Evidence for bacteriophage activity causing community and performance changes in a phosphorus-removal activated sludge


  • Jeremy J. Barr,

    1. Advanced Water Management Centre (AWMC), The University of Queensland (UQ), Brisbane, Qld, Australia
    2. Environmental Biotechnology Cooperative Research Centre, Sydney, Australia
    Search for more papers by this author
  • Frances R. Slater,

    1. Advanced Water Management Centre (AWMC), The University of Queensland (UQ), Brisbane, Qld, Australia
    Search for more papers by this author
  • Toshikazu Fukushima,

    1. Advanced Water Management Centre (AWMC), The University of Queensland (UQ), Brisbane, Qld, Australia
    2. Division of Environmental Studies, Graduate School of Frontier Sciences, The University of Tokyo, Chiba, Japan
    Search for more papers by this author
  • Philip L. Bond

    1. Advanced Water Management Centre (AWMC), The University of Queensland (UQ), Brisbane, Qld, Australia
    2. Environmental Biotechnology Cooperative Research Centre, Sydney, Australia
    Search for more papers by this author

  • Editor: Julian Marchesi

Correspondence: Philip L. Bond, Advanced Water Management Centre (AWMC), The University of Queensland (UQ), Level 4, Gehrmann Building (60), Brisbane, Qld 4072, Australia. Tel.: +61 7 3346 7841; fax: +61 7 3365 4726; e-mail: phil.bond@uq.edu.au


Bacteria are known to play important roles in biogeochemical cycles and biotechnology processes, but little is known about the influence of bacteriophage on these processes. A major impediment to the study of host–bacteriophage interactions is that the bacteria and their bacteriophage are often not available in a pure culture. In this study, we detected an unexpected decline in the phosphorus-removal performance of a granular laboratory-scale wastewater treatment reactor. Investigations by FISH, transmission electron microscopy and proteomics led us to hypothesize that a bacteriophage infection of the uncultured Candidatus‘Accumulibacter phosphatis’ was responsible for the decline in performance. Further experiments demonstrated that the addition of a putative bacteriophage-rich supernatant, obtained from the previous failed reactor to phosphorus-removal reactors, caused a decrease in the abundance of Accumulibacter in both granular and floccular activated sludges. This coincided with increases in bacteriophage-like particles and declining phosphorus-removal performance. The granular sludge did not recover after the attack, but the floccular sludge regained Accumulibacter numbers and phosphorus-removal performance. These findings suggest that bacteriophage may play a significant role in determining the structure and function of bacterial communities in activated sludges.


There is increasing awareness of the potential of bacteriophage to affect microbial community composition and performance. In natural microbial systems, such as sea water and fresh water, bacteriophage have been shown to play major roles in biogeochemical cycles, community dynamics and genetic diversity through cell lysis and horizontal gene transfer (Paul et al., 1991; Fuhrman, 1999; Jardillier et al., 2005; Suttle, 2007). The majority of these attacks on bacterial communities appear to affect primarily the dominating bacterial population, known as ‘kill the winner’ attacks (Weinbauer, 2004; Weinbauer & Rassoulzadegan, 2004). There is much interest in understanding the bacteriophage–host interactions that shape these ecosystems, especially in relation to population dynamics, coevolution, lysogeny, host resistance and nutrient cycling in food webs. Much knowledge to date has been derived from studying pure cultures, especially coliphages and their hosts, and more recently, the marine cyanobacteria Synechococcus and their bacteriophage (Paul, 2008). Recently, artificially mixed cultures of Synechococcus, their specific bacteriophage and heterotrophic bacteria were studied to examine interactions in a simplified ecosystem (Lennon & Martiny, 2008). Other investigations of phosphorus-removal reactors and membrane bioreactors treating industrial wastewater demonstrated that bacteriophage–host interactions occur within complex microbial communities over an extended period of time (Lee et al., 2004; Shapiro et al., 2010). However, due to complex microbial interactions, the lack of pure-culture hosts and the lack of techniques, there is still a very limited understanding of host–bacteriophage interactions in mixed-culture systems.

Enhanced biological phosphorus removal (EBPR) is a widely applied wastewater treatment process for the removal of phosphorus and carbon. This removal occurs through the intracellular accumulation of phosphate by polyphosphate-accumulating organisms (PAO) (Hesselmann et al., 1999; Liu et al., 2001; Gu et al., 2008). The metabolism of PAO is unique in that carbon uptake occurs anaerobically and is coupled with the energy-rich release of a phosphate group from stored, intracellular polyphosphate. The stored carbon is then utilized aerobically through the tricarboxylic acid cycle for biomass growth, while a portion is used for excess uptake of external phosphate and regeneration of intracellular polyphosphate stores (Mino et al., 1987; Oehmen et al., 2007).

