Present address: Bärbel U. Foesel, Bereich Mikrobiologie, Department Biologie I Ludwig-Maximilians-Universität München, Germany. Eddie Cytryn, Department of Soil Water and Climate and Biotechnology Institute, University of Minnesota, Saint Paul, MN, USA.
Editor: Michael Wagner
Correspondence: Andreas Schramm, Department of Biological Sciences, Microbiology, University of Aarhus, Ny Munkegade, Building 1540, DK-8000, Aarhus C, Denmark. Tel.: +45 8942 3248; fax: +45 8942 2722; e-mail: email@example.com
Zero-discharge marine aquaculture systems are an environmentally friendly alternative to conventional aquaculture. In these systems, water is purified and recycled via microbial biofilters. Here, quantitative data on nitrifier community structure of a trickling filter biofilm associated with a recirculating marine aquaculture system are presented. Repeated rounds of the full-cycle rRNA approach were necessary to optimize DNA extraction and the probe set for FISH to obtain a reliable and comprehensive picture of the ammonia-oxidizing community. Analysis of the ammonia monooxygenase gene (amoA) confirmed the results. The most abundant ammonia-oxidizing bacteria (AOB) were members of the Nitrosomonas sp. Nm143-lineage (6.7% of the bacterial biovolume), followed by Nitrosomonas marina-like AOB (2.2% of the bacterial biovolume). Both were outnumbered by nitrite-oxidizing bacteria of the Nitrospira marina-lineage (15.7% of the bacterial biovolume). Although more than eight other nitrifying populations were detected, including Crenarchaeota closely related to the ammonia-oxidizer ‘Nitrosopumilus maritimus’, their collective abundance was below 1% of the total biofilm volume; their contribution to nitrification in the biofilter is therefore likely to be negligible.
Declining marine fish stocks have induced a rapid growth in marine aquaculture (FAO, 2004), an industry that can have deleterious effects on local environments via uncontrolled discharge of nutrients and organic matter (Wu, 1995; Christensen et al., 2000; Holmer et al., 2003). As an environmentally friendly alternative, a zero-discharge marine aquaculture system has been developed with integrated microbial biofilters that couple nitrification with sludge digestion and denitrification (Cytryn et al., 2003; Gelfand et al., 2003). Nitrification in such a system has to be efficient enough to keep ammonia concentrations below toxic levels (e.g.<30 μM for juvenile gilthead seabream; Wajsbrot et al., 1993), and yet dynamic enough to buffer fluctuations in ammonium production, e.g., caused by changing fish population size, age, and feeding. This challenge is quite different from nitrification in activated sludge and wastewater biofilms, where ammonium concentrations are often high and ammonia-oxidizing bacteria (AOB) with a low substrate affinity but high reaction rates dominate (Wagner et al., 2002; Koops et al., 2003). Marine nitrifying biofilters have not been evaluated in detail, and nitrifier diversity has sometimes been described on the basis of only a few clone sequences. Such diverse nitrifiers as Nitrosomonas europaea (Hovanec & Delong, 1996), Nitrosomonas cryotolerans, and the nitrite-oxidizing bacterium (NOB) Nitrospira marina (Tal et al., 2003), Nitrosomonasmarina (Grommen et al., 2005), or Nitrosomonas aestuarii and Nitrosomonas sp. Nm143 (Itoi et al., 2006) have been detected in marine biofilters. Recently, even a crenarchaeal ammonia oxidizer (AOA), ‘Nitrosopumilus maritimus’, was isolated from a marine aquarium (Könneke et al., 2005), and subsequently archaeal ammonia monooxygenase genes (amoA) were detected in marine samples and activated sludge (Francis et al., 2005; Park et al., 2006). However, quantitative data on the nitrifying community structure of marine biofilters are still lacking, and the importance of the respective species remains unclear. Therefore, the goal of this study was to provide a comprehensive and quantitative analysis of the AOB and NOB communities in a marine, nitrifying biofilter.
