strongly suggest that Sulfurovum-like bacteria in Epsilonproteobacteria play a significant ecological and biogeochemical role in the oxidation and reduction of reduced sulfur compounds in the highly sulfidic fish-farm sediments.

Fig. 1 .
Fig. 1. (A) Study area in the coastal waters near Yeosu, South Korea; (B) 'farm' indicates the sampling site where sediment samples were taken directly underneath the fish cages, whereas 'reference' denotes the sampling site located ~50 m away from the fish cages Environmental parameters of the bottom water and surface sediments and vertical flux of particulate materials in the sediment trap at the farm and reference sites.DO: Dissolved oxygen; SPM: suspended particulate matter; POC: particulate organic carbon; TN: total nitrogen.Surface sediment values are mean (±1 SD) of 2 sediment cores and are down to 10 cm depth (for porosity and density) or 2 cm depth (POC, TN and acetic acid).Sediment trap data is from NIFS (2010) Fig. 3. Vertical distributions of (A) sulfate reduction rate (SRR) and (B) DNA copy number of the dsrA gene at farm and reference sites

Fig. 4 .
Fig. 4. Phylogenetic tree based on partial sequences of dsrA from sulfate-reducing prokaryotes, including partial sequences based on environmental dsrA gene amplicons.Sequences were retrieved from sediment (1−2 cm depth) of the farm (YF) and reference (YC).The tree was constructed using the neighbor-joining method with Thermodesulfovibrio yellowstonii and T. islandicus.Bootstrap values are based on 1000 replicates and are indicated at branch nodes for values >50% bootstrap support.Numbers in square brackets indicate the number of dsrA gene sequences detected from YF and YC.GenBank accession numbers for each sequence are indicated in parentheses

Fig. 5 .
Fig. 5. Phylogenetic relationships of epsilonproteobacterial clones obtained from the fish-farm sediments.The phylogenetic tree is based on the 16S rRNA gene sequences.The tree was constructed using the neighbor-joining method with Aquifex pyrophilus.The clones from 0−1 cm (1) and 1−2 cm (2) depths at the farm (YF, red) and reference (YC, blue) sites are denoted in bold.Numbers in square brackets indicate the number of clones found in the clone libraries.GenBank accession numbers for each sequence are indicated in parentheses

Table 3 .
Bacterial composition and number of clones closely related to Gamma-, Epsilon-and Deltaproteobacteria found at the farm and reference sites.Numbers in parentheses represent the relative abundance of each clone.The putative function of closely related species (only sequence similarities > 95%) is indicated.Chem O: chemoorganotrophy; F: fermentation; FeR: Fe(III) reduction; NR: nitrate reduction; SO: sulfur oxidation; SR: sulfate reduction.The total number of clones is the total number of bacterial 16S rRNA gene clones from TableS1