All samples were collected by the RAS biofilter (UWM biofilter) at the University of Wisconsin-Milwaukee Great Lakes Aquaculture Facility. Measured from the base, the biofilter is ~2.74 m high and ~1.83 m in diameter. The water level within the biofilter is ~2.64 m from the base, with the fluidized sand filter matrix extending to a height of Extends ~1.73 m from the base. The biofilter is filled with Wedron 510 silica sand, which is fluidized to ~200% starting sand volume through the use of 19 Plan 40 PVC probes, each 3.175 cm in diameter. The probes receive inflow from the solid waste clarifier, which rises through the filter matrix. Samples for this study were collected at three depths within the fluidized sand biofilter, defined as surface (~1.32–1.42 m from the biofilter base), middle (~0.81–0.91 m from the biofilter base), and bottom (~0.15–0.30 m, made from biofilter base). Images of the UWM biofilter and sampling locations are shown in Figure 1  . The maximum flow rate of the biofilter inflow is 757 L per minute, resulting in a hydraulic retention time of ~9.52 min. Typical system water quality parameters are as follows (mean ± standard deviation): pH 7.01 ± 0.09, oxidation-reduction potential 540 ± 50 (mV), water temperature 21.7 ± 0.9 (°C), and dissolved oxygen (DO) of biofilter effluent 8.20 ± 0.18 mg/l. The biofilter is designed for maximum operation at 10 kg of feed per day, which is based on predicted ammonia production from fish protein breakdown at this feeding rate (  Timmons and Ebeling, 2013  ).

FIGURE 1. ILLUSTRATION OF UW-MILWAUKEE'S RECYCLING AQUACULTURE SYSTEM (RAS) FLUID SAND BIOFILTER  . For illustrative purposes only a single inflow pipe is shown. Nineteen of these pipes are present in the system. Water flow is shown with directional arrows, sample locations are marked with circles, and biofilter elevation is listed.

Sample collection, processing and DNA extraction

Samples from the top of the biofilter matrix were collected in autoclaved 500 mL polypropylene bottles. Two samples from the surface of the biofilter were collected during the last 2 months of a yellow perch rearing cycle and then immediately before the start of a new rearing cycle in the system. After the system was stocked with fish, samples were collected approximately every week for the first half of the new rearing cycle (the yellow perch strains present during this study take approximately 9 months to grow to market size). After collection, water from the biofilter matrix samples was decanted into a second sterile 500 mL bottle for further processing. Then, approximately 1 g of wet weight sand was removed from the sample bottle and frozen at −80 °C for storage prior to DNA extraction. Water samples were filtered to 0. 22 μm filters (47 mm mixed cellulose esters, EMD Millipore, Darmstadt, Germany) frozen at −80 °C and macerated with a sterilized spatula before DNA extraction. To address the spatial distribution of bacterial taxa separately, depth samples were collected from the filter matrix using 50 ml syringes with attached weighted Tygon tubing (3.2 mm ID, 6.4 mm OD; Saint-Gobain SA, La Défense, Courbevoie , France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA and macerated with a sterilized spatula before DNA extraction. To address the spatial distribution of bacterial taxa separately, depth samples were collected from the filter matrix using 50 ml syringes with attached weighted Tygon tubing (3.2 mm ID, 6.4 mm OD; Saint-Gobain SA, La Défense, Courbevoie , France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA and macerated with a sterilized spatula before DNA extraction. To address the spatial distribution of bacterial taxa separately, depth samples were collected from the filter matrix using 50 ml syringes with attached weighted Tygon tubing (3.2 mm ID, 6.4 mm OD; Saint-Gobain SA, La Défense, Courbevoie , France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. Depth samples were extracted from the filter matrix using 50 mL syringes with attached weighted Tygon tubing (3.2 mm ID, 6.4 mm OD; Saint-Gobain SA, La Défense, Courbevoie, France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. Depth samples were extracted from the filter matrix using 50 mL syringes with attached weighted Tygon tubing (3.2 mm ID, 6.4 mm OD; Saint-Gobain SA, La Défense, Courbevoie, France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA 4mm OD; Saint-Gobain SA, La Défense, Courbevoie, France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA 4mm OD; Saint-Gobain SA, La Défense, Courbevoie, France). Samples were categorized according to the approximate distance from the filter base as surface, center, and bottom. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA. The tubing was sterilized with 10% bleach and rinsed three times with sterile deionized water between samplings. DNA was extracted separately from biofilter sand and water samples (~1 g wet weight and 100 mL, respectively) using the MP Bio FastDNA^®  SPIN Kit for Soil (MP Bio, Solon, OH, USA) according to the manufacturer's instructions, except that each sample was beaten for 2 minutes with the beads included in the MP Bio FastDNA ® SPIN Kit at the single operating speed of the Mini-BeadBeater  ^-  16 (Biospec Products, Inc., Bartlesville, OK, USA). DNA quality and concentration were checked using a NanoDrop  ^®  Lite (Thermo Fisher Scientific Inc., Waltham, MA, USA). Sample details and associated environmental data and molecular analyzes are presented in Table S1.