Culture-independent methods have identified the Rhodocyclus-related Betaproteobacteria, Candidatus‘Accumulibacter phosphatis’ (henceforth referred to as Accumulibacter) as a major PAO responsible for EBPR (Bond et al., 1995; Hesselmann et al., 1999; Crocetti et al., 2000; He et al., 2007; Gu et al., 2008; Barr et al., 2010a). Recently, the metagenomes of two laboratory-scale, Accumulibacter-enriched EBPR sludges were sequenced (García Martín et al., 2006). These metagenomic sequences provide a much-needed blueprint for a systems-level understanding of EBPR, have confirmed metabolic and biochemical pathways and facilitated in situ analysis of the transcriptome, proteome and metabolome (He et al., 2007, 2010; Wilmes et al., 2008). Further analysis of the metagenomes revealed a high degree of similarity between the dominant Accumulibacter strains (García Martín et al., 2006). However, there was a large degree of variability between extracellular polymeric substance gene cassettes and clustered regularly interspaced short palindromic repeat (CRISPR) elements, both of which have been demonstrated to be involved with host defence against bacteriophage predation (Sutherland, 2001; Barrangou et al., 2007; Kunin et al., 2008).

Typically, full-scale EBPR systems operate as floccular activated sludge, in which small biofilm aggregates (30–200 μm) of microorganisms treat wastewater (de Kreuk et al., 2007). However, there has been recent interest in operating activated sludge as aerobic granules, which are larger biofilm aggregates (200–2000 μm) (De Bruin et al., 2004; Liu & Tay, 2004; de Kreuk et al., 2007). Granular sludge has been shown to exhibit faster settling characteristics and higher biomass concentrations compared with floccular sludge, making it operationally and economically advantageous (Morgenroth et al., 1997; Beun et al., 1999; de Kreuk et al., 2007). However, aerobic granular sludge has not yet been applied to full-scale wastewater treatment plants (WWTP), mainly due to the inability to satisfactorily control, couple and maintain the granulation and EBPR processes.

Full-scale activated sludge systems, for the most part, perform reliable biological nutrient removal. There are, however, occasional instances of EBPR performance deterioration and failure, even under favourable operational conditions at both laboratory scale and full scale (Blackall et al., 2002; Crocetti et al., 2002; Thomas et al., 2003), and maintaining stable performance is a major topic of interest. EBPR failure is commonly attributed to the presence of glycogen-accumulating organisms (GAO), which can compete anaerobically with PAO for the uptake of carbon substrates, such as volatile fatty acids (VFA). Because GAO do not uptake and cycle polyphosphate, anaerobic carbon stores are converted to glycogen as their main energy storage molecule (Oehmen et al., 2005). Major GAO identified to date include Candidatus‘Competibacter phosphatis’ (henceforth referred to as Competibacter, also known as the GB lineage), a deeply branching cluster within the Gammaproteobacteria (Crocetti et al., 2002; Kong et al., 2002), and Defluvicoccus-related organisms within the Alphaproteobacteria (Wong et al., 2004; Meyer et al., 2006).

Bacteriophage infection of PAO cells is another possible cause of EBPR failure and, presently, little is known about this potentially important topic in relation to activated sludge performance and ecology. Bacteriophage numbers are shown to be between 4.2 × 107 and 3.0 × 109 mL−1 in a full-scale activated sludge WWTP, appearing as dynamic populations with the rapid emergence and disappearance of particular bacteriophage types (Otawa et al., 2007). Previous studies show that lytic-bacteriophage replication is characterized by the rapid proliferation of bacteriophage, followed by host lysis and release of further bacteriophage (Williamson et al., 2008). Lytic infections tend to decimate local populations quickly (Pantasticocaldas et al., 1992) and are considered a biological factor affecting the bacterial population dynamics in activated sludge (Lee et al., 2007). Laboratory-scale activated sludge reactors are reported to contain high numbers of bacteriophage (Ewert & Paynter, 1980; Otawa et al., 2007; Kunin et al., 2008); however, very little is known about the nature of these infections (i.e. lytic/lysogenic) and the effects of bacteriophage on both EBPR performance and granular sludge.

In the current study, bacteriophage were identified in a failing laboratory-scale EBPR reactor enriched with Accumulibacter, which led us to hypothesize that a bacteriophage infection of Accumulibacter may have been responsible for the poor EBPR performance. Cell-free supernatant, thought to be enriched with Accumulibacter-associated bacteriophage, was collected and stored. This supernatant was used to reinfect a number of small, Accumulibacter-enriched reactors operating either as floccular or as granular sludges. This corroborated the hypothesis that the previous reactor failure was due to bacteriophage predation on the Accumulibacter cells and not a result of PAO–GAO competition.

Materials and methods

Sequencing batch reactor (SBR) operation, analysis and sample preparation

A laboratory-scale SBR (named Granular Crash) was operated for EBPR and fed with synthetic wastewater containing VFA and orthophosphate (P-PO43−) for the enrichment of PAO. Granular Crash SBR had a working volume of 8 L and a 6-h cycle time, consisting of 10 min of decant, 6 min of feed, 120 min of the anaerobic phase, 218 min of the aerobic phase, 4 min of waste and 2 min for settling. Eight 5-mL liquid samples were collected over a single reactor cycle and filtered with a 0.2-μm filter to remove all bacteria for phosphorus and VFA analysis. The solids retention time (SRT) was approximately 18 days. A dosing of 0.5 M HCl and 2 M NaOH was used to control the pH during the anaerobic phase between 7.8 and 8. Dissolved oxygen was maintained at 0 mg L−1. during the anaerobic phase via nitrogen sparging and maintained between 3.0 and 4.5 mg L−1 during the aerobic phase via compressed air sparging. The synthetic feed composition for all reactors consisted of a COD : P ratio of 20 : 1 and is further described in Lu et al. (2006). VFA was alternated between acetate and propionate, with a switching frequency of every 2 weeks, in order to confer a selective advantage to PAO over GAO. For additional information on the reactor set-up, operation and feeding strategy, please see Supporting Information and Lu et al. (2006). Mixed liquor from the Granular Crash reactor was collected during week 38 of operation. The mixed-liquor sample was homogenized using a manual homogenizer in an icebox, centrifuged at 4000 g for 10 min at 4 °C to remove cellular biomass and filtered through 0.2-μm filters to remove bacterial cells and obtain a high-bacteriophage supernatant. This Granular Crash supernatant was subsequently stored at 4 °C before being used to infect the mini-SBR (mSBR).

mSBR set-up, operation and sample preparation

Four mSBR, with an 800 mL working volume, were operated for EBPR simultaneously and were tested by a challenge with a membrane-filtered Granular Crash supernatant, hypothesized to contain bacteriophage, but few or no bacterial cells. Inoculum sludges for mSBR were obtained from granular and floccular parent reactors, which had been operated previously for a period of 6 months to achieve a stable EBPR performance. Two mSBR were operated with granular sludge (Granular Control and Granular Infect) from a parent SBR (8 L) and two with floccular sludge (Floccular Control and Floccular Infect) from a parent SBR (4 L). As an inoculum for each mSBR, 600 mL of parent SBR mixed liquor was taken and concentrated to 150 mL. Control reactors received an additional 150 mL of 0.2 μm filtered, cell-free parent SBR supernatant for a total volume of 300 mL. Infect mSBR received 75 mL of 0.2 μm filtered, cell-free parent SBR supernatant and 75 mL of stored, 0.2 μm filtered, cell-free Granular Crash supernatant for a total volume of 300 mL. The mSBR were operated for 14 days on a 6-h cycle, with 10 min of decant, 6 min of feed, 124 min of the anaerobic phase, 180 min of the aerobic phase and 40 min of settling. The mSBR were operated under reduced feed conditions (1/6th volume) and no decant until day 3, when full feed conditions and decant were introduced. At the beginning of each cycle, 200 mL of synthetic feed was fed into the mSBR, resulting in a hydraulic retention time of 24 h. The synthetic feed composition for the mSBR was the same as described previously in Lu et al. (2006), with only acetate used as the sole source of VFA. For details of mSBR set-up, please see Fig. S1.

Transmission electron microscopy (TEM) sampling and viewing

Samples of mSBR biomass were prepared for TEM analysis. These were stabilized using 2.5% glutaraldehyde and 75 mM lysine in 0.1 M cacodylate buffer for 10 min to minimize structural damages arising from the dehydration procedure (Jacques & Graham, 1989). For detailed information relating to the TEM sample processing, please see the Supporting Information. Bacteriophage were visualized by negative staining and samples were fixed according to Patel et al. (2007) and filtered through 0.2-μm filters before processing. To allow sufficient numbers of bacteriophage to be visualized, 80 μL of sample was centrifuged onto carbon-coated copper grids using an EM-90 rotor in a Beckman Airfuge spun at 90 000 g for 20 min. Grids were washed briefly in water, dried and stained for 30 s in 1% uranyl acetate and imaged using a JEOL 1010 TEM operated at 80 kV.

Quantitative FISH analyses

Biomass samples were taken during the aerobic phase of mSBR cycles and fixed in 4% paraformaldehyde in phosphate-buffered saline at 4 °C for 2 h. FISH was performed as described previously (Amann, 1995). Slides were hybridized with EUB Mix (Amann et al., 1990; Daims et al., 1999), GAO Mix (Crocetti et al., 2002) and PAO Mix (Crocetti et al., 2000) probes (see Table S1 for detailed probe information). Fluorescent DNA probes were visualized and images were captured with a Zeiss LSM 510 Meta confocal laser-scanning microscope (CLSM; Carl Zeiss, Jena, Germany) using a Zeiss Neofluor × 40/1.3 oil objective. Images were analysed using daime version 1.2 (Daims et al., 2006). Image segmentation was carried out using default parameters. The relative abundance of different bacterial groups was determined by calculating the average proportion of the area targeted by EUB Mix probes (all bacteria) that was also targeted by group-specific probes (Accumulibacter or Competibacter). Between 20 and 30 images were acquired for each sample and the artefact rejection tool was set at a congruency threshold of 75%.

Bacteriophage-like particle (BLP) counts

Biomass samples were taken during the aerobic phase of mSBR cycles and filtered through 0.2-μm filters and fixed with 0.02 μm filtered formaldehyde to a 2% v/v final concentration (Noble & Fuhrman, 1998; Patel et al., 2007). Between 200 μL and 1 mL of the filtered bacteriophage supernatant was used for counts (depending on bacteriophage concentrations), filtered onto a 0.02-μm anodisc filter (Whatman), followed by SYBR staining (Invitrogen, Vic., Australia) and visualization of bacteriophage using CLSM as before. 0.02 μm of filter-autoclaved MilliQ water was added to prevent the meniscus from concentrating bacteriophage particles near the filter ring (Patel et al., 2007). Images were analysed using daime version 1.2 (Daims et al., 2006). Object counts were performed from 20 to 30 images acquired from a filter using the count objects function. BLP per millilitre of supernatant were calculated as described previously (Patel et al., 2007).

Particle size analysis

To determine the particle size distribution of granules and flocs, 30 mL of mixed liquor was sampled from the SBR at the end of the aeration period and applied to a Malvern laser light scattering instrument, Mastersizer 2000 series (Malvern Instruments, Worcestershire, UK).

Proteomic extraction and analyses

For protein extraction, sludge was collected and homogenized using a manual homogenizer in an icebox. One millilitre was transferred to a microcentrifuge tube and centrifuged at 15 000 g for 15 min at 4 °C. The supernatant was discarded and the sludge pellet was resuspended in 1 mL of urea–thiourea–CHAPS (UTCHAPS) buffer and proteins were extracted. Proteins extracted using UTCHAPS were analysed by LC–MS/MS. The resulting peptide mass fingerprints were searched against the three metagenomic databases (OZ sludge, Phrap assembly; US sludge, Phrap assembly; US sludge, Jazz assembly) (García Martín et al., 2006; Markowitz et al., 2006) and the LudwigNR database using the MASCOT search tool (http://www.matrixscience.com) via the Australian Proteomics Computational Facility (http://www.apcf.edu.au). For detailed information regarding the protein extraction procedure and LC–MS/MS analysis, please see Supporting Information.


Accumulibacter-associated bacteriophage detected during performance decrease in EBPR laboratory-scale SBR

The Granular Crash SBR was inoculated with floccular EBPR sludge from a domestic WWTP and operated for EBPR and for the formation of aerobic granules. The Granular Crash reactor was operated over a 26-week period, during which the microbial community became dominated by Accumulibacter (75% of bacteria as determined by quantitative FISH) and the SBR showed stable EBPR performance, demonstrated by a mean (± SE) effluent P-PO43− concentration of 2.31 (± 3.04) mg L−1 (n=12) and P-PO43− release during the anaerobic phase of 66.59 (± 15.18) mg L−1 (n=12). The reactor phosphate profiles were typical of a well-performing EBPR reactor, with phosphate release in the anaerobic phase and subsequent uptake in the aerobic phase (e.g. Week 26 in Fig. 1). Aerobic granules were obtained by week 26 of the reactor operation, demonstrated by a median particle size of 460 μm. However, from weeks 26 to 38, EBPR performance decreased (Fig. 1) although anaerobic consumption of VFA continued (data not shown). By week 38, the Accumulibacter population had decreased from 75% to 22% of the total bacterial population, as determined by quantitative FISH. The reactor was then shut down and a small amount of sludge samples was stored at 4 °C.

Figure 1.

 Phosphorus concentration (mg L−1) of the Granular Crash reactor treating synthetic wastewater enriched with Accumulibacter population over a 12-week period (week 26–week 38) after granulation. Before week 26, the reactor had been successfully performing EBPR.

Analysis of TEM samples from Granular Crash taken at week 26 (good EBPR performance) showed that the dominant cells were approximately 1 μm in diameter and contained large, dark intracellular stores of polyphosphate (Crocetti et al., 2000) and were therefore presumably Accumulibacter (Fig. 2a). However, TEM images of samples from week 36 (poor EBPR performance) showed sections of cellular membrane surrounding polyphosphate granules, indicative of apoptosis or lysed Accumulibacter cells (Fig. 2b). Further TEM analysis of the supernatant from Granular Crash at week 36 provided a visual representation of BLP (Fig. 2c).

Figure 2.

 TEM images of sludge from the Granular Crash reactor during weeks 26 and 36 of operation. Healthy Accumulibacter cells during week 26 of reactor operation with intact cell membrane, scale bar=1 μm (a). Possible lysed Accumulibacter cells during week 36 of reactor operation, scale bar=500 nm (b). Bacteriophage visualized by negative staining during week 36 of reactor operation, scale bar=200 nm (c).

A proteomic analysis was performed on the total proteins extracted from the Granular Crash sludge at week 30 (Table S2). The three most prominent proteins (according to peptide spectral counts) detected in the sludge sample were putatively identified as phage tail sheath proteins. This indicated further evidence of a possible bacteriophage attack (Finn et al., 2008). One of those was a phage protein FI associated with Accumulibacter, with a 99% sequence homology to a putative prophage gene from the Accumulibacter genome. The other two, initially appearing as unknown proteins against the US sludge Phrap assembly, were revealed to contain putative conserved domains within the phage tail sheath protein superfamily from the pfam database (Table S3). As a comparison, prominent proteins were analysed from extracts prepared from a well-performing EBPR sludge (Table S4). However, from that sludge, we did not detect bacteriophage-related proteins in high abundance. Likewise, other studies on well-performing EBPR sludges have not identified bacteriophage proteins in high abundance (Wilmes et al., 2008). These results suggest that an Accumulibacter-associated bacteriophage protein was unusually abundant within the Granular Crash reactor during the time of failing EBPR performance.

Accumulibacter-associated bacteriophage from the Granular Crash SBR were used to reinfect EBPR sludges

In order to test whether bacteriophage were responsible for the EBPR performance deterioration in the Granular Crash SBR, four mSBR were operated simultaneously for EBPR. Inoculum parent sludges had a mean effluent P-PO43− concentration of 4.38 (± 3.68) mg L−1 (n=3) and P-PO43− release during the anaerobic phase of 49.69 (± 2.90) mg L−1 (n=3) in the granular inoculum, and a mean effluent P-PO43− concentration of 0.40 (± 0.32) mg L−1 (n=3) and P-PO43− release in the anaerobic phase of 55.35 (± 5.11) mg L−1 (n=3) in the floccular inoculum. Accumulibacter enrichments in the granular and floccular inoculum sludges were 80% and 88%, respectively, as assessed by quantitative FISH. Granular inoculum sludge was fully granulated, with a median particle size of 1165 μm, while the floccular inoculum sludge had a median particle size of 140 μm. Infect mSBR were inoculated with a membrane-filtered supernatant from the Granular Crash reactor while Control mSBR received a membrane-filtered supernatant only from parent reactors. There was no evidence from proteomics analysis of bacteriocins or any other proteinaceous antimicrobials within the membrane-filtered inoculum. mSBR were then run for 14 days and monitored closely using quantitative FISH, BLP counts and EBPR process performance to assess whether bacteriophage attack on the infect reactors was responsible for the deterioration in EBPR performance.

By day 3 after inoculation with the bacteriophage supernatant, both Infect reactors showed a drastic decrease in Accumulibacter and an increase in Competibacter population abundance, as assessed by quantitative FISH (Fig. 3). Furthermore, at this time, both the Infect reactors showed an increase in BLP (Fig. 4), with the Granular Infect showing an order-of-magnitude increase. EBPR performance was monitored on day 3 after inoculation; however, a comparison of phosphorus-removal performance between the Infect and the Control reactors was extremely difficult due to mSBR variability and large differences in the phosphorus concentrations. The Infect reactors, particularly the Granular Infect, showed high P-PO43− concentrations in the liquid phase on day 3 of around 400 mg L−1 (extremely high for an EBPR system; Fig. S2), most probably resulting from lysis of Accumulibacter cells causing a release of phosphorus from intracellular polyphosphate stores. As a result, the influent synthetic wastewater feed had the effect of diluting infect reactor phosphorus concentrations, resulting in negative anaerobic phosphorus release profiles. In order to compare mSBR phosphorus-removal performance, the phosphorus concentration in the effluent was subtracted from the phosphorus concentration within the reactors after feed to yield an approximate measure of the total phosphorus removed from the synthetic wastewater per cycle (Fig. 5). On day 3 after inoculation, the Granular Infect showed negative phosphorus removal, strongly indicative of phosphate release as a result of cell lysis during that cycle, and had no anaerobic VFA consumption (data not shown), while the Floccular Infect also showed negative phosphorus removal, but still maintained anaerobic VFA consumption. In comparison, on day 3, both Control reactors removed net phosphorus and achieved complete anaerobic VFA consumption.

Figure 3.

 Representative FISH images indicating the bacterial populations present during the period of the mSBR operation. Cells are detected with the PAO-specific probe (Accumulibacter) appearing in blue, the GAO-specific probe (Competibacter) appearing in yellow and all other bacteria appearing in green, scale bar=100 μm. The column on the left shows images from the Granular Parent (top) and Floccular Parent (bottom) reactors. The remaining columns are images taken over the operating period. The top two rows are from the granular-control and infect mSBR. The bottom two rows are from the floccular-control and infect mSBR. The quantitative FISH results are included for PAO and GAO at the base of each image.

Figure 4.

 BLP counts per millilitre of sludge determined from mSBR experiments over the 14-day operating period. Error bars represent the SE of the mean from 20 to 30 fields of view from one filter.

Figure 5.

 Phosphorus-removal graph indicating the net phosphorus removal (mg L−1) from mSBR over the 14-day operating period. Phosphorus removal was calculated by subtracting the phosphorus concentration in the effluent from the phosphorus concentration within mSBR after feed.

Recovery of the granular and floccular sludges following initial bacteriophage infection

Following the reinfection of the Accumulibacter-enriched EBPR sludge and initial bacteriophage attack, the mSBR were further investigated to assess whether there was recovery of Accumulibacter populations and/or phosphorus-removal performance in either the granular or the floccular sludges. Granular Infect showed an almost complete disappearance of the Accumulibacter population from days 3 to 14 as assessed by quantitative FISH (Fig. 3). Furthermore, FISH images from days 7 and 10 in the Granular Infect showed an increase in filamentous organisms and an apparent increase in Competibacter. During this time, the BLP counts also decreased back down to the initial seed concentrations (Fig. 4) and there was no net phosphorus removal (Fig. 5) and no anaerobic VFA uptake. The Floccular Infect, however, showed an increase in the relative Accumulibacter population abundance from days 3 to 14 (Fig. 3) along with a slower decrease in the BLP counts (Fig. 4). By day 14, the Floccular Infect had recovered net phosphorus removal (Fig. 5) and the relative abundance of Accumulibacter had reverted to seed levels. Granular Control from days 10 to 14 showed a decrease in Accumulibacter relative abundance, an increase in the BLP counts and a decrease in phosphorus-removal performance, and by day 14, showed negative phosphorus removal. These results are similar to the Infect reactors following inoculation and indicate a possible delayed bacteriophage attack within Granular Control. This infection may have resulted from contaminating bacteriophage particles through sampling, aeration and close proximity to the Infect mSBR (Fig. S1) or from some other lytic-bacteriophage infection. Floccular Control from days 3 to 14 did not show a significant decrease in Accumulibacter abundance or an increase in the BLP counts and maintained net phosphorus-removal performance.


There is great interest to better understand the nature of host–bacteriophage interactions in real, natural ecosystems as they likely influence nutrient cycling, system respiration, species distributions and bacterial biodiversity (Fuhrman, 1999; Suttle, 2007). Previous studies of bacteriophage infections and host population dynamics are largely conducted on pure cultures of specific host–bacteriophage systems (Lenski & Levin, 1985), and recently, within artificially mixed-culture systems (Lee et al., 2006; Lennon & Martiny, 2008; Shapiro et al., 2010). To date, the occurrence and analysis of bacteriophage in activated sludge have been preliminary, focusing on quantifying bacteriophage and determining their community dynamics (Otawa et al., 2007; Nordgren et al., 2009), with little done to investigate host–bacteriophage interactions and their effect on process performance. In the current study, we used molecular techniques and microscopy to study the microbial ecology and process performance of a naturally arising bacteriophage infection within a mixed culture, activated sludge system. The resulting infection was found to remove a high proportion of the Accumulibacter population, resulting in significant changes to the community dynamics, process performance and granular sludge stability. Importantly, the study highlights the significant effect that bacteriophage may have on activated sludge process performance and raises doubts about the viability of granular sludge as a novel wastewater treatment technology. Furthermore, this study demonstrates the potential of EBPR laboratory reactors to study intricate ecological questions. It must be noted that studying bacteriophage within mixed-culture ecosystems is very challenging. The key organisms of EBPR, such as Accumulibacter and their antagonists Competibacter, have not been cultured in isolation. It is therefore extremely difficult to isolate their respective bacteriophage and to study these bacteriophage–host interactions, which are proving crucial for ecosystem understanding (Lindell et al., 2007; Suttle, 2007). The current study was only possible due to the fortuitous detection of a possible Accumulibacter-associated bacteriophage and the retrieval of suspected bacteriophage samples from the Granular Crash reactor. Consequently, only a limited amount of bacteriophage supernatant was available for the mSBR inoculation, limiting the scope for repeat experiments. The study was also somewhat constrained by the small size of the mSBR, which limited the volume of samples that could be removed and therefore the extent of analysis performed.

Evidence for Accumulibacter-associated bacteriophage lytic events and impacts on community structure

A previous study has identified temperate bacteriophage in the Accumulibacter metagenome (García Martín et al., 2006), indicating a possible lysogenic bacteriophage integration into the host genome. Further Accumulibacter-associated lytic-bacteriophage events are thought to have occurred in EBPR systems following the analysis of the CRISPR region in Accumulibacter DNA fragments (Barrangou et al., 2007; Kunin et al., 2008), a region that provides acquired bacteriophage resistance in prokaryotes (Sorek et al., 2008). In the current study, there were some minor operational differences between the parent SBR and the mSBR, such as a reduced feeding level for the first 3 days, to simulate slight starvation conditions. Environmental conditions that cause physiological stress, such as starvation, are known to induce lytic-bacteriophage transition events (Weinbauer, 2004; Williamson et al., 2007), and our induced starvation conditions may have contributed to subsequent lytic-bacteriophage events.

Importantly, in this study, there was evidence of Accumulibacter-associated lytic-bacteriophage events. A high abundance of Accumulibacter-associated phage proteins and possible lysed Accumulibacter cells were detected within the Granular Crash reactor. The proteomic analyses from the Granular Crash and the well-performing EBPR sludge suggest a high relative abundance of phage-associated proteins present at the time of the failing EBPR reactor. It should be noted that the proteomic approach used here can be applied for quantitative comparisons of protein abundance (Bantscheff et al., 2007); however, in this instant, this snapshot of detected proteins was not intended for absolute quantitative purposes. Further support of the Accumulibacter-associated bacteriophage events includes the increase in BLP counts coinciding with decreases in Accumulibacter population abundance in the two Infect mSBR.

During the mSBR experiment, the decrease in the Accumulibacter population was concomitant with an apparent increase in the Competibacter population in the Infect mSBR. Bacterial population abundances were measured using quantitative FISH, which measures the volume of a target population relative to a general population (in this case Accumulibacter or Competibacter against all bacteria, as defined by the EUB probes). Because of this relative assessment, a large and rapid decrease in Accumulibacter abundance, as seen on day 3, would result in a skewed representation of increased Competibacter abundance. Therefore, the increase in Competibacter abundance seen within Infect reactors was partially due to the large relative decrease in Accumulibacter abundance. This is further confirmed by partial VFA consumption, from days 3 to 14 in the Granular Infect, during the anaerobic periods (data not shown), indicating that Competibacter was not consuming all anaerobic VFA and therefore was not out-competing Accumulibacter. The resulting VFA was then available at the start of the aerobic period and was consumed rapidly by aerobic filamentous organisms. There was evidence of an increased abundance of filamentous bacteria, which can be regular components of activated sludge/EBPR systems (Eikelboom & Geurkink, 2002; Nielsen et al., 2009), within the Granular Infect mSBR, from days 7 to 10. The failure in the Granular Infect raises an interesting hypothesis as to the actual cause of EBPR system failures. It is conceivable that bacteriophage infections are overlooked as the initial cause of microbial community changes. We hypothesize that bacteriophage infections may be responsible for initially reducing the relative abundance of a particular group of organisms (e.g. PAO), thereby reducing competition for that niche space and resulting in increases in the relative abundances of other populations (e.g. GAO). Population changes such as these may result in a deterioration of EBPR performance.

Evidence for bacteriophage affecting EBPR performance and relevance to full-scale systems

In the current study, it is possible that a bacteriophage infection resulted in the EBPR failure. Before seeding the mSBR, both granular and floccular sludges demonstrated good phosphorus-removal performance. It may have been preferable to operate the mSBR for a stable period before infection with bacteriophage. However, due to the possible divergence of microbial populations in replicate activated sludge reactors run over long periods of time (i.e. weeks or months) (Boon et al., 2000; Slater et al., 2010), this approach was not adopted. It should be noted that a negative phosphorus-removal profile, indicative of a large anaerobic P-PO43− release with lower aerobic uptake (as seen in Granular and Floccular Infect on day 3), should not typically occur, even in nonperforming EBPR sludges. These large negative removal profiles suggest significant lysis of Accumulibacter cells resulting in the release of intracellular polyphosphate stores. It is interesting to note that negative phosphorus-removal profiles were only seen in the mSBR during times of suspected bacteriophage infection.

Here, we produce evidence that bacteriophage cause deterioration in EBPR performance in laboratory-scale systems. A key question is whether these effects could also be seen at full scale. The synthetic wastewater used in the current study was less complex and less variable in composition than real wastewater and reactor operational parameters were tightly controlled, resulting in a simpler community with a greater enrichment of Accumulibacter than would ever likely occur at full scale. Full-scale systems typically have Accumulibacter population abundances of between only 5% and 20% of the total community (Gu et al., 2008) compared with between 60% and 90% in the laboratory-scale SBR used in the current study. It is therefore unlikely that such a drastic decrease in population abundance and process performance would be seen at full scale. Laboratory-scale EBPR enrichments are more likely composed of near-clonal Accumulibacter strains, making them ideal targets for ‘kill-the-winner’ bacteriophage predation (Thingstad & Lignell, 1997; Kunin et al., 2008). In comparison, full-scale EBPR systems likely contain a greater diversity of Accumulibacter and other PAO organisms (Kong et al., 2005; He et al., 2008; Lopez-Vazquez et al., 2008; Flowers et al., 2009), which may maintain EBPR performance in the event of an Accumulibacter-specific bacteriophage infection. Understanding the role of bacteriophage infections in laboratory-scale systems is the first, necessary step towards understanding their potentially significant and complex role in full-scale WWTP.

Consequences for aerobic granular sludge technology

Both granular and floccular EBPR sludges showed signs of bacteriophage infection on day 3 after inoculation. However, there were noticeable differences between them. The Granular Infect demonstrated an intense, short BLP peak following inoculation, while the Floccular Infect demonstrated a less intense BLP peak, but with a longer duration. Furthermore, the Granular Infect Accumulibacter population remained depleted and EBPR performance poor, while Floccular Infect appeared to recover both the Accumulibacter population and the EBPR performance. The community compositions of both sludges were similarly enriched with Accumulibacter; however, the compact structure and tight clustering of bacterial colonies within granular structures (Barr et al., 2010b) may have resulted in a more intense, rapid infection compared with the less ridged structure of floccular sludge, resulting in a slower, prolonged infection.

This preliminary study raises the possibility that granular sludge may be less adept at recovering from a bacteriophage attack within laboratory-scale EBPR systems. A possible explanation relates to shorter solids retention times (SRT) applied to floccular sludges in comparison with granular sludges. Floccular SBR typically have an SRT between 4 and 8 days (McMahon et al., 2002; García Martín et al., 2006; Lu et al., 2006), resulting in a rapidly regenerating biomass. However, aerobic granular SBR have longer biomass generation times, requiring an SRT between 15 and 20 days (Lemaire et al., 2008; Yilmaz et al., 2008). It is conceivable that a faster-growing floccular biomass would recover more quickly from a bacteriophage infection than a slower-growing granular sludge, resulting in a return to the original bacterial community and a recovery in performance. The physical structure of the granules may also contribute to the inability of granular sludge recovery. Granules are larger and much more stratified in terms of microenvironmental conditions, such as redox potential, substrate concentrations and microbial community structure (Lemaire et al., 2008). Bacteriophage infection of a dominant bacterial population may result in the deterioration of granule structure, exposing previously inner-granule sections, changing the microconditions experienced by the cells. Largely, the effects of the loss of a dominant bacterial population on granular systems are unknown. Comparatively, floccular sludge has a more loose structure, and the microbial communities may not be so conditioned to stratified environments. Whether differences in the recovery of granular and floccular sludges to bacteriophage infection occur at full scale remain to be seen. As there are currently no full-scale granular sludge WWTP, and full-scale systems are more complex microbial systems, testing this hypothesis will prove challenging.


This work was funded by the Environmental Biotechnology Cooperative Research Centre (EBCRC), which was established and funded by the Australian Government together with industry and university partners. J.J.B. acknowledges EBCRC for funding of PhD Scholarship. F.R.S. acknowledges the Australian Research Council for funding of a postdoctoral research fellowship. T.F. acknowledges the Japan Society for the Promotion of Science (JSPS) for funding of a JSPS research fellowship for young scientist. P.L.B. acknowledges EBCRC for part funding of a senior research fellowship. The authors would like to acknowledge Rick Webb from the Centre for Microscopy and Microanalysis (CