Autotrophic ammonia-oxidising bacteria (AOB) are a crucial component of the microbial communities of nitrifying wastewater treatment systems. Nitrification is known to occur in reactors of different configuration, but whether AOB communities are different in reactors of different design is unknown. We compared the diversity and community structure of the betaproteobacterial AOB in two full-scale treatment reactors – a biological aerated filter (BAF) and a trickling filter – receiving the same wastewater. Polymerase chain reaction (PCR) of 16S ribosomal RNA (rRNA) gene fragments with AOB-selective primers was combined with denaturing gradient gel electrophoresis (DGGE) to allow comparative analysis of the dominant AOB populations. The phylogenetic affiliation of the dominant AOB was determined by cloning and sequencing PCR-amplified 16S rRNA gene fragments. DGGE profiles were compared using a probability-based similarity index (Raup and Crick). The use of a probability-based index of similarity allowed us to evaluate if the differences and similarities observed in AOB community structure in different samples were statistically significant or could be accounted for by chance matching of bands in DGGE profiles, which would suggest random colonisation of the reactors by different AOB. The community structure of AOB was different in different sections of each of the reactors and differences were also noted between the reactors. All AOB-like sequences identified, grouped within the genus Nitrosomonas. A greater diversity of AOB was detected in the trickling filters than in the BAF though all samples analysed appeared to be dominated by AOB most closely related to Nitrosococcus mobilis. Numerical analysis of DGGE profiles indicated that the AOB communities in depth profiles from the filter beds were selected in a non-random manner.
Aerobic autotrophic ammonia-oxidising bacteria (AOB) are found within two phylogenetic groups based on comparison of 16S ribosomal RNA (rRNA) sequences [1–3]. One group comprises strains of Nitrosococcus oceani and Nitrosococcus halophilus within the class gammaproteobacteria and the other contains Nitrosomonas and Nitrosospira spp. within the class Betaproteobacteria. It should be noted that Nitrosococcus mobilis is misnamed and belongs to the betaproteobacterial genus Nitrosomonas not the genus Nitrosococcus in the Gammaproteobacteria. A continually expanding database of AOB 16S rRNA gene sequences has led to the description of distinct clusters within the betaproteobacterial AOB from the family ‘Nitrosomonadaceae’, five within the genus Nitrosomonas and five within the genus Nitrosospira. The genus Nitrosomonas also contains a lineage currently represented only by Nitrosomonas cryotolerans-like sequences . All betaproteobacterial ammonia oxidisers form a phylogenetically coherent group within which all organisms exhibit the same primary physiology. Therefore all known cultured bacteria clustering within this AOB radiation are autotrophic ammonia oxidisers.
AOB present in the environment are generally found at comparatively low concentrations, are slow growing and difficult to isolate in pure culture . Only 16 named species have been described thus far [5–9]. Consequently most of the physiological and kinetic data available for AOB has been based on a select group of cultured isolates. In particular many of these studies have focused on one AOB, Nitrosomonas europaea, readily available from national and international culture collections. However, cultured isolates may not represent the dominant AOB in the environment [3,10–13]. The use of molecular techniques allows a more complete understanding of the diversity and distribution of AOB in natural environments than is offered by cultivation-based methods alone.
A number of studies suggest that physiological and ecological differences exist between the betaproteobacterial AOB genera and clusters identified from culture-independent studies [14–17]. Natural, low ammonia environments commonly harbour Nitrosospira spp. [10,16] whereas engineered high ammonia environments often exhibit a predominance of Nitrosomonas spp. [12,18–20]. Several studies have suggested that environmental factors such as salinity and ammonia concentrations select for certain species of AOB [21–24]. Environmental factors may also influence the extent of AOB diversity. In wastewater treatment systems (full- and lab-scale) there appears to be selection for either predominance of a single AOB population (e.g. N. europaea-like organisms [25,26]; N. mobilis-like bacteria ; Nitrosospira-like bacteria ) or several different AOB populations occur together [13,28,29]. It has recently been suggested that the level of AOB diversity found in a reactor relates to the stability of the reactor . Hence engineering a system with a greater diversity may improve performance and stability. However, it is important to increase the diversity of the most suitable AOB. To do this we need to understand more about the conditions, which select for certain species and promote species richness, within and between different wastewater treatment reactors. For example it has recently been suggested that effluent with a high industrial input selects for N. mobilis-like AOB . This predominance, of a salt-tolerant AOB, is attributed to the high concentration of salt often encountered in industrial wastewaters.
Of course we must be cautious about the interpretation of these studies because they are essentially observational in nature. No one has yet undertaken robust, controlled and statistically designed experiments on the effect of environmental conditions on AOB distribution.
The purpose of this study was to examine the diversity of AOB in two different full-scale wastewater treatment reactors treating the same mixed domestic and industrial waste. A range of culture-independent techniques was employed to do this. Polymerase chain reaction (PCR) was used to amplify 16S rRNA gene fragments using primers selective for the betaproteobacterial AOB . The PCR-amplified fragments were analysed by denaturing gradient gel electrophoresis (DGGE) and the profiles were numerically analysed to allow statistically significant comparison of the dominant AOB populations in the reactors. In addition, a selection of cloned PCR-amplified 16S rRNA gene fragments obtained from the reactors was sequenced to determine the phylogenetic affiliation of the predominant AOB-like sequences.
2Materials and methods
2.1Wastewater treatment plant (WWTP)
Samples were collected from a WWTP run by Yorkshire Water, which receives a mixed waste of domestic and industrial origin (agrochemicals and speciality chemicals). The treatment of the wastewater at the Yorkshire Water Plant is in the form of a multistage treatment process. Two reactors were studied from the WWTP, a biological aerated filter (BAF) and filter bed reactor, which operate in parallel. Typical loadings for the WWTP are: TBOD, 3400 kg day−1; COD 12 200 kg day−1; NH4+–N 650 kg day−1; TSS, 3400 kg day−1; and typical values for the final effluent quality for the plant are: TBOD 9.3 mg l−1; COD 107.4 mg l−1; NH4+–N, 1.2 mg l−1; TSS, 15.7 mg l−1.
Effluent from a predominantly industrial source (70% industrial and 30% domestic, hydraulic loadings of 30 000 kg day−1 and 13 000 kg day−1 respectively) is initially pre-treated using a high rate activated sludge (HRAS) process. The effluent from the pre-treatment plant is mixed with more of the domestic wastewater (2:1 – domestic/pre-treatment plant effluent) and routed through either the BAF or two trickling filter beds in series. The BAF system is composed of three linked units optimised for denitrification (anoxic), carbon removal (oxic) and nitrification (oxic). Flow through the BAF unit is in the order above and operates with 50% recycle. Both the BAF and filter bed reactors are attached growth, continuous flow reactors.
Following treatment the effluent is discharged to the local river systems.
Duplicate biomass samples were taken from the backwash flow from each unit in the BAF system. The top, middle and bottom (0, 0.9, 1.8 m depth) of the primary and secondary filter beds were sampled (in duplicate) by manual excavation. Biomass was preserved immediately on sampling in 50% ethanol and stored at −20°C prior to analysis.
DNA was extracted from biomass using a modification of the method described by Curtis et al. . Approximately 2 ml of sample was centrifuged at 13 000×g for 10 min. The supernatant was removed and the pellet was resuspended in 500 μl of 0.12 M sodium phosphate buffer (pH 8.0). The mixture was transferred to sterile Hybaid ribolyser tubes suitable for DNA extraction from bacteria (Blue matrix, Hybaid Ltd, Middlesex, UK) and 700 μl of phenol:chloroform:isoamyl alcohol (25:24:1) was added. The samples were lysed for 30 s at 6.5 m s−1 in a Ribolyser (Hybaid Ltd, Middlesex, UK) and subsequently centrifuged at 13 000×g for 10 min to separate the phases. The upper aqueous phase was collected and extracted with phenol:chloroform:isoamyl alcohol. The mixture was vortexed briefly before being centrifuged at 13 000×g for 10 min. The phenol–chloroform extraction was repeated twice. After the final phenol–chloroform extraction a measured volume of the aqueous phase was removed and transferred to a fresh tube to which two volumes of PEG 6000 (30% w/v in 1.5 M NaCl) were added. The samples were refrigerated at 4°C for 1 h to precipitate the DNA. The DNA was pelleted for 15 min at 13 000×g and the supernatant was removed. The pellet was re-dissolved in 200 μl of TE buffer [10 mM Tris–HCl, 1 mM ethylenediaminetetra-acetate (EDTA), pH 8.0] and further purified by the addition of two volumes of ice-cold (−20°C) absolute ethanol and 20 μl of 3 M sodium acetate (pH 5.2). The samples were incubated at −20°C for 10 min and then pelleted by centrifugation at 13 000×g for 10 min. The supernatant was removed and the pellet was air-dried. The dried pellet was dissolved in 50 μl of TE buffer. Nucleic acids were quantified spectrophotometrically at 260 nm and the DNA was diluted to 20 μg ml−1 for PCR amplification.
2.5PCR amplification of 16S rRNA genes from bacterial and AOB populations
PCR amplification of bacterial 16S rRNA gene fragments was performed using primer 2 (5′-ATTACCGCGGCTGCTGG-3′) and primer 3 (5′-CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCACGGGGGGCCTACGGGAGGCAGCAG-3′) . The betaproteobacterial AOB were analysed using PCR amplification with primers CTO189f (5′-CCGCCGCGCGGCGGGCGGGGCGGGGGCACGGGGGGAGRAAAGYAGGGGATCG-3′) and CTO654r (5′CTAGCYTTGTAGTTTCAAACGC-3′) . For DGGE analysis, nested amplification of the PCR products obtained with the AOB specific primers was done using primer 2 and primer 3 . Nested amplification was not required for bacterial DGGE analysis. AOB 16S rRNA gene fragments to be cloned were also obtained using a nested amplification. The first round of amplification was conducted using primers pA (5′AGAGTTTGATCCTGGCTCAG-3′) and pHr (5′-AAGGAGGTGATCCAGCCGCA-3′, ) followed by amplification with the AOB specific primers CTO189f and CTO654r . PCR was conducted using an Omn-E programmable thermal cycler (Hybaid Ltd, Middlesex, UK). Different cycling parameters were used depending on the primers employed (Table 1). The reactions were carried out in a 50 μl volume containing 1 μl of template DNA, 10 pmol each of the forward and reverse primers, 2 units of Biotaq™ DNA polymerase (Bioline, London, UK), 1×NH4 buffer (no Mg2+, Bioline, London, UK), 1.5 mM MgCl2 (Bioline, London, UK), 10 nmol of each deoxyribonucleoside triphosphate and molecular biology-grade water.
Table 1. Summary of PCR conditions
aNo. of cycles for denaturation, annealing and extension steps.
24 cycles followed by a further 15 at annealing temperature of 53°C
10 min, 72°C
3 min, 95°C
1 min, 95°C
1 min, 57°C
1 min, 72°C
10 min, 72°C
DGGE analysis was conducted using the D-Gene DGGE system (Bio-Rad, Hercules, CA, USA). Polyacrylamide gels (10% polyacrylamide, 0.75 mm thick, 16 by 16 cm) were run in 1×TAE buffer (40 mM Tris–acetate, 1 mM EDTA, pH 8.3). A denaturant gradient ranging from 30 to 60% denaturant (100% denaturant is 7 M urea plus 40% vol/vol formamide in 1×TAE) was used. Gels were run at 60°C for 4 h at a constant 200 V and stained for 30 min in SYBR green I (Sigma, Poole, UK; diluted 1/10000 in 1×TAE). Stained gels were viewed with an ultraviolet transilluminator (UVP, San Gabriel, CA, USA) and photographed with a Polaroid camera (CU-5, GRI, Great Dunmow, Essex, UK).
2.7Cloning and sequencing of 16S rRNA gene fragments from AOB populations
PCR-amplified 16S rRNA gene fragments from AOB populations were cloned using an AdvanTAge™ PCR Cloning Kit (Clontech Laboratories UK Inc., Hampshire, UK) according to the manufacturer′s instructions. PCR products were ligated with plasmid pT-Adv and competent Escherichia coli TOP 10F′ cells were transformed with the ligated DNA. Plasmid DNA was extracted by boiling in TE buffer for 3 min. The boiled cell preparation was used as a DNA template in PCR to screen individual clones for the presence of inserts of the expected size.
Amplification of insert DNA was performed using the same reagent mixture as described above and the same temperature programme was used as used for primers pA and pHr (Table 1). The primers used were pUCr (5′-CAGGAAACAGCTATGAC-3′) and pUCf (5′-GTTTTCCCAGTCACGAC-3′). PCR products of the correct size from 30 randomly selected clones were screened by ARDRA and DGGE to identify different clone types.
2.8DGGE screening of cloned 16S rRNA genes
The conditions for DGGE analysis of cloned 16S rRNA gene fragments were as described above. Cloned 16S rDNA fragments were analysed by DGGE to relate bands in DGGE profiles from the original samples with the cloned DNA. DGGE bands from the original profile that co-migrated with cloned sequences were excised, re-amplified and sequenced. This confirmed that the bands in the original DGGE profile genuinely corresponded to cloned sequences that co-migrated.
2.9ARDRA screening of cloned 16S rRNA genes
PCR products were digested simultaneously with HaeIII and RsaI (Promega, Madison, WI, USA) for 4 h according to the manufacturer′s instructions. The restriction fragments from each clone were separated on a 3% agarose gel run for 1 h at 100 V in 1×TAE buffer. Gels were pre-stained with ethidium bromide (0.15 μg ml−1), visualised with an ultraviolet transilluminator (UVP, San Gabriel, CA, USA, and photographed with a Polaroid camera (CU-5, GRI, Great Dunmow, Essex, UK). Restriction patterns were compared by eye and clones with identical patterns were grouped.
Representatives of each group of clones identified, by DGGE and ARDRA analysis, were sequenced. The sequences of the complete cloned 16S rRNA gene fragments (ca 460 bp) were determined. Sequencing was achieved using the dye terminator method on an Applied Biosystems ABI Prism 377 automated DNA sequencer.
The partial 16S rRNA gene sequences from the reactor samples were aligned with published 16S rRNA gene sequences from betaproteobacterial AOB and related non-AOB were used as outgroup sequences. A phylogenetic distance tree was generated using the Jukes and Cantor correction  and the neighbour-joining algorithm  as implemented in the TREECON software package .
2.11Nucleotide sequence accession numbers
The 16S rDNA partial sequences obtained from the reactors in this study are available from the EMBL nucleotide sequence database under accession numbers AF527013–AF527028.
2.12Statistical analysis of the DGGE gels
The Quantity One® 4.1.1 software package (Bio-Rad, Hercules, CA, USA) was used to analyse DGGE profiles. To correct for variations across the gel, a marker sample was run on either side of samples. Bands were assigned and matched automatically and the band assignments were checked manually. Band matching data was stored as a binary matrix and analysed using Raup and Crick′s index of similarity . The index of similarity was calculated as described by Legendre and Legendre  using the RCSI program (author – Peter Cejchan) on a PC running SuSE Linux 6.6. RCSI was obtained from the PaleoNet FTP site (http://www.nhm.ac.uk/hosted_sites/paleonet/) run by N. Macleod. The Raup and Crick index of similarity  was used to test if similarities observed within and between samples were greater (<0.05) or less (>0.95) than would be expected by chance.
2.13Raup and Crick's index of similarity
WWTPs exploit complex undefined communities of microorganisms to treat a range of wastes. The assembly of these microorganisms to form communities within WWTP is poorly understood but may be best considered in two ways. (1) The communities observed might arise through selection of specific organisms that are best adapted to the conditions in the WWTP; this exemplifies deterministic selection. (2) The communities present may arise through colonisation by organisms present in the external environment, and the particular organisms that colonise a WWTP are dictated by their chance arrival and proliferation in the plant. To rigorously test whether similarities in microbial communities observed in a comparison of two WWTP are genuinely the result of deterministic selection, the observed data must be statistically tested against a null hypothesis [36,38]. In this case the null hypothesis is that the similarities measured can be accounted for by the chance occurrence of the same organisms in the two WWTP. In the specific case of this study it is the chance occurrence of the same band in two DGGE profiles that is being tested. Most similarity coefficients commonly used to compare DGGE profiles are unable to discriminate similarities that are due to chance matching of bands or matching of bands at level greater than can be expected by chance alone (i.e. evidence of deterministic selection) and a statistical test of the significance of observed similarities against a null hypothesis is rarely explicitly considered in such comparisons. The Raup and Crick method  employs a randomisation procedure to detect similarities that are greater than can be accounted for by the chance matching of bands in two profiles and permits the statistical significance of the similarities to be tested against the null hypothesis. The Raup and Crick index of similarity compares the number of species common to two sites (observed data) with the number of species common to two sites that would be expected if the species were selected at random (expected/randomised data) from the source population (influent). Differences in the observed data compared to the randomised data correlate with the level of similarity or dissimilarity between the two sites. The index of similarity is defined by the probability that the expected similarity (randomised data) would be greater than or equal to the observed similarity. Raup and Crick thus proposed that a similarity value (SRC)between 0.05 and 0.95 is indicative of random occurrence of the same organisms (DGGE bands) in two samples (null hypothesis not falsified) whereas values below 0.05 or above 0.95 indicate deterministic selection (null hypothesis rejected) and are due either to significant dissimilarity or significant similarity respectively. To compare the DGGE data from within and between each of the reactors (primary filter bed, secondary filter bed and BAF reactor) Raup and Crick similarities were calculated. All possible comparisons between and within reactors were made. In addition DGGE data from each of the reactors was pooled to compare the total AOB populations.
3Results and discussion
A number of recent studies on AOB in wastewater treatment systems have suggested that different plants support different populations and different levels of species richness. For instance a domestic wastewater biofilm from a lab-scale reactor in Japan was dominated by N. europaea-like AOB  while AOB populations from lab- and full-scale plants in Germany were dominated by Nitrosospira-like bacteria or N. mobilis-like AOB respectively [27,12]. Other lab- and full-scale reactors in Germany contained more diverse AOB populations [13,28,29]. To examine whether particular AOB are selected in full-scale reactors of different configuration receiving identical wastewater, the AOB present in reactors treating waste of the same origin were characterised using culture-independent methods. PCR, DGGE and sequence analysis were combined to determine the dominant AOB populations in the BAF and trickling filter reactors.
3.1DGGE analysis of the BAF and trickling filter reactors
Analysis of duplicate samples by DGGE revealed that the profiles obtained were reproducible and for clarity, comparisons of single samples from each part of the reactor are shown. Visual comparison of the DGGE profiles of bacterial and AOB 16S rRNA gene fragments from the filter beds and the BAF reactor revealed some different populations in different sections of each of the reactors and differences between the reactors (Figs. 1 and 2).
Bacterial and AOB DGGE data for the BAF suggest that the conditions within the reactor differentially select for some different populations (Figs. 1 and 2, lanes 7–9). The BAF reactor is composed of three linked basins each of which is optimised for different processes. A change in the level of aeration appears to have an effect on the bacterial populations present. This is evident by a visual comparison of the banding pattern of each of the three stages (Figs. 1 and 2, lanes 7–9).
Considerable differences in bacterial and AOB DGGE data were also observed between different depths of the filter bed and between the primary and secondary filter bed (Figs. 1 and 2, lanes 1–6). For example, more bands were detected in samples from the bottom of the primary filter (Fig. 1, lane 3) and in the secondary filter bed (Fig. 1, lanes 4–6) than at the top of the primary filter (Fig. 1, lanes 1 and 2). Furthermore, particularly in the AOB DGGE profiles, there appears to be a successional change in bacterial populations through the filter beds. This is apparent by the loss of some bands down the filter bed profile and the appearance of others (Fig. 1, lanes 1–6).
A comparison of the DGGE profiles between the filter beds and the BAF reactor revealed that although both reactors were fed the same waste they harboured distinct bacterial and AOB populations (Figs. 1 and 2). The BAF reactor harboured a lower detectable diversity of AOB compared with the filter beds. In particular, a number of bands that migrated further in the DGGE gel were noted in samples from the filter beds that were not present in the BAF reactor. Nevertheless both reactors appeared to have a common predominant population (marked X on Fig. 1). To confirm the identity of AOB represented by bands in DGGE gels PCR-amplified 16S rDNA from the WWTP samples was cloned and sequenced.
3.2Characterisation of AOB in the BAF and trickling filter reactors
Clone libraries of betaproteobacterial AOB 16S rRNA genes were constructed from samples from each of the reactors selected as those containing the greatest diversity on the basis of DGGE profiles (top and bottom, 0 and 1.9 m depth, of the secondary filter bed) or exhibiting active nitrification (the nitrification unit of the BAF). Two methods, DGGE and ARDRA, were employed to screen 30 clones from each sample. Both methods resulted in similar groupings of clones; however, DGGE proved more discriminatory than ARDRA, hence ARDRA appeared to underestimate the AOB diversity in these reactors.
The DGGE screening of the clone libraries indicated that sequences that co-migrated with most of the predominant bands from the original DGGE analysis were recovered in clone libraries (Fig. 3). However, some components of the DGGE profiles were not recovered in our screening of the clone libraries. Conversely some clones present at low frequencies in the clone libraries did not have a corresponding band in the DGGE gel (e.g. clone 19Fb). Nevertheless, the intensity of most of the bands reflected the frequency of different clones in the libraries (e.g. 45 BAF dominant DGGE band and clone Figs. 3 and 4).
Nucleotide sequences were determined for each clone type from the clone libraries and were compared to the GenBank database using FASTA 3 . All AOB sequences recovered had between 96 and 99% identity with previously identified AOB 16S rRNA gene sequences with the exception of clones 21Fb and 55Fb which had 95% identity with N. cryotolerans Nm55 and an environmental clone related to Nitrosomonas spp. respectively (Table 2).
Table 2. Nearest neighbour of the cloned sequences from the top and bottom of the secondary trickling filter and the nitrification unit of the BAF reactor
aFt – top of filter, Fb – bottom filter, BAF – biological aerated filter.
A detailed analysis of the sequences recovered from the BAF and filter bed revealed that all were derived from the Betaproteobacteria (except clone 1BAF which grouped most closely, but with low identity (92%) to an uncultured Gammaproteobacterium). Furthermore, most of the sequences (97% BAF, and 80% top filter and 85% bottom filter) were most closely related to the betaproteobacterial ammonia oxidisers. The rest of the sequences recovered were usually closely related to either Thauera spp. (e.g. clone 47Ft) or Dechlorimonas agitatus-like bacteria (e.g. clone 43Ft, Table 2).
AOB-like sequences from the nitrification unit of the BAF and top and bottom of the filter bed reactors were all recovered with the genus Nitrosomonas (Fig. 5). The clone library from the BAF reactor had a preponderance of N. mobilis-like sequences (clone 45BAF; 90%, n=30, Fig. 4 and Table 2) (Nitrosomonas cluster 7). Whereas the filter bed samples were more diverse and three or four different predominant clone types were identified; N. mobilis-like sequences, (clone 25Ft, 33%; clone 36Fb, 27%; n=30), Nitrosomonas spp. cluster 6b sequences (clone 50Ft, 10%; clone 11Fb, 23%; n=30), cluster 6a sequences (clones 19Fb, 10%; 21Fb, 7%; 26Ft, 23%; 17Ft, 14%; n=30) and Nitrosomonas spp. Cluster 5 sequences (clone 55Fb, 20%; n=30, Fig. 4 and Table 2). The sequences most closely related to cluster 6a (50Ft, 11Fb), 6b (21Fb) and 5 (54Fb, 55Fb) had relatively low sequence identity with the most similar sequences in the sequence databases and their association with these groups based on bootstrap analysis was not robust. This may be a consequence of the relatively short sequences used in the analysis. Analysis of longer 16S rRNA gene sequences would help place these sequences more confidently with one of the recognised AOB clusters. The nearest neighbours to the sequences recovered were originally reported from a range of environments but a number group with sequences from halotolerant AOB or sequences recovered from saline environments or enrichments (36Fb-N. mobilis and 55Fb-clone AEM-5). N. mobilis-like sequences were found in all the reactors, but were proportionally more abundant in the clone library from the BAF (90% of sequences) than in the filter beds (33% top of secondary, 27% bottom of secondary, Fig. 4 and Table 2). It should be noted that the AOB-selective primers that were used in this study are not completely universal for all betaproteobacterial AOB and primer CTO189f and CTO654r have mismatches with some sequences from Nitrosomonas cluster 7 and cluster 6a, and CTO654r has more than two mismatches with all cultured members of the N. communis cluster (cluster 8) . This might to some extent explain why we did not detect any sequences closely related to Nitrosomonas cluster 8, while we did detect sequences that were related to all other currently described Nitrosomonas clusters, albeit in some cases with relatively low sequence identity. However, in clone libraries generated from some of the samples analysed in this study, using primers Nso190 and Nso1225, which have no mismatches with the N. communis cluster , no members of this group were detected (unpublished data).
The cloning and sequencing analysis indicated that comparable levels of AOB diversity were identified when the amplified rRNA gene fragments were analysed by DGGE or cloning. DGGE and sequencing analysis therefore appeared to indicate that particular AOB were selected for in the different reactors. To examine whether the selection of particular AOB in the different reactors was statistically significant Raup and Crick simulations  were employed.
3.3Statistical analysis of DGGE gels
In our study we detected N. mobilis-like organisms as an important component of the AOB communities in all samples examined. It is tempting to impose an interpretation when such patterns are observed. For example, saline wastewaters favour the occurrence of N. mobilis. However, it is inappropriate to make an interpretation of ecological data until we have established that the patterns we observe cannot be explained by chance alone . It is therefore necessary first to show that the patterns we observe cannot be attributed to the random distribution of species before inferring deeper causal relationships between the environmental and physiological factors and the occurrence of specific organisms. For this reason, we tested the null hypothesis that the similarity between patterns observed in our DGGE data from different samples could be explained by chance alone (Figs. 6–8).
3.4Comparison of AOB communities down the profile of the trickling filter beds
The similarity between the total pooled AOB data and any given sampling point in the trickling filter was no greater than could be accounted for by chance alone (P=0.501). Neither was there any significant similarity between the top of the reactor and the subsequent depths (Fig. 6). Taken in isolation these data suggest that the similarities in the AOB communities present at different depths are no greater than can be accounted for by chance. However, regression analysis demonstrated a statistically significant (P=0.006) relationship between the depth in the reactor and the SRC value.
Moreover, adjacent sampling points were similar (SRC>0.95; Fig. 7) suggesting that selection for different AOB with depth in the filter was non-random and the similarities in AOB communities were statistically greater than could be expected from chance matching of bands in the DGGE profiles. Clearly there is a significant gradient in AOB diversity down the reactor. A similar pattern of succession was also observed in the bacterial DGGE profiles (Figs. 6 and 7). This may be due to the stratification that exists in unmixed biofilm reactors as the waste is degraded and changes in composition as it passes through the filter bed . The waste entering the primary filter bed will contain a high amount of organic matter, and as the waste is degraded the level of organic matter will be reduced and concomitantly oxygen will be consumed. In the presence of high levels of organic matter heterotrophs can out-compete autotrophic ammonia oxidisers for oxygen [4,41] and ammonia [42,43]. As the organic matter declines, the number of heterotrophs decreases and consequently AOB can proliferate. Hence there should be a higher number of AOB at the base of the primary filter bed and in the secondary filter bed. AOB were not quantified in this study but operational data suggest that the secondary filter beds remove more ammonia than the primary beds (04/01/99 to 16/08/01 NH4+ removal – primary 45%±9.0; secondary 87%±4; mean±95% CI); shortly prior to sampling primary 74%; secondary 81%) suggesting the presence of a higher number or metabolically more active AOB in the secondary filter.
Interestingly, no statistically significant differences or gradients were observed in the BAF even though this is also a fixed film reactor. This may be, at least in part, attributable to the more homogenous conditions engendered by the vigorous mixing and recycle regime in the BAF reactor.
3.5Comparison of AOB communities between the BAF and trickling filter reactors
Comparison of DGGE data from the BAF and trickling filter revealed that for all but two of the reactor comparisons the similarities observed were no greater than would be expected by random association of bands in the DGGE profiles (0.05<SRC<0.95; Fig. 8). Therefore any differences observed between the BAF and filter beds could be explained by random processes without the need to invoke differences in reactor design as a causal factor.
Nevertheless, the similarity values for each comparison of the BAF with the primary filter bed were generally all very high (SRC>0.943) whereas the values for each comparison of the BAF with the secondary filter bed were lower and successively decreased with depth (e.g. BAF nitrification B vs St, SRC=0.734; B vs Sm SRC=0.473; B vs Sb SRC=0.083; Fig. 8).
Although not statistically significant on the basis of the Raup and Crick analysis, it is apparent that the similarities between the BAF and the primary filter bed are considerably greater than the similarities between the BAF and the secondary filter bed. This may reflect the fact that the primary filter and the BAF reactor receive identical influent whereas the influent to the secondary filter is the effluent from the primary filter and as such the composition will be different.
Interestingly, reports from practitioners suggest that the filter beds are more robust and suffer less from nitrification failure than other systems . The comparison of pooled DGGE data from the BAF with the secondary filter bed suggested that the AOB populations in the BAF and secondary filter bed were significantly divergent (SRC=0.044). This was reflected in the extent of AOB diversity within each plant. The filter beds had a greater AOB diversity (four or five different AOB identified) than the BAF (two different AOB identified). This could explain why the filter beds perform better and are inherently more stable than the BAF reactor.
It has been suggested that the level of AOB diversity within a WWTP has a major influence on process stability ; the greater the diversity, the more stable the process. A plant with greater diversity should cope better with changing conditions since a reduction in numbers of one organism may not mean process failure as other organisms better adapted to the new conditions may proliferate resulting in a more functionally stable system. Our data support this notion since the filter beds where AOB diversity was greater are the inherently more stable reactors. If diversity does play a major role in wastewater treatment then as previously suggested by Daims and his co-workers , the engineering of plant to have higher diversity may make processes such as nitrification more stable. However, the reason why different AOB populations are present under different conditions must be understood to permit the development of engineering solutions to obtain the most suitable diverse AOB population in WWTPs.
The observation of gradients in the diversity of the AOB communities suggests that environmental selection is partly influencing the diversity observed. However, an N. mobilis-like organism appears to be ubiquitous at this site. This must mean that either this organism is well adapted to the environmental conditions at all points in all reactors or it is so abundant in the source community that it appears, by chance, at all locations . Traditionally microbial ecologists have taken a more deterministic stance . However, we cannot distinguish between these two possibilities in this, or other, purely observational, studies. It is likely a number of mechanisms are at work and the relative importance of each mechanism will vary. In our study for example, the N. mobilis-like AOB would be dominant at all depths in the filter bed, if community assembly was random and the N. mobilis-like AOB were abundant in the source population. In reality, other AOB sequences increase in proportional abundance deeper in the reactor, an observation consistent with deterministic selection. The reactors studied here were fed with effluent from the same HRAS reactor. Interestingly, although we have no data on the composition of the AOB communities in the effluent from the HRAS reactor, analysis of cloned 16S rRNA gene sequences obtained from this reactor using AOB-selective primers indicated that the clone libraries contained a low frequency of AOB sequences (26%, n=30), but those that were identified were most closely related to N. mobilis (data not shown).
In this study we have shown important and significant differences in the AOB diversity of two full-scale WWTP of different conformation. The differences in diversity may very well be related to the differences in performance. The challenge now is to elucidate the mechanism underlying the differences so that these mechanisms may be incorporated into WWTP design.
We are grateful to AstraZeneca Global SHE and Yorkshire Water Services for funding of the project and supplying samples. We would also thank the anonymous reviewers for their thoughtful comments which helped improve the manuscript.