Ammonia and nitrite measurements

For both the time series and depth profiles, a Seal Analytical AA3 Autoanalyzer (Seal Analytical Inc., Mequon, WI, USA) was used to quantify ammonia and nitrite using the manufacturer-supplied phenol and sulfanilamide protocols on two separate channels became. To quantify nitrite only, the cadmium reduction column was not installed in the Auto Analyzer. RAS operators recorded all other chemical parameters from submerged probes that measured temperature, pH and oxidation-reduction potential. Following laboratory standard operating procedures, RAS operators used Hach colorimetric kits to measure ammonia and nitrite concentrations in the rearing tank.

16S rRNA gene sequencing

To maximize read depth for a temporal study of biofilter surface communities, we used the Illumina HiSeq platform and separately targeted the V6 region of the 16S rRNA gene for  archaea  and  bacteria  . In total, we received community data from 15 dates for temporal analysis. To interrogate changes in the spatial distribution of taxa across depth in the biofilter and obtain increased taxonomic resolution, we used 16S rRNA gene V4-V5 region sequencing on an Illumina MiSeq. We obtained samples from three depths  n  = 5 for the surface,  n  = 5 for the middle and  n = 4 for the bottom. Example metadata is listed in Table S1. Extracted DNA samples were sent to the Josephine Bay Paul Center at the Marine Biological Laboratory (V6  Archaea  and V6  Bacteria  ; V4-V5 samples from 12/8/2014 to 2/18/2015) and to the Great Lakes Genomic Center (V4-V5 samples from 11/18 .2014, 12/2/2014, 12/18/2014) for massively parallel 16S rRNA gene sequencing using previously published bacterial (  Eren et al., 2013  ) and archaeal (  Meyer et al., 2013  ) V6 Illumina HiSeq and bacterial V4 V5 Illumina MiSeq chemistry (  Huse et al., 2014b  ;  Nelson et al., 2014 ). Reaction conditions and primers for all Illumina runs are listed in the citations above and can be accessed at: https://vamps.mbl.edu/resources/primers.php#illumina. Sequence run processing and quality control for the V6 data set are in  Fisher et al. (2015)  , while CutAdapt was used to trim the V4-V5 data from low quality nucleotides (phred score <20) and primers (  Martin, 2011  ;  Fisher et al., 2015  ). Trimmed reads were merged using Illumina Utils as described previously (  Newton et al., 2015  ). Minimum Entropy Decomposition (MED) was implemented for each dataset to group sequences (MED nodes = operational taxonomic units, OTUs) for sample community composition and diversity analysis ( Eren et al., 2015  ). MED uses the information uncertainty calculated via Shannon entropy at all nucleotide positions of an alignment to divide sequences into sequence-like groups (  Eren et al., 2015  ). The sequence datasets were parsed with the following minimum substantial abundance settings: Bacteria V6, 377; archaeal V6, 123; bacterial V4-V5, 21. The minimum substantial threshold sets the abundance threshold for inclusion of MED nodes (i.e. OTU) in the final data set. Minimum substantial frequencies were calculated by dividing the total number of 16S rRNA gene sequences per data set by 50,000 as suggested in MED Best Practices (sequence counts are listed in Table S2). The Global Alignment for Sequence Taxonomy (GAST) algorithm was used to assign a taxonomy to sequence reads ( Huse et al., 2008  ) and the Visualization and Analysis of Microbial Population Structures (VAMPS;  Huse et al., 2014a  ) website was used for uses data visualization.

