Which are the polyphosphate accumulating organisms in full-scale activated sludge enhanced biological phosphate removal systems in Australia?

Authors


R.J. Seviour, Biotechnology Research Centre, La Trobe University, Bendigo, Vic., Australia (e-mail: r.seviour@latrobe.edu.au).

Abstract

Aims:  To see if the compositions of the microbial communities in full scale enhanced biological phosphorus removal activated sludge systems were the same as those from laboratory scale sequencing batch reactors fed a synthetic sewage.

Methods:  Biomass samples taken from nine full scale enhanced biological phosphate removal (EBPR) activated sludge plants in the eastern states of Australia were analysed for their populations of polyphosphate (polyP)-accumulating organisms (PAO) using semi-quantitative fluorescence in situ hybridization (FISH) in combination with DAPI (4′-6-diamidino-2-phenylindole) staining for polyP.

Results:  Very few betaproteobacterial Rhodocyclus related organisms could be detected by FISH in most of the plants examined, and even where present, not all these cells even within a single cluster, stained positively for polyP with DAPI. In some plants in samples from aerobic reactors the Actinobacteria dominated populations containing polyP.

Conclusions:  The PAO populations in full-scale EBPR systems often differ to those seen in laboratory scale reactors fed artificial sewage, and Rhodocyclus related organisms, dominating these latter communities may not be as important in full-scale systems. Instead Actinobacteria may be the major PAO.

Significance and Impact of the Study:  These findings illustrate how little is still known about the microbial ecology of EBPR processes and that more emphasis should now be placed on analysis of full-scale plants if microbiological methods are to be applied to monitoring their performances.

Introduction

Many full-scale activated sludge systems designed to remove phosphorus microbiologically by the process of enhanced biological phosphorus removal (EBPR) operate around the world. Yet with a few exceptions (Crocetti et al. 2000; Zilles et al. 2002; Saunders et al. 2003; Kong et al. 2004,2005) our current understanding of their microbiology comes largely from data generated from studies with laboratory scale or pilot plant systems (Seviour et al. 2003). These have usually been run under artificially controlled operating conditions and fed a synthetic sewage containing only a single energy source, usually acetate. Hence the selective pressures experienced by these microbial communities are markedly different to those likely to be encountered in full-scale systems (Juretschko et al. 2002; Loy et al. 2002; Seviour et al. 2003; Kong et al. 2004,2005).

A combination of culture independent 16S rRNA based techniques used with many of these laboratory systems suggests that the gammaproteobacterial Acinetobacter spp., the early favourite candidates as polyphosphate (polyP)-accumulating organisms (PAO), are not as important as earlier data from culture dependent methods had concluded (Seviour et al. 2003). Instead, Betaproteobacteria, closely related to members of the genus Rhodocyclus appear to dominate many such EBPR communities, especially those in acetate-fed anaerobic: aerobic sequencing batch reactors (SBRs) (Bond et al. 1999; Hesselmann et al. 1999; Crocetti et al. 2000; Kong et al. 2002a). However, other populations including members of the Alphaproteobacteria (Christensson et al. 1998; Kawaharasaki et al. 1999; Crocetti et al. 2000; Levantesi et al. 2002), Actinobacteria (Bond et al. 1999; Hesselmann et al. 1999; Crocetti et al. 2000; Liu et al. 2000; Lee et al. 2003), and the Bacteroidetes (Dabert et al. 2001) and ‘Chloroflexi’ (Björnsson et al. 2002) divisions have all been detected in large numbers by fluorescence in situ hybridization (FISH) in such systems. These populations generally have not been identified further, and whether they were accumulating polyP there has rarely been demonstrated.

Even in laboratory scale systems, complex feeds can change markedly the EBPR community composition (Hesselmann et al. 1999), but how other operational or plant configurational parameters may influence their population dynamics is largely unknown. Eschenhagen et al. (2003) indicated that the levels of biodiversity in the communities from a Phoredox configured plant (Seviour and Blackall 1999) were lower by T-RFLP but not FISH analyses than those in an A/O configured plant community. These plant communities also differed in their population structures, and the putative PAO in both communities were dominated more by actinobacterial Tetrasphaera spp. (Maszenan et al. 2000b) than by the Rhodocyclus related organisms. Lee et al. (2003) also showed that EBPR communities differed quantitatively between those with and without nitrogen removal, although Betaproteobacteria including Rhodocyclus related organisms dominated both.

Relatively few published studies deal with full-scale EBPR communities. While Crocetti et al. (2000) showed a direct parallel between the numbers of Rhodocyclus related organisms and EBPR capacity in a full-scale plant, Zilles et al. (2002) provided evidence that both the numbers and proportions of Rhodocyclus-related organisms in the total PAO community were much higher in a University of Cape Town (UCT) configured plant (Seviour and Blackall 1999) than in an aerobic: anoxic plant lacking a true anaerobic zone. Furthermore, about half of the PAO cells in the latter community appeared to lack intracellular polyP, and Zilles et al. (2002) considered these played no role in EBPR. Both Kong et al. (2004) and Wong et al. (2005) showed that numbers of Rhodocyclus related organisms varied in communities from different full-scale EBPR plants operating with the same and different configurations. From FISH/micro-autoradiography data, these populations behaved as expected of PAO from the biochemical models (Kong et al. 2004). Again, in both studies, not all cells of the Rhodocyclus related organisms stained positively for polyP even within a single PAO cluster. Later, (Kong et al. 2005), they presented evidence that Actinobacteria closely related to Tetrasphaera were behaving as PAO in some full-scale systems, but with an in situ physiology quite different to that seen with the Rhodocyclus related PAO.

Since little else is known about full-scale EBPR communities, this study was undertaken using semi-quantitative FISH analyses to see which PAO were dominant in full-scale plants of different configuration, and especially whether those which appear to be the major PAO in laboratory scale systems (Loy et al. 2002; Seviour et al. 2003), were represented equally there. Instead of following a single EBPR plant community over time, it was decided instead to look at a range of EBPR plants of different configurations, although only a single sample of biomass was examined from each, an approach similar to that adopted in other studies (Kong et al. 2004). Such information is considered useful if microbiological parameters are ever to be the preferred way to monitor the performance of full-scale EBPR plants.

Materials and methods

EBPR plants investigated in study

A range of full-scale EBPR plants treating mainly domestic wastes was used in this study, differing in their configurations (Seviour and Blackall 1999) and their EBPR capacities, measured in terms of their final effluent P levels. Chemical data given were supplied by the individual plant staff, but were usually generated from nationally certified analytical laboratories. The main operational features of these plants are summarized in Table 1, and while their final effluent phosphorus levels varied, in most cases it was below 1·5 mg l−1.

Table 1.  Operational data of EBPR activated sludge plants sampled in this study
PlantABCDEFGHJ
  1. na, data not available.

ConfigurationModified UCT3 stage bardenphoCarousel/ UCT3 stage bardenphoModified UCTCarousel/ UCTModified UCTModified UCTModified UCT
Sludge age (days)9–122521nana2525na16
PrefermentationYesYesNoneYesYesYesNoneNoneNone
Agitation/aerationSubmergedMechanicalSubmergedSubmergedSubmergedMechanicalMechanicalSubmergedSubmerged
SystemDiffusersSurfaceTurbinesDiffusersDiffusersSurfaceSurfaceDiffusersDiffusers
Phosphorus in (mg l−1)9·413·512107·56101212
Phosphorus out (mg l−1)4·50.50·11·20·21·5>20·22
Phosphorus removal capacity (mg l−1)4·91311·98·87·34·5<811·810
N removal?YesYesYesYesYesYesVariableYesVariable
Operating temperature2525232526na171619
Equivalent persons80 00025 00060 00032 00044 00025 000120 00030 00060 000

Microscopic examination of biomass samples

Biomass samples were collected during October 2002 from the effluent leaving the aerobic and anaerobic zones of these plants. They were immediately fixed in either paraformaldehyde for gram-negative organisms or absolute ethanol for gram-positive organisms (Amann et al. 1990) for subsequent FISH analysis.

Community FISH analyses of samples from the aerobic reactors of the plants studied were carried out with the FISH probes given in Table 2. The conditions for hybridization stringencies were those described in the original publications. Quantification of each population fluorescing with a specified probe and/or stain was determined using images captured with a Nikon PMX1200 digital camera attached to a Nikon E800 microscope with the appropriate filters as described by Beer et al. (2004). Briefly, for each probed sample, at least 10 frames, with an estimated total of >2000 bacteria present there were captured at random, similar to the analytical protocol used in other similar studies (e.g. Kong et al. 2004). For each frame, three identical fields were captured. The first field had been stained with SYBR-gold to determine the total area of nucleic acid containing cells, the second allowed the total cell area responding to each of the targeted probes used in this study to be determined, and in the third field the cell areas fluorescing with DAPI stain (i.e. the polyP accumulating bacteria) in samples to which these individual FISH probes were applied was calculated. For most samples, replicate analyses were carried out. All images were converted into binary images using Adobe PhotoShop Windows version 7.0, and the fluorescing cell areas measured using Scion Image beta 4.0.2 Windows version (Scion Frederick, Washington DC, USA). All values given are expressed as percentages of the total cell areas fluorescing with SYBR-gold. FISH probes were labelled with the Cy3 fluorochrome and all were purchased from ProOligo. Combined FISH/nile blue A staining to identify those cells accumulating poly β-hydroxyalkanoates (PHA) was also performed (Liu et al. 2001). Methylene blue stains were carried out on some samples (see Results) on unfixed biomass from aerobic reactors (Beer et al. 2004) prior to FISH analyses, and metachromatically stained cells were taken to indicate their polyP accumulation. Fixation was then carried out as described above for FISH analysis on biomass already attached to a slide.

Table 2.  16S/23S rRNA targeted oligonucleotide probes used in this study
TargetProbe nameSequence (5′-3′)F(%)Reference
  1. *Applied as a mixture, EUBmix.

  2. †Applied as a mixture, PAOmix.

  3. ‡Applied as a mixture, LGCmix.

Most bacteriaEUB338-1*GCTGCCTCCCGTAGGAGT35Amann et al. 1990
PlanctomycetalesEUB338-II*GCAGCCACCCGTAGGTGT35Daims et al. 1999
VerrucomicrobialesEUB338-III*GCTGCCACCCGTAGGTGT35Daims et al. 1999
AlphaproteobacteriaALF968GGTAAGGTTCTGCGCGTT20Neef et al. 1999
Amaricoccus spp.MAR839CTGCGACACCGAACGGCAAGCC20Maszenan et al. 2000a
Putative SBR GAOSBR9-1aCAACCTGGGAACTTGCGCTT35Beer et al. 2004
BetaproteobacteriaBET42aGCCTTCCCACTTCGTTT35Manz et al. 1992
CompetitorBET-compGCCTTCCCACATCGTTT35Manz et al. 1992
Candidatus Accumulibacter phosphatisPAO651†CCGTCATCTACWCAGGGTATTAC35Crocetti et al. 2000
Candidatus Accumulibacter phosphatesPAO846†CCCTCTGCCAAACTCCAG35Crocetti et al. 2000
Rhodocyclus related PAORHC175TGCTCACAGAATATGCGG20Hesselmann et al. 1999
GammaproteobacteriaGAM42aGCCTTCCCACATCGTTT35Manz et al. 1992
CompetitorGAM-compGCCTTCCCACTTCGTTT35Manz et al. 1992
Candidatus Competibacter phosphatisGAOQ431TCCCCGCCTAAAGGGCTT35Crocetti et al. 2002
Candidatus Competibacter phosphatisGAOQ989TTCCCCGGATGTCAAGGC35Crocetti et al. 2002
Putative GAOGB_G2TTCCCCAGATGTCAAGGC35Kong et al. 2002b
Putative GAOGB_4GGCTCCTTGCGGCACCGT35Kong et al. 2002b
Acinetobacter spp.ACA23aATCCTCTCCCATACTCTA35Wagner et al. 1994
ActinobacteriaHGC69aTATAGTTACCACCGCCGT25Roller et al. 1994
Actinobacterial PAOActino-221CGCAGGTCCATCCAGAC30Kong et al. 2005
Actinobacterial PAOActino-658TCCGGTCTCCCCTACCAT40Kong et al. 2005
Micropruina glycogenicaMIC184CATTCCTCAAGTCTGCC20Kong et al. 2001
Microlunatus phosphatusMIC179GAGCAAGCTCTTCTGAAACCG10Kawaharasaki et al. 1998
Tetrasphaera japonica/australiensisTET63GCTCCAGGGTCACCGTTC20Kong et al. 2001
Tetrasphaera elongataActino1011TTGCGGGGCACCCATCTCT20Liu et al. 2001
Non-EBPR cloneKSBR531TTTCACTCCCGACGCAACAGAC20Kong et al. 2001
Most FirmicutesLGC354A‡TGGAAGATTCCCTACTGC35Meier et al. 1999
Most FirmicutesLGC354B‡CGGAAGATTCCCTACTGC35Meier et al. 1999
Most FirmicutesLGC354C‡CCGAAGATTCCCTACTGC35Meier et al. 1999
BacteroidetesCF319ATGGTCCGTGTCTCAGTAC35Manz et al. 1992
ChloroflexiCFX1223CCATTGTAGCGTGTGTGTMG35Björnsson et al. 2002
Type 1851CHL1851AATTCCACGAACCTCTGCCA30Beer et al. 2002
  MIC179   

Results

Influent and effluent measured P values would suggest that most of these plants (with the possible exception of plants A and F) were functioning as EBPR systems at the time samples were taken for analyses, since their biomasses were removing >7 mg l−1 phosphate. However, sludge phosphate levels were not determined.

FISH analyses of EBPR plant biomass samples

Application of the EUBmix FISH probes to biomass samples from the aerobic reactors of these plants showed that about 82–97% of the cell area fluorescing with the SYBR-gold stain also fluoresced with the EUBmix probes (Table 3), similar to the values obtained in most other studies with activated sludge communities (Seviour et al. 2003). One exception was with plant B, where only approximately 70% of the SYBR-gold positive cell area fluoresced with the EUBmix probe. Such a low response suggests that either many cells were metabolically inactive, or the presence of many organisms lacking the EUBmix target sites of members of the Bacteria, because these SYBR-gold positive signals did not derive from eukaryotic organisms like protozoa in this sample. Although Archaea have been reported in activated sludge systems (Gray et al. 2002), they were not probed for here. Community members of the proteobacterial subdivisions and Actinobacteria were well represented in all plants examined here (Table 3). Few or no members of the Firmicutes were detected in any of the biomass samples, and Meier et al. (1999) have suggested from their earlier data that these are frequent but always minor components (usually <3%) of conventional activated sludge communities (Table 3).

Table 3.  Semi-quantitative FISH analysis of samples of biomass from EBPR full-scale plants
Target organismABCDEFGHJ
XYXYXYXYXYXYXYXYXY
  1. FISH, fluorescence in situ hybridization; nd, none detected.

  2. All data are presented as area percentages of SYBR-gold positive cells. The standard deviations are given after the ± symbol.

  3. X, values refer to the area percentages of the SYBR-gold positive cell area for each plant sample, which hybridized with each of the designated FISH probes.

  4. Y, values refer to the area percentages of the SYBR-gold positive cells for each of the designated probes, which then stained positively with DAPI for polyP.

  5. The following probes hybridized with <1% of the area of the biomass area staining with SYBR-gold: LGCmix, KSBR531, RHC175, GB_G2, GB_4, MIC184, TET63, Actino-221 and Actino-658.

All bacteria89 ± 69 ± 370 ± 2514 ± 582 ± 629 ± 683 ± 720 ± 886 ± 89 ± 383 ± 1030 ± 892 ± 617 ± 997 ± 222 ± 1089 ± 89 ± 6
Alphaproteobacteria9 ± 4019 ± 6<129 ± 128 ± 316 ± 8057 ± 13031 ± 15<129 ± 11<129 ± 9<129 ± 9<1
SBR 9-1 clone<5013 ± 11011 ± 50<50<10nd24 ± 133 ± 2<50<50
Amaricoccus spp.<1<1nd<10ndnd<10ndnd9 ± 2<1
Betaproteobacteria15 ± 610 ± 44 ± 3<1<5<113 ± 98 ± 3<1<15 ± 4<58 ± 4<516 ± 115 ± 312 ± 83 ± 2
Accumulibacter phosphatis<5<5<1<1nd<1<1<1<1<5<59 ± 65 ± 45 ± 34 ± 3nd
Gammaproteobacteria12 ± 48 ± 5<5<113 ± 66 ± 26 ± 4<15 ± 407 ± 33 ± 35 ± 2<17 ± 4<515 ± 73 ± 2
Acinetobacter spp.ndnd<508 ± 70<50<10nd <10<10<10
Bacteroidetes<5<1<5011 ± 5<5<5011 ± 8<59 ± 7<1<5013 ± 3<1<50
Actinobacteria7 ± 34 ± 316 ± 811 ± 318 ± 710 ± 535 ± 1225 ± 619 ± 710 ± 413 ± 57 ± 311 ± 74 ± 213 ± 87 ± 36 ± 54 ± 2
Microlunatus phosphorus<5<5<5<5<5<5<10<5<5<5<5<1<1<10<5<5
Tetrasphaera elongata<1<1ndnd<10<5<1ndndndnd
Firmicutes<50<5<1<10<1<1<1<1nd<50<10<50
Chloroflexi22 ± 60<10<508 ± 4025 ± 12020 ± 90<50<5013 ± 60
Eikelboom type 1851<50nd<10<5013 ± 708 ± 30ndnd<10
DAPI +ive cell area before FISH5 ± 1430 ± 1225 ± 921 ± 1011 ± 715 ± 920 ± 726 ± 710 ± 9
DAPI +ive cell area after FISH33 ± 1320 ± 718 ± 624 ± 1010 ± 418 ± 1217 ± 720 ± 711 ± 9

What proportion of these cells contained polyP?

The specificity and sensitivity of DAPI (4′-6-diamidino-2-phenylindole) staining in detecting intracellular polyP is still unclear. However, Rodrigues et al. (2002) have suggested that polyP containing Pi from <50 to over 800 orthophosphate chain length will be detected. Furthermore, other studies have employed DAPI to detect intracellular polyP in activated sludge populations, sometimes in combination with FISH or flow-cytometry (Kawaharasaki et al. 1999; Liu et al. 2001; Tsai and Liu 2002; Lee et al. 2003; Kong et al. 2004,2005).

When biomass samples from the aerobic reactors of the plants were stained for polyP with DAPI, the percentage of the biomass which was DAPI positive varied from 10% of the SYBR-gold positive area in plant J to 35% in plant A. Combining FISH with DAPI staining appeared to underestimate the percentages in some samples of the DAPI positive cell areas detected, although given the errors inherent in such quantifications, these differences in most cases were considered very small (Table 3). There appeared to be no consistent correspondence between values for DAPI-positive bacterial cell areas and the EBPR capacity of the different PAO communities based on phosphate removal levels. This is to be expected if the polyP storage capacities of individual cells in the same population and of cells in different PAO populations vary. Thus, although the percentage of area of the PAO in the total EUB mix fluorescing area in plant A was high at 34%, the amount of P removed was much lower than in other plants with lower DAPI positive percentages of the EUB mix positive cell areas in their communities (Table 3). These data are summarized in Fig. 1. Bearing in mind that some individual bacterial cells can contain very high polyP levels, the percentage of PAO in the total community may only need to be at the lower end of the range found in this survey for effective EBPR capacity (Hesselmann et al. 1999; Loy et al. 2002). It is also possible that some cells with high polyP contents may no longer be sufficiently metabolically active to respond to FISH probing.

Figure 1.

Phosphorus removal performance in mg/L (bsl00011) of each of the full-scale EBPR plants with relationship to the percentage of the area of cells fluorescing with SYBR-gold which responded to the EUBmix (bsl00023) the Beta42a (bsl00004) and the HGC69a (bsl00022) FISH probes, and which also stained positively with DAPI for polyP

Which populations are the possible PAO in these communities?

Most of the communities in the aerobic reactors contained very high cell area percentages responding to the FISH probe for Alphaproteobacteria, but with the exception of plant C biomass (Table 3), <1% of these contained polyP after DAPI staining. The major surprise was the very low cell area percentages of the Betaproteobacteria detected by FISH in some of the plants surveyed (usually approximately 5% of the EUB mix positive cell area), although in some plants e.g. A, D, H and J, these values were much higher. However, although reasonably high area percentages of these Betaproteobacteria were DAPI positive, in almost all cases only a small percentage (<5%) of these populations responded to the RHC and PAOmix probes of Hesselmann et al. (1999), Crocetti et al. (2000) and Zilles et al. (2002), respectively, targeting the Rhodocyclus related PAO (Fig. 2). In some aerobic reactor biomass samples, none in fact could be detected with these probes. Of those Rhodocyclus related PAO that did respond to the targeted probes (Fig. 2a,b), not all contained polyP after DAPI staining (Table 3), which agrees with data from other studies with full-scale EBPR plants (Zilles et al. 2002; Kong et al. 2004).

Figure 2.

Light micrographs showing EBPR populations in the full-scale plants after application of FISH probes and histochemical stains. (a) Plant G biomass after hybridization with the PAOmix probes. (b) Same field of view showing DAPI positive cells, indicating presence in some cells of polyP. (c) Plant G biomass after hybridization with the HGC69a probe. (d) Same field of view showing DAPI positive cells. (e) Plant C from the anaerobic reactor biomass after hybridizing with the Acinetobacter spp. ACA23a FISH probe. (f) Same field of view showing nile blue A positive cells, indicating possible presence of PHA (g) Plant C biomass stained with methylene blue, where metachromatically stained cells indicate presence of polyP. (h) Same field of view with the Acinetobacter spp. ACA23a FISH probe. (i) Plant E biomass after hybridizing with the type 1851 specific FISH probe. (j) Same field of view after DAPI staining. Bar, 10 μm in all cases

Most of the communities with high percentages of cells accumulating DAPI positive polyP in samples from the aerobic reactors were dominated instead by Actinobacteria responding to the HGC69a probe. Their area percentages varied from as high as approximately 35% of SYBR-gold positive cell area in plant D to only approximately 7% in plant A, which contained high numbers of DAPI positive Betaproteobacteria and Gammaproteobacteria instead (Table 3). Furthermore, in most cases more than 50% of the cell area of these Actinobacteria contained polyP after DAPI staining (Fig. 2c,d). What the precise identity of these Actinobacteria might be is not yet clear, since none responded to the TET63 probe (Kong et al. 2002a) for Tetrasphaera spp., which were considered a possible PAO candidate from pure culture studies (Maszenan et al. 2000b), and were detected in high numbers in a full-scale Japanese A2O EBPR plant community (Wong et al. 2005). Nor did they respond to the Actino-221 and Actino-658 FISH probes (Table 1) designed by Kong et al. (2005) for their actinobacterial Tetrasphaera related PAO. However, in plant E (Table 3) some cells did fluoresce with the Actino 1011 probe of Liu et al. (2001), thought to detect Tetrasphaera elongata (Hanada et al. 2002), another proposed PAO.

Applying other probes identified some of the Actinobacteria in these samples. Cells were seen (up to 5% of the SYBR-gold positive area) responding to the MIC179 probe designed against Microlunatus phosphovorus (Kawaharasaki et al. 1998), a possible PAO candidate, and often all cells of these contained polyP (Table 3). Their morphology varied from very small cocci in clusters to much larger coccobacilli. It was also clear that many filamentous Actinobacteria including Gordonia amarae-like organisms with right-angled branching, and acutely branched Skermania piniformis-like organisms contained DAPI positive granules (data not shown). Where seen in plant biomasses (Table 1), some of the filamentous bacterial morphotype ‘Candidatus Nostocoida limicola II’, now known to be Tetrasphaera spp. (C. McKenzie et al. unpublished) also contained polyP granules, agreeing with earlier observations with this filament (Liu et al. 2001).

However, when FISH/PHA combined staining was carried out on samples taken from the anaerobic reactor of several plants with high actinobacterial cell area percentages (plants B, D, E and F), none of the Actinobacteria stained positively for PHA. Kong et al. (2005) also failed to detect intracellular PHA with nile blue A staining in their putative actinobacterial PAO either. Neither could these cells assimilate acetate anaerobically, but instead appeared to utilize amino acids as determined with FISH/MAR (Kong et al. 2005). If true generally of the actinobacterial PAO, this might explain why these populations are not seen often in high numbers in laboratory scale SBR systems fed synthetic sewage with acetate. Other studies, where unidentified Actinobacteria were detected by FISH as major populations in EBPR communities (e.g. Lee et al. 2003) did not report whether they were capable of accumulating PHA anaerobically.

Major cell populations in the aerobic reactors of plants examined in this study were shown with FISH in some but not all of these communities (Table 3) to be members of the Gammaproteobacteria and the Bacteroidetes. Thus, in plant J, approximately 15% of the total SYBR-gold positive areas were Gammaproteobacteria. Furthermore in some plants, these appeared to be major polyP accumulating populations (Table 3), and in both plants A and C aerobic biomass samples, many of the Gammaproteobacteria contained polyP. One possibility was that these might include Acinetobacter spp., and when their presence was sought with FISH, they were detected in some aerobic reactor biomass samples. Thus, approximately 8% of the SYBR-gold positive cell area in the sample from plant C responded to the ACA23a probe (Table 1) for members of this genus (Wagner et al. 1994). Some cells were scattered as coccobacilli in pairs through the flocs while others were in the tight clusters so distinctive of EBPR activated sludge biomass (Seviour et al. 2003). In some samples they were in chains, appearing as the filamentous morphotype 1863 (Wagner et al. 1994). However, none of these ACA 23a positive cells in these different arrangements stained positively for polyP with DAPI (Table 3), but when samples taken from the anaerobic zones of these plants were examined after nile blue A staining, some cells in clusters responding to the ACA23a probe appeared to contain PHA (Fig. 2e,f). These results directly contradict the behavioural patterns shown by Acinetobacter spp. in pure culture (Seviour et al. 2003), but it is possible that lipid material other than PHA was stained. There was also no indication of polyP accumulation when cells responding to the ACA23a probe and taken from the aerobic reactor were stained with methylene blue (Fig. 2g,h) (Crocetti et al. 2000).

Björnsson et al. (2002) reported that the ‘Chloroflexi’ were often dominating populations in EBPR systems, including full-scale plants, but not whether they played any role in EBPR. When they were sought here in samples from aerobic reactors with their CFX1223 FISH probe, they were detected in most, and in some cases in large numbers (Table 3), agreeing with these earlier data. Almost all were filamentous (Fig. 2i) and morphologically very similar under the microscope. They appeared to be type 1851-like from their morphology, but often lacking any attached growth, and in fact most fluoresced (Fig. 2i) with the FISH probe designed against this filament (Beer et al. 2002). However, none of the ‘Chloroflexi’ ever stained positively for polyP with DAPI in any sample (Fig. 2j), suggesting they were probably not contributing to EBPR.

What other populations are in these communities?

Fluorescence in situ hybridization probes (Table 2) were also used to see whether other bacterial populations, reported elsewhere in previous EBPR community studies could be detected in aerobic reactors in these full-scale plants. These included Amaricoccus spp. (Maszenan et al. 2000a), examples of the so-called ‘G-bacteria’ or tetrad-forming organisms, the TFO (Tsai and Liu 2002), and considered in some early studies to be responsible possibly for EBPR failure (Seviour et al. 2003). None could be detected, except in plant J, a system often showing variable EBPR capacity (unpublished results).

When FISH analyses were performed with the probes designed against the gammaproteobacterial putative ‘glycogen-accumulating organisms’ (GAO) described by Crocetti et al. (2002) and Kong et al. (2002b)), again very few (<1% of total cell area) were detected in any samples taken from the aerobic reactors. In an earlier study, the actinobacterial Micropruina glycogenica and a coccus (from FISH analysis using a probe designed against a clone sequence, KSBR531), dominated the community of an anaerobic: aerobic reactor fed glucose and showing no EBPR capacity, and the former was held mainly responsible for the high glycogen content of the biomass (Kong et al. 2001). Neither could be detected by FISH in any of the communities examined here. However, TFO cells responding to the SBR9-1 probe (Table 2) were seen in all but plant F. This probe was designed against a possible alphaproteobacterial GAO dominating a community in a laboratory scale SBR with poorer EBPR capacity (Beer et al. 2004).

Discussion

This study has used semi-quantitative FISH analysis to determine the community composition of samples of biomass from nine full-scale EBPR plants, and together with DAPI staining to see which populations were storing polyP in aerobic reactor samples. Only a single sample of biomass was analysed with FISH from each plant. It might be argued that this approach would not accommodate the probability that not all cells would be behaving at the single sampling time in a similar metabolic fashion (Seviour et al. 2003). However, these data are sufficient in our view to suggest that the bacterial communities of full-scale EBPR plants may differ markedly to those developing in laboratory scale anaerobic: aerobic SBRs fed a synthetic sewage containing acetate, and add support to data from other similar recent work (Lee et al. 2003; Kong et al. 2005). This is probably not too surprising considering the complexity of substrates in sewage. However, very few data are available upon which to base such an opinion, a situation, which prompted this present study. Of particular interest were the very low numbers of Rhodocyclus related organisms sometimes detected with the FISH probes designed to target these in aerobic reactors of plants showing high EBPR capacity. These Rhodocyclus percentages were much lower than those detected by FISH in some other studies. However, similar values were reported for the communities in the aerobic: anaerobic full-scale systems examined by Zilles et al. (2002) and Kong et al. (2005). Low percentages of Rhodocyclus related PAO were also reported by Saunders et al. (2003) in their FISH study with several full-scale EBPR systems of different configurations, and were often detected at levels below 10% of the total cells. These values are close to those obtained for Acinetobacter spp. with FISH analysis, which were used to discount its importance as a possible PAO (Wagner et al. 1994). Saunders et al. (2003) did not determine whether some or all of their Rhodocyclus related PAO were actually accumulating polyP, but several other studies would suggest that not all of these organisms always assimilate acetate anaerobically (Lee et al. 2003; Kong et al. 2004) or, as in this and other work, store polyP aerobically (Zilles et al. 2002; Kong et al. 2004; Wong et al. 2005). This may not necessarily mean that these cells are not actively participating in EBPR, since cell physiologies within full-scale plants are not synchronized in the same way they might be in SBR systems. However, variations in DAPI staining responses among individual cells within a single cluster were frequently seen, where such criticism would not apply.

Previous evidence that Actinobacteria are probably important PAO (Eschenhagen et al. 2003; Kong et al. 2005) receives qualified support from the data presented here. However, in the plants examined here their precise identity is still unclear, although they are probably diverse phylogenetically. They failed to fluoresce with the FISH probes used by Kong et al. (2005) for their actinobacterial Tetrasphaera-related PAO, but analysis of 16S rRNA based clone libraries obtained from biomass from for example, plant D containing high numbers of DAPI-positive Actinobacteria may provide clues as to their identity. No evidence supporting their ability to synthesize PHA anaerobically in any of the biomass samples was obtained. This is not conclusive proof that these bacteria may never behave as expected of PAO in terms of the biochemical models (Seviour et al. 2003). Kong et al. (2005) observed that the Tetrasphaera-related PAO seen in large numbers in their full-scale EBPR communities neither synthesized PHA nor assimilated acetate under anaerobic conditions. Clearly their ability for anaerobic substrate assimilation and subsequent aerobic phosphate assimilation and polyP storage (Kong et al. 2005) demonstrates that Actinobacteria are important in EBPR processes, and identification of the anaerobically stored material in these populations requires urgent attention. To define PAO totally on whether they behave as the early biochemical models expect is misleading and fails to take account of an ability for anaerobic synthesis of storage material other than PHA (Kong et al. 2005). One approach to reflect our current incomplete understanding of the biochemistry of these populations is to refer to any organisms, which accumulates phosphorus at levels in excess of those required for growth, and store it as polyP as PAO.

In summary, these results add to the view that EBPR in full-scale plants may not be a consequence primarily of the presence of Rhodocyclus related organisms. They also illustrate the dangers inherent in generalizing at this stage about PAO identity, as well as emphasizing that much still remains to be understood about the microbial ecology of EBPR systems, and many PAO populations probably still await identification (Wong et al. 2005).

Ancillary