Materials and methods
System operation and biofilm sampling
Biofilm samples originated from the trickling filter of a zero-discharge marine aquaculture system (Cytryn et al., 2003; Gelfand et al., 2003) that was stocked with gilthead seabream (Sparus aurata). The trickling filter consisted of 1 m3 polyvinyl chloride (PVC) cross-flow medium with a surface area of 240 m2 (Jerushalmi Ltd., Ashod, Israel). The top of the filter was continuously sprinkled with water from the fish pool (pool volume 2.3 m3) at a rate of 60 L min−1 (±20 L min−1). Normal inlet ammonia concentrations were c. 20 μM with peaks of up to 100 μM after feeding; outlet concentrations ranged from 6 to 42 μM (Gelfand et al., 2003), and salinity was kept at c. 20 psu. The system was restarted in spring 2003; for sampling of intact biofilm, plastic strips were inserted into the trickling filter. Samples were taken after several months of stable operation in October 2003 and April/May 2004 by removing the strips that were either directly frozen (for DNA extraction) or fixed in 4% paraformaldehyde (for FISH; Amann et al., 1990b). Vertical sections (thickness, 14–30 μm) were prepared from fixed biofilms as described earlier (Schramm et al., 1996). Additionally, biofilm was scraped off the PVC substratum and fixed for qualitative FISH-comparison of the nitrifier community on the original PVC substratum and the inserted plastic strips.
Biofilm samples (1–2 cm2 on plastic strips) were incubated in a laminar flow chamber with artificial seawater (20 psu Red Sea salt, Red Sea Fish Pharm, Eilat, Israel) amended with 25 μM NH4+ at a flow velocity of 1 cm s−1. Vertical microprofiles of dissolved O2 and NOx− (nitrate+nitrite+N2O) were measured with amperometric microsensors (Revsbech, 1989) and microbiosensors (Larsen et al., 1997), respectively. Depth-resolved NOx− production rates were computed from the profiles using the software profile v.1.0 (Berg et al., 1998).
DNA was first extracted from 3 cm plastic strip fragments by bead-beating (=DNA extraction method A), based on a modified version of the FastDNA® SPIN Kit for soil (Qbiogene Inc., Carlsbad, CA). Extraction tubes contained-acid washed and baked glass beads (0.36 g with a diameter of 106 μm; Sigma-Aldrich, St Louis, MO, and six to eight beads with a diameter of 5 mm; Marienfeld, Lauda-Koenigshofen, Germany). Wash buffer was 50 mM NaCl, 10 mM Tris-HCl (pH 7.5), 2.5 mM EDTA, and 50% (v/v) ethanol. As this method proved insufficient for yielding DNA of AOB, the FastDNA® Kit (Qbiogene) was combined with an enzymatic digestion step (Juretschko et al., 1998), hereafter referred to as the DNA extraction method B: biofilm pellet (0.25 g) was resuspended in 300 μL DNA extraction buffer [100 mM Tris-HCl (pH 8.0), 100 mM sodium EDTA (pH 8.0), 100 mM sodium phosphate (pH 8.0), 1.5 M NaCl, 1% cetyltrimethylammonium bromide], transferred to a Lysis matrix E tube (Qbiogene), and bead-beaten (2 × 15 s at speed 4.0 and 1 × 15 s at speed 4.5) using a FastPrep® Instrument (Qbiogene). Then, 50 μL of enzyme mixture I [lysozyme (Sigma-Aldrich, Steinheim, Germany), lipase typ7 (from Candida rugosa, Sigma), pectinase (from Aspergillus niger, Sigma), and β-glucuronidase (from Helix pomatia, Sigma), 10 mg L−1 each] was added and the biofilm was incubated for 30 min at 37 °C, followed by addition of 50 μL of enzyme mixture II [proteinase K (Roche Diagnostics GmbH, Mannheim, Germany), protease typ9 (from Bacillus polymyxa, Sigma), and pronase P (from Streptomyces griseus, Sigma), 10 mg L−1 each] and incubation for 30 min at 37 °C. Finally, the sample was incubated with 75 μL of 20% dodecyl sulfate for 2 h at 65 °C, then 800 μL cell-lysis solution (CLS)-TC (Qbiogene) was added for a second bead-beating round (2 × 15 s at speed 5.0 and 1 × 15 s at speed 5.5), and DNA extraction was continued with the FastDNA® Kit following the manufacturer's instructions. In addition, the DNA extraction protocol described by Burrell et al. (2001) was tested, which includes enzymatic digestion with lysozyme and proteinase K.
PCR amplification of 16S rRNA genes and amoA
Primer sets GM3/GM4 (Muyzer et al., 1995), 616V (Juretschko et al., 1998)/NSO1225R (published as probe by Mobarry et al., 1996), and NSMR76F (published as probe by Burrell et al., 2001)/NSO1225R were used for amplification of bacterial, β-AOB, and Nitrosomonas marina-like 16S rRNA genes, respectively. The presence of Nitrobacter spp. was tested using primers 26F (Hicks et al., 1992)/NIT3R (published as a probe by Wagner et al., 1996), and the presence of γ-AOB was tested by following the protocol of Ward et al. (2000). PCR mixtures contained 75 mM KCl, 10 mM Tris-HCl pH 8.8, 1.5 mM MgCl2, 0.4 mg mL−1 bovine serum albumin, 125 μM of each dNTP, 10 pmol of each primer, 1 μL of template, and 1 U of Taq DNA-polymerase (Sigma). Thermal cycling included an initial denaturation step of 94 °C for 2–5 min, followed by 35 cycles of denaturation at 94 °C for 1 min, annealing for 1 min at 42 °C (GM3/GM4), 52 °C (616V/NSO1225R), 54 °C (NSMR76F/NSO1225R), or 60 °C (26F/NIT3R), and extension at 72 °C for 2 min; cycling was completed by a final elongation step at 72 °C for 5 min.
Bacterial amoA fragments were amplified using primer pair amoA-1F/amoA-2R (Rotthauwe et al., 1997), PCR mixture as described above, and thermal cycling with initial denaturation of 94 °C for 2 min, followed by 35 cycles of denaturation at 94 °C for 30 s, annealing at 58 °C for 40 s, and extension at 72 °C for 45 s (+1 s cycle−1); final elongation was 5 min at 72 °C. Crenarchaeal amoA fragments were amplified using the primer pair CrenamoA-F2 (5′-TGG TCT GGY TWA GAC GMT GTA-3′) and CrenamoA-2R (5′-CCC AYT TTG ACC ARG CGG CCA-3′), PCR mixture as described above but with 100 pmol of primers, and thermal cycling with initial denaturation of 94 °C for 10 min, followed by 40 cycles of denaturation at 94 °C for 30 s, annealing at 58 °C for 30 s, and extension at 72 °C for 45 s (+1 s cycle−1); final elongation was 10 min at 72 °C (J.R. de la Torré & D.A. Stahl, unpublished data). Primer and PCR conditions described by Francis et al. (2005) were also tested but without positive results.
Cloning, clone screening, and sequencing
Nitrosomonas marina 16S rRNA gene-specific and crenarchaeal amoA PCR products were run on a 1.5% Nusieve 3 : 1 agarose gel (FMC Bioproducts, Rockland, ME). The desired bands (991 and 630 bp, respectively) were excised, purified (Wizard® SV Gel and PCR Clean-Up System, Promega, Madison, WI), and cloned using the pGEM-T Easy Vector System (Promega) according to the manufacturer's protocol. All other PCR products were directly cloned. Clones were screened by PCR using vector-specific primers, and grouped after restriction fragment length polymorphism (RFLP) analysis with restriction enzymes RsaI (16S rRNA gene library), AluI (AOB library), or HhaI and RsaI (amoA library). For each RFLP group, one to three clones were selected for plasmid extraction (Wizard Plus Minipreps Purification System, Promega) and automated sequencing (Macrogen Inc., Seoul, South Korea), followed by blast search (Altschul et al., 1997).
The coverage and genotype richness of the clone libraries were estimated using Good's coverage (Good, 1958) and the bias-corrected Chao1 estimator (Chao, 1984), calculated with the freeware program estimates (Colwell, 2005). RFLP groups with highly similar sequences were merged. Thresholds for the definition of operational taxonomic units (OTU) were 97% sequence similarity (Stackebrandt & Goebel, 1994) and 80% (Koops et al., 2003) for 16S rRNA and amoA gene sequences, respectively. In addition, collector's curves were fitted with a hyperbola [NOTU=A × (NClones/B+NClones)] in which the asymptote (fit-parameter A from the formula) describes the expected number of OTUs (Hill et al., 2003) (supplementary Fig. S1).
Nucleotide sequences were added to prealigned 16S rRNA and amoA gene databases using the arb program package (http://www.arb-home.de; Ludwig et al., 2004). After automated alignment and manual refinements, phylogenetic trees were calculated with distance-matrix, maximum-parsimony, and maximum-likelihood methods implemented in arb; the results are displayed as neighbor-joining trees with branchings that were not supported by the other treeing methods drawn as multifurcations (Ludwig et al., 1998). Only Escherichia coli positions 27-1225, i.e. between primers 616V and NSO1225R, were used to calculate 16S rRNA gene-based trees. AmoA trees were calculated based on nucleic acid sequences with a filter using only positions between primers amoA-1F and amoA-2R and excluding the third codon position.
Probe design and testing
Probes specific for the Nitrospira marina group and the different clone clusters of ammonia oxidizers (see Table 1) were designed and evaluated in silico using the ARB probe tools (Ludwig et al., 2004). Probe specificity and hybridization conditions were optimized by Clone-FISH as described previously (Schramm et al., 2002), using pGEM-T as the vector and E. coli JM109(DE3) (Promega) as the host strain. Mismatch-clones (three mismatches) as well as mismatch-probes (one central mismatch) served as negative controls; a list of clones and probes used for Clone-FISH can be found in the supplementary Table S1.
Table 1. Probes for FISH designed in this study with hybridization conditions
Microscopic images of hybridized samples were recorded on either a confocal laser scanning microscope (Zeiss LSM 510, Carl Zeiss, Jena, Germany) or, for quantification, on a Zeiss ApoTome (Carl Zeiss, Jena, Germany) using an AxioCam MRm camera and the axiovision software (version 4.4). Biovolume fractions of the major AOB (Nitrosomonas sp. Nm143- and Nitrosomonas marina-related species) and NOB (Nitrospira marina-like species) populations were quantified using the daime software package (Version 1.1, Daims et al., 2006). For each probe combination (specific CY3-labelled probe/CY5-labelled probes EUB338-I to III), a set of random optical sections (86, 120, and 40 image sets for Nitrosomonas sp. Nm143, Nitrosomonas marina, and Nitrospira marina, respectively, thickness 1 μm) was recorded. Thresholds were set manually for each image series before image defragmentation and biovolume determination. The final biovolume fractions were calculated after correction with the negative control probe NON338. Where possible, images were recorded separately from three layers (top, middle, and bottom, each c. 100 μm thick) along the vertical biofilm sections, and AOB biovolume fractions were calculated separately for each layer.
CO2 uptake under nitrifying conditions was tested for the major nitrifying populations and a group of potential AOB by combining MAR with specific FISH analysis (Lee et al., 1999). Short-term, oxic incubations (8.25 h) of intact biofilm on plastic strips received 1.5 MBq NaH14CO3, leading to a final molar activity of 220 kBq (μmol CO32−)−1, and contained 2 mM ammonium; analysis was performed as described previously (Andreasen & Nielsen, 1997; Lee et al., 1999).
Nucleotide sequence accession numbers
The 16S rRNA and amoA gene sequences obtained in this study were deposited at EMBL under accession numbers AM295508–AM295533 for 16S rRNA gene sequences of AOB, accession numbers AM295534–AM295536 for 16S rRNA gene sequences of ‘potential AOB’, accession numbers AM295537–AM295545 for 16S rRNA gene sequences of NOB, accession numbers AM295546–AM295575 for bacterial amoA sequences, and accession numbers AM295170–AM295174 for archaeal amoA sequences.
Biofilm structure and microenvironment
Biofilm thickness was on average 150 μm (n=100 sections), spanning a range of 50–500 μm but rarely exceeding 300 μm. Under normal in situ ammonium concentrations (c. 20 μM), O2 penetrated the whole biofilm, and nitrification occurred across the entire biofilm depth without apparent stratification at an NOx− production rate of 0.596 nmol cm−3 s−1 (Fig. 1).
In total, four 16S rRNA gene libraries using different DNA extraction methods and PCR primers were necessary before a satisfying congruency with the FISH results was obtained. In an initial bacterial library (primers GM3/GM4, 300 clones screened) based on extraction method A (bead-beating, no enzymes), about 20% of the analyzed sequences had 97% similarity to the NOB Nitrospira marina but not a single AOB sequence was found. Therefore, a semi-specific β-AOB library (primers 616V/NSO1225R) was constructed from the same DNA extract. Fifty-seven of the 61 clones screened belonged to a group termed Nitrosomonas marina cluster 1 (Fig. 2a and 3a), while four sequences were affiliated to non-ammonia-oxidizing Betaproteobacteria. As these findings were not consistent with the FISH results, the DNA extraction was modified to include enzymatic lysis (method B), and the semi-specific PCR amplification (primers 616V/NSO1225R) and cloning were repeated. This new library was dominated by Nitrosomonas sp. Nm143-like sequences (two clusters, 34 of 59 clones screened), followed by a group of sequences (13 clones) with 90% sequence similarity to the AOB Nitrosospira briensis. Because the latter lineage only included uncultured species and branched off before the deepest AOB branch point (Fig. 3a), it was provisionally termed ‘potential AOB’. Of the remaining clones, two sequences affiliated with Nitrosomonas marina cluster 1, one with Nitrosospira sp. Nsp58, and the remaining nine with non-AOB sequences (Figs. 2b and 3a). A third DNA extraction with enzymatic lysis (Burrell et al., 2001) yielded similar results and was not analyzed further (data not shown). Sequences for the second most abundant AOB, Nitrosomonas marina cluster 2 (see FISH results below), could first be retrieved after cloning a more specific PCR reaction (primers NSMR76F/NSO1225R) tailored for that group (Fig. 3a). Neither γ-AOB nor NOB of the genus Nitrobacter were detected by specific PCR.
Analysis of amoA largely confirmed the 16S rRNA gene-based AOB community; in an amoA library constructed from DNA extraction method A, exclusively Nitrosomonas marina-like sequences were detected, while method B yielded both Nitrosomonas marina and Nitrosomonas sp. Nm143-like sequences (Fig. 2c and d, Fig. 3b). In contrast to the 16S rRNA gene data, which suggested two groups of Nitrosomonas marina-like AOB, only one Nitrosomonas marina-like amoA sequence type was found; it is unclear to which of the two 16S rRNA gene-defined Nitrosomonas marina cluster it belongs. One could, however, speculate that it may be from cluster 1, because it was detected in the first DNA extract, in which, at the 16S rRNA gene level, exclusively Nitrosomonas marina cluster 1 sequences were obtained. No deep-branching amoA sequences corresponding to the ‘potential AOB’ 16S rRNA gene sequence group were found. Instead, crenarchaeal amoA sequences with 97% similarity to the recently described AOA ‘Nitrosopumilus maritimus’ (Könneke et al., 2005) were detected.
All clone libraries described above appear to be well sampled as indicated by their high coverage (98–100%) and consistent richness estimates (Table 2 and supplementary Fig. S1). Therefore, the differences between the libraries cannot be ascribed to insufficient sampling/screening efforts.
Table 2. Observed and estimated numbers of OTU in the AOB-specific clone libraries
NOTU, number of OTU found in the various AOB-specific libraries; thresholds for OTU definition were 97% and 80% sequence similarity for 16S rRNA and amoA gene sequences, respectively; Nclones, number of clones screened.
† Number of OTU estimated by the bias-corrected estimator Chao1.
Number of OTU estimated by hyperbolic curve fitting [NOTU=A × (NClones/B+NClones)].
§ NA, not applicable (cannot be calculated with less than three data points).
Initial FISH analysis using published probes indicated that the nitrifier community was dominated by Nitrospira-like NOB (as detected by probes Ntspa712 and Ntspa662), and by AOB of the Nitrosomonas europaea/eutropha- (probe NEU23a) and the Nitrosomonas marina-lineage (probe NSMR76). However, no 16S rRNA gene or amoA sequences affiliated with the Nitrosomonas europaea/eutropha-lineage could be detected, and the Nitrosomonas marina cluster 1-like sequences obtained were not targeted by probe NSMR76; in addition, sequences of the Nitrosomonas sp. Nm143-lineage had been retrieved, for which no FISH probes were available. Therefore, (1) probe NEU23a was compared with the Nitrosomonas sp. Nm143-like sequences, (2) new probes were designed based on the Nitrosomonas sp. Nm143- and Nitrosomonas marina-like sequences retrieved, and (3) probe NSMR76 was used as a primer to obtain the missing sequence information for that particular Nitrosomonas marina group. Sequence analysis revealed that probe NEU23a has only one weak mismatch to the Nitrosomonas sp. Nm143-like sequences, which is not covered by the NEU23a-competitor probe CTE (supplementary Fig. S2; Wagner et al., 1995). Probe NEU23a was subsequently proven to hybridize to Nitrosomonas sp. Nm143-like cell clusters by double hybridization using clone-specific probes Nm143-446-1+2 (Table 1) in combination with probe NEU23a. It appeared that these Nitrosomonas sp. Nm143-like cell clusters were surrounded by a highly autofluorescent layer or envelope, putatively of extracellular polymeric substances (EPS) that enclosed every cell tetrade (supplementary Fig. S3). Probe Nmar831 was specifically designed for Nitrosomonas marina cluster 1 (Table 1). Although the probe worked well in the clone-FISH tests, no cells could be detected in situ, indicating that Nitrosomonas marina cluster 2 (as targeted by probe NSMR76) was the in situ dominant Nitrosomonas marina group.
New probes were also designed and optimized for the Nitrospira marina-like sequences and the group of ‘potential AOB’ (Table 1). Probe Nspmar62 hybridized to almost all Nitrospira-like cells (detected with probe Ntspa662). The ‘potential AOB’ were identified in situ as small cell clusters with a typical AOB morphology (Fig. 4g).
Nitrifier community structure
Initial FISH comparison between biofilm grown on the original PVC support and biofilm grown on the inserted plastic strips revealed no visible differences in community composition, i.e., the plastic strips did not select for (or against) certain nitrifiers compared with the original biofilm (data not shown). Therefore, all further analyses were performed on the plastic strip-grown biofilm. Biovolume fractions for Nitrosomonas sp. Nm143- and Nitrosomonas marina-like AOB were not significantly different along a vertical biofilm profile (supplementary Table S2), i.e. no spatial stratification of nitrifiers was observed in this mostly rather thin biofilm; therefore, quantification was averaged over the whole biofilm depth. Using the optimized probe set (Tables 1 and 2), Nitrosomonas sp. Nm143-like and Nitrosomonas marina cluster 2-like AOB were quantified to account for c. 75% and 25% of the total AOB community, respectively. Several other AOB were occasionally detected but only in very low numbers, accounting together for less than 1% of the total AOB community (Table 3, Fig. 4). Compared with the total biofilm community, an AOB biovolume fraction of 8.9% was determined, corresponding to 6.7%Nitrosomonas sp. Nm143-like AOB and 2.2%Nitrosomonas marina-like AOB. The ‘potential AOB’ group accounted for less than 1% of all bacteria. Crenarchaeota were also detected in situ (Fig. 4h) but only in low abundance, about 0.1% of all cells.
Table 3. Contribution of AOB, AOA and NOB tested to overall community composition
Numbers give the fraction of all bacteria, numbers in brackets give the fraction of AOB or NOB, respectively; +, detected, but in very low numbers; −, not detected.
Nitrosomonas communis/nitrosa lineage
Nitrosomonas marina-like clones, cluster 1
Nitrosomonas marina-like clones, cluster 2
Nitrosomonas oligotropha lineage
Nitrosomonas sp. Nm143-like clones, cluster 1
Nitrosomonas sp. Nm143-like clones, cluster 2
Nitrospira marina-like clones
Nitrospira moscoviensis rel. group
The NOB community encompassed 15.7% of the total biofilm volume and consisted almost exclusively of Nitrospira marina with very few Nitrospira moscoviensis-like cell clusters (Table 3, Fig. 4i); Nitrobacter spp. were not detected.
Potentially active nitrifiers
After oxic incubation of intact biofilm with 2 mM ammonium, both types of numerically dominant AOB and the NOB were clearly MAR-positive (supplementary Fig. S4 a–d). This result indicates that all in situ dominating nitrifiers can take up CO2 under nitrifying conditions and were therefore potentially active. In contrast, the ‘potential AOB’ group was MAR-negative under the assay conditions applied (supplementary Fig. S4 e and f).
The cloning-FISH discrepancy
Difficulties arose when trying to match the initially retrieved sequence information to the nitrifier populations detected by FISH. These discrepancies were due to insufficient DNA extraction, PCR bias, and mis-hits of one FISH probe. This combination and the absence of the most dominant AOB after screening >60 AOB-specific clones (Table 3;Fig. 2) might seem extreme; however, in many studies it would not even surface because the ‘full cycle’ rRNA approach (Amann et al., 1995) is often not completed. Without verification of sequence data in situ (e.g. by FISH), or without relating quantitative FISH data to sequence-defined populations, artifacts and biases may easily remain undetected. In our case, the inability to detect Nitrosomonas sp. Nm143 in the first 16S rRNA and amoA gene libraries could be clearly traced to the DNA extraction method A, because both genes were detected when using the improved extraction method B. DNA extraction methods are known to influence microbial diversity studies (e.g. Miller et al., 1999; Krsek & Wellington, 1999; Martin-Laurent et al., 2001), and especially cluster-forming AOB have been proven difficult to extract (Juretschko et al., 1998; Schramm, unpublished data). It can only be speculated, based on microscopic observations (supplementary Fig. S3) and the success of enzymatic lysis (method B, Burrell et al., 2001; Itoi et al., 2006), that a thick layer of EPS prevented mechanical lysis (method A) of the Nitrosomonas sp. Nm143 cells.
Even after the improved DNA extraction, neither the 16S rRNA gene nor the amoA library reflects the different AOB's in situ ratio. Because the two genes yield diametrically opposed percentages for Nitrosomonas sp. Nm143, the most likely explanation are the different degrees of PCR bias (e.g. von Wintzingerode et al., 1997; Polz & Cavanaugh, 1998; Mahmood et al., 2006) for the 16S rRNA gene and the amoA gene PCR. In addition, different copy numbers of the 16S rRNA and AmoA genes in the different AOB may contribute to the observed skewed ratio in the clone library.
Finally, the hybridization of probe NEU23a to a nontarget sequence with a weak mismatch (supplementary Fig. S2) suggests that checking published probes in silico may not be sufficient to ensure probe specificity. Empirical testing may be warranted as new sequences become available.
The nitrifying community
Nitrosomonas sp. Nm143-like bacteria were the most abundant AOB in the system. The Nitrosomonas sp. Nm143-lineage has been described only recently (Purkhold et al., 2003) based on a few coastal isolates and clone sequences. During the last two years, additional sequences of this lineage were detected in estuarine (Bernhard et al., 2005; Freitag et al., 2006) and coastal samples (Urakawa et al., 2006a, b), and in a marine aquaculture system (Itoi et al., 2006). The closest cultured relative of the AOB detected in our system, isolate Nitrosomonas sp. NS20 (sequence similarity, 98–99% for 16S rRNA gene, 97% for amoA), was isolated from a heavily polluted and eutrophicated area in Tokyo Bay, Japan (Urakawa et al., 2006a). Despite the diverse habitats, AOB of the Nitrosomonas sp. Nm143-lineage seem to share a moderate salt requirement (brackish to coastal sites) and a preference for low to moderate (5–300 μM) ammonium concentrations. Therefore, the Nitrosomonas sp. Nm143-like AOB appear to be well adapted to the marine biofilters. The other in situ relevant AOB, Nitrosomonas marina cluster 2, can be linked to several clone sequences (Burrell et al., 2001; Grommen et al., 2005) and one isolate (Nitrosomonas sp. Is343; A. Bollmann et al., unpublished data), all originating from low ammonium environments with various salinities. Thus, it is likely that the occurrence of these Nitrosomonas marina-like AOB is indicative of low ammonium concentrations rather than of a certain salinity. Given this differing habitat range of the two dominant AOB groups, one may speculate that both Nitrosomonas sp. Nm143- and Nitrosomonas marina-like AOB thrive well under the normally low ammonium concentrations in the marine aquaculture system but that the former have an advantage under occasional substrate peaks (e.g. after feeding). The coexistence of multiple AOB populations in biofilms as a consequence of fluctuating substrate (or O2) concentrations has been postulated previously (Daims et al., 2001b; Gieseke et al., 2001, 2003; Koch, 2007).
Besides the two dominant AOB groups, five bacterial and one archaeal ammonia-oxidizing populations were detected, albeit in very low numbers, i.e.,<0.1% of all cells in the biofilm. This AOB community composition was stable over several months before the first sampling (Koch, 2007), and the community structure did not change during the 6 months between the two sampling periods. Although these marginal populations may become functionally relevant under certain conditions, it appears more likely that they are simply remnants of the start-up period of the reactor, when conditions were most different (J. van Rijn, pers. commun.).
The maximum genetic distance within the known betaproteobacterial AOB is about 89% (Koops et al., 2003). With 90% 16S rRNA gene sequence similarity to Nitrosospira briensis and the typical AOB aggregate morphology (Fig. 4g), the ‘potential AOB’ were possible candidates for a novel AOB group. However, extended phylogenetic analyses did not support a monophyletic origin of the ‘potential AOB’ with the betaproteobacterial AOB; instead, the ‘potential AOB’ belong to one of several groups of uncultured bacteria from diverse habitats (e.g. soil, bioreactors, or activated sludge), which are clearly placed outside the AOB-Gallionella branchpoint (Fig. 3a). This tree topology, together with the absence of a deep-branching betaproteobacterial amoA and the lack of CO2 fixation under nitrifying conditions do not support the authors' hypothesis of ‘potential AOB’. Even though one might speculate that the unnaturally high ammonium concentration in the MAR incubation (2 mM vs. 20 μM in situ) may inhibit a low ammonia-adapted AOB population, the combined results render a nitrifying metabolism of the ‘potential AOB’ very unlikely.
Crenarchaeal ammonia oxidizers have been implicated in marine nitrification (Beman & Francis, 2006; Hallam et al., 2006), and diverse archaeal amoAs can be readily detected in marine samples and activated sludge (Francis et al., 2005; Park et al., 2006). In the present system, archaeal amoA diversity was low, and even if one assumes that all CARD-FISH-detected Crenarchaeota (0.1% of the total biofilm volume) were ammonia oxidizers, their contribution to overall ammonia oxidation would be minor.
The dominant NOB were closely related to the marine nitrite oxidizer Nitrospira marina. Sequences of this so-called Nitrospira lineage IV (Daims et al., 2001a) have been identified previously in saltwater aquaria (Hovanec et al., 1998) and a marine biofilter (Tal et al., 2003). The high abundance of Nitrospira marina-like NOB (nearly twice the biovolume of AOB, which, due to their smaller cell size, will translate into fourfold higher cell numbers compared with AOB) resembles the situation in low ammonia freshwater environments (Schramm et al., 1999; Altmann et al., 2003), in which freshwater Nitrospira outnumbered AOB. This numerical dominance of various nitrospiras over AOB might therefore be a general characteristic of ammonium-limited systems.
In summary, although at least 10 different nitrifying populations were detected in the biofilm, only three of them (Nitrosomonas sp. Nm143, Nitrosomonas marina cluster 2, and Nitrospira marina) were quantitatively relevant in situ. Arriving at this result required repeated rounds and optimization of DNA extraction, PCR-cloning, and FISH analysis. A possible explanation for the coexistence of two AOB populations might be niche differentiation: one high-affinity/low-activity population (possibly Nitrosomonas marina) may be active under the normally low ammonium concentrations, while a second population (possibly Nitrosomonas sp. Nm143) may thrive better during occasional ammonium peaks. This hypothesis remains to be proven.
The authors thank Britta Poulsen and Pernille Thykier for excellent technical assistance, Rikke L. Meyer (University of Aarhus, Denmark) for performing the profile analyses, Dirk de Beer (Max Planck Institute for Marine Microbiology, Bremen, Germany) for helpful discussion and comments on the manuscript, and David A. Stahl (University of Washington, Seattle, WA) for pointing out the archaeal ammonia oxidizers to us. This project was financially supported by GIF Grant no. I-732–165.8/2001, and by the Universities of Bayreuth, Germany, and Aarhus, Denmark.