Comammox  -amoA-  PCR

To target comammox  Nitrospira amoA  for PCR and subsequent cloning and sequencing,  amoA  nucleotide sequences were obtained from  van Kessel et al. (2015)  and  Daims et al. (2015)  were aligned with MUSCLE (  Edgar, 2004  ). The alignment was imported into EMBOSS to   generate  an amoA consensus sequence ( Rice et al., 2000  ). Primer sequences were identified from consensus using Primer3Plus (Untergasser et al., 2012)  ,  and  the candidates along with those described by  van Kessel et al. (2015), were evaluated from the consensus sequence in SeqMan Pro (DNAStar) using MUSCLE (  Edgar, 2004  ). The  pmoA  forward primer (  Luesken et al., 2011  ) and the candidate primer COM_amoA_1R (this study; Table  1  ) provided the best combination of read length and specificity and were subsequently used to   amplify amoA genes from our samples.

.

 

Construction of a clone library and phylogenetic analysis

Multiple endpoint PCR approaches were used to   examine the nitrifying community composition of the RAS fluidized sand biofilter for amoA  (  Gammaproteobacteria, Betaproteobacteria, Archaea,  and Comammox  Nitrospira  ),  nxrA  (  Nitrobacter  spp.), and  nxrB  (Non-  Nitrobacter NOB). The primer sets and reaction conditions used are   listed in Table 1 . All endpoint PCR reactions were performed with a volume of 25 μl: 12.5 μl 2x Qiagen PCR master mix (Qiagen, Hilden, Germany), 1.5 μl appropriate primer mix (F&R), 0.5 μl bovine serum albumin ( BSA), 0.75 μl 50 mM MgCl  ₂  and 1 μl DNA extract.

DNA samples of biofilter water and sand from four different time points of the rearing cycle were used to create clone libraries of  archaeal amoA  and  Nitrospira  sp. nxrB  .  A sample from the middle of the sand biofilter was used to construct clone libraries for Betaproteobacteria  -amoA  and Comammox  -amoA . The middle biofilter sample was selected because it produced well-defined amplicons suitable for cloning target  amoA genes. All PCR reactions for cloning libraries were constructed using a TOPO-PCR 2.1 TA cloning kit plasmid (Invitrogen, Life Technologies, Carlsbad, CA). Libraries were sequenced on an ABI 3730 Sanger sequencer with M13 forward primers. Vector plasmid sequence contamination was removed using DNAStar (Lasergene Software, Madison, WI).

Cloned sequences of  Betaproteobacteria amoA, Archaea amoA,  and  Nitrospira nxrB  from this study were added to ARB alignment databases from previous studies (  Abell et al., 2012  ;  Pester et al., 2012  ,  2014  ). Comammox  -amoA  sequences from this study were compared with those from  van Kessel et al. matched. (2015)  ,  Pinto et al. (2015)  and  Daims et al. (2015)  using MUSCLE and imported into a new ARB database where the alignment was heuristically corrected prior to phylogenetic tree reconstruction. For the AOA, AOB and  Nitrospira amoA phylogenies, relationships were determined using maximum likelihood (ML) with RAxML on the Cipres Science Gateway (  Miller et al., 2010  ;  Stamatakis, 2014  ) and Bayesian inference (BI). calculated by MrBayes with a significant posterior probability of <0.01 and the associated consensus tree (  Abell et al., 2012  ;  Pester et al., 2012  ,  2014  ) integrated by ARB into a tree block within the input nexus file to reduce the calculation time (  Miller et al., 2010  ;  Ronquist et al., 2012  ). Consensus trees were then calculated from the ML and BI reconstructions using ARB's consensus tree algorithm (  Ludwig et al., 2004  ).

The  Nitrospira nxrB  sequences generated in this study were significantly shorter than those used for the  nxrB phylogenetic  reconstruction in  Pester et al. (2014)  , therefore we did not perform phylogenetic reconstructions as with the other marker genes. Instead, the UWM Biofilter and  Candidatus  Nitrospira nitrificans sequences were added to the majority consensus tree by  Pester et al. (2014)  using the quick-add parsimony tool of the ARB package (  Ludwig et al., 2004  ). This tool uses sequence similarity to add sequences to pre-existing trees without changing the tree topology.

 

qPCR assays for target marker genes

Quantitative PCR assays were developed to   distinguish two Nitrospira nxrB  genotypes and two  Nitrosomonas amoA genotypes in our system. Potential qPCR primer sequences were identified using Primer3Plus (  Untergasser et al., 2012  ) on MUSCLE (  Edgar, 2004 ) generated alignments in DNAStar (Lasergene Software, Madison, WI). Primer concentrations and annealing temperatures were optimized for specificity for each reaction target. Primers were checked using Primer-BLAST on NCBI to ensure assays matched their target genes. The newly designed primers were tested for cross-reactivity between genotypes using the non-target genotype sequence in both endpoint and real-time PCR dilution series. After optimization, all assays amplified only the target genotype. Due to the high sequence similarity between the two  archaeal amoA  genotypes (>90% identity) in our system, a single qPCR assay was developed to target both genotypes using the steps described above. The two closely related sequence types were pooled in equimolar amounts for reaction standards. A Komammox amoA  qPCR primer set was developed using the same methods as the other assays presented in this study. All test conditions are listed in Table  1  . All qPCR assays were performed on an Applied Biosystems StepOne Plus thermal cycler (Applied Biosystems, Foster City, CA).  Cloned target genes were used to generate standard curves from 1.5 × 10  ^{6 to 15 copies per reaction.} All reactions were performed in triplicate, with melting curve and endpoint confirmation of the assays (qPCR standard curve parameters and efficiency are listed in Table S3).

 

Statistics and data analysis

Taxonomy-based data were visualized with heatmaps created in the R statistical language (  R Core Team, 2014  ) by implementing functions from the gplots, Heatplus from Bioconductor Lite, VEGAN, and RColorBrewer libraries. MED nodes were used in all sample diversity metrics. The EnvFit function in the VEGAN (  Oksanen et al., 2015  ) R package was used to test the relationship between RAS observation data and changes in biofilter bacterial community composition. Pearson's correlations were calculated using the Hmisc package in R (  Harrell, 2016  ) to test whether 16S rRNA,  amoA  and  nxrB  gene copies were correlated over time. Kruskal-Wallis rank sum tests were performed in the R basic statistics package ( R Core Team, 2014  ) to test whether the populations of the above genes were stratified by depth. VEGAN's ADONIS function was used on the V4-V5 depth dataset to test the significance of the observed Bray-Curtis dissimilarity as a function of the categorical factors of depth, with strata = ZERO because the same biofilter was sampled multiple times.

 

Biomass model

To determine whether the observed ammonia removal could provide the energy required to support the number of potential ammonia-oxidizing microorganisms (AOM) in the biofilter as quantified via qPCR, we included the steady-state biomass concentration from the measured ammonia oxidation modeled by the following equation:

: 

 X  _AO  is  defined in previous models (Mußmann et al., 2011) as the biomass concentration of ammonia oxidizers in milligrams per liter, but in this study we converted to cells per wet gram of sand by finding the mean grams of sand per liter of water in Biofilter. Θ  _x  is the mean cell residence time (MCRT) in days and was unknown for the system. Θ is the hydraulic residence time in days, which in this system is ~9.52 min or 0.0066 days. Y  _AO  is the growth yield of ammonia oxidizers and  b  _AO  is the endogenous respiration constant of ammonia oxidizers, which  ^{were estimated to be 0.34 kg volatile suspended solids (VSS)/kg NH4 +}  -N and 0.15 d  -1 by  Mußmann et al. (2011)  . Δ  S  _NH  3  is the change in substrate ammonia concentration between inlet and outlet in mg/l. To  calculate  _{_}  _  _  _  _  _ *C/μm  ³  (  Mußmann et al., 2011  ) to relate the biovolume to endogenous respiration. The modeled biomass concentration was plotted against a range of potential MCRT for a RAS fluidized sand filter (Summerfelt, personal communication). The results of all amoA  qPCR assays were combined to estimate total ammonia-oxidizing microorganism biomass in copy numbers per gram wet weight of sand. The modeled biomass was then compared to our AOM qPCR assay results. A commented R script for the model is available on GitHub ( https://github.com/rbartelme/BFprojectCode.git  ).

 

NCBI sequence accession numbers

Bacterial V6, V4-V5, and archaeal V6-16S rRNA gene sequences generated in this study are available from the NCBI SRA (SRP076497; SRP076495; SRP076492). Partial gene sequences for amoA and nxrB are available through NCBI Genbank and have accession numbers KX024777–KX024822. Results Biofilter chemistry results RAS operational data were examined from the beginning of a yellow perch rearing cycle to approximately 6 months thereafter. The mean biofilter influent concentrations of ammonia and nitrite were 9.02 ± 4.76 and 1.69 ± 1.46 μM, respectively. Biofilter wastewater ammonia concentrations (3.84 ± 7.32 μM) remained within the toxicological limitations (

< 60 μM) of P. flavescens grown in the system. Occasionally, nitrite accumulated above the recommended threshold of 0.2 μM in both the rearing tank (0.43 ± 0.43 μM) and the biofilter effluent (0.73 ± 0.49 μM). No major fish diseases were reported during the RAS's operational period. Environmental and operational data are listed in Table S1. Bacteria and archaea accumulations within the biofilter Characterization of the RAS biofilter bacterial community showed that both the sand-associated and aquatic communities were diverse at a broad taxonomic level; Seventeen phyla averaged >

0.1% in each of the biofilter sand and water bacterial communities (see Table S2 for an example taxonomic characterization of the genus). Proteobacteria (on average 40% of biofilter sand community sequences and 40% of water sequences) and Bacteroidetes (18% in sand, 33% in water) dominated both water and sand bacterial communities. In taxonomic classification at the family level, the community associated with biofilter sand was different from the aquatic community. The majority of sequences in the sand samples were assigned to the bacterial groups Chitinophagaceae (average relative abundance 12%), Acidobacteria family unknown (9%), Rhizobiales family unknown (6%), Nocardioidaceae (4%), Spartobacteria family unknown ( 4%) and Xanthomonadales - family unknown (4%). Water samples were dominated by sequences classified as Chitinophagaceae (14%), Cytophagaceae (8%), Neisseriaceae (8%), and Flavobacteriaceae (7%). At the genus level , Kribbella, Chthoniobacter, Niabella, and Chitinophaga were the most numerous taxa classified, each with an average of >3% relative abundance in the biofilter samples. Using Minimum Entropy Decomposition (MED) to obtain highly discriminatory sequence classification, we identified 1261 nodes (OTUs) in the entire bacterial dataset. A MED-based comparison of bacterial community composition (Figure 1 ) supported the patterns observed using a broader taxonomic classification, indicating that the biofilter sand-associated community was distinct from the assemblage present in the biofilter water. In contrast to the high diversity in the bacterial community, we found that the archaeal community is dominated by a single taxonomic group belonging to the genus Nitrososphaera This taxon accounted for >99.9% of the Archaea -classified sequences identified in the biofilter samples (Table S2). This taxon was also almost entirely represented by a single sequence (>95% of sequences classified as Archaea ) that was identical to a number of Thaumarchaeota sequences deposited in the database, including the complete genome of Candidatus Nitrosocosmicus oleophilus (CP012850), together with clones from activated sewage sludge, wastewater treatment and freshwater aquariums (KR233006, KP027212, KJ810532–KJ810533). Initial characterization of the biofilter community composition revealed distinct communities between the biofilter sand and decanted biofilter water (Figure 2 ). Based on these data and the fact that fluidized bed biofilter nitrification occurs primarily in particle-bound biofilms ( Schreier et al., 2010 ), we focused our further analyzes on the biofilter sand matrix. In the sand samples, we observed a significant change in bacterial community composition (MED nodes) over time (Table 2 ). The early part of the study, which included a period when market size yellow perch were present in the system (sample −69 and −26), a fallow period after fish removal (sample 0), and the period after restocking of mixed juveniles (Samples 7 and 14) had a more variable bacterial community composition (mean Bray-Curtis similarity 65.2 ± 6.5%) than the remaining samples ( n = 9), which were collected at time points after an adult food source had been started (20.0 ± 6.4%, Figure 3 ). Several operational and measured physical and chemical parameters, including oxidation-reduction potential, feed size, conductivity, and nitrite from the biofilter, were correlated ( S < 0.05) with the time-dependent changes in bacterial community composition (see Table 2 for environmental correlation results).

 

.

	FIGURE 2. DENDROGRAM ILLUSTRATING THE BACTERIAL COMMUNITY COMPOSITIONAL RELATIONSHIPS BETWEEN BIOFILTER SAND AND BIOFILTER WATER SAMPLES. A complete linkage dendrogram is represented using Bray-Curtis sample dissimilarity relationships based on nodal distributions of minimum entropy decomposition between samples (V6 dataset). The leaves of the dendrogram are labeled with the day count, with 0 representing the start of a fish rearing cycle. Negative numbers are days before a new breeding cycle. The day count is followed by the date of sampling (MM.DD.YY). See Table S1 for example metadata.
	TABLE 2. CORRELATIONS BETWEEN ENVIRONMENTAL VARIABLES AND BACTERIAL COMMUNITY COMPOSITION.
	FIGURE 3. NON-METRIC MULTIDIMENSIONAL SCALING DIAGRAM OF BRAY-CURTIS BACTERIAL COMMUNITY COMPOSITION DISsimilarity BETWEEN SAMPLING POINTS. nMDS stress = 0.07 and dimensions (k) = 2. Arrows show the progression of the sample through time from the end of a rearing cycle (day number -69 and -26) to a period without fish (0) and into the subsequent rearing cycle ( 7-126). The circle shows samples taken after the fish had grown to a size where feed type and amount were stabilized (3 mm pelleted feed and 3-7 kg feed per day).
Using a second sequence data set (V4-V5 16S rRNA gene sequences), we examined sand-associated bacterial community composition across a depth gradient (surface, middle, bottom). We found that the bacterial communities in the top sand samples were different from those in the middle and bottom (ADONIS  R  2  = 0.74,  p  = 0.001; Figure  4  ). The  planctomycetes  represented a larger proportion of the community in the surface sand (on average 15.6% of the surface sand versus 9.6% of the middle/lower sand), while the middle and lower layers harbored a larger proportion of  Chitinophagaceae  (7.4% in the surface sand vs. 16.8% middle/bottom) and  Sphingomonadaceae (2.4% in the surface vs. 7.9% in the middle/bottom; Figure  4  ).
	FIGURE 4. DEPTH COMPARISON OF BACTERIAL BIOFILTER COMMUNITY COMPOSITION. A heatmap is presented for all bacterial families with a relative abundance of ≥ 1% in each sample. Relative taxon abundance was generated from V4–V5 16S rRNA gene sequencing and is presented on a scale of 0 to 25%. The dendrogram represents the Bray-Curtis dissimilarity between sample community composition. Sample IDs are listed and sample depth is indicated by on the graph next to the dendrogram. Sample names correspond to sample metadata in Table S1.

 

Nitrification community composition and phylogeny

The massively parallel 16S rRNA gene sequencing data showed that bacterial taxa not associated with nitrification comprised the majority (~92%) of the sand biofilter bacterial community. In contrast, >99.9% of archaeal 16S rRNA gene sequences were assigned to a single taxon associated with known AOA. Among bacterial taxa, Nitrosomonas represented 1% of the total community in all samples, and no Nitrobacter sequences were obtained. We were also unable to amplify Nitrobacter nxrA genes (Figure S1) with a commonly used primer set ( Poly et al., 2008 ; Wertz et al., 2008 ). In contrast, Nitrospira was quite abundant, accounting for 2–5% of the total bacterial community (Table S2).

 

---

Context: 

 

ID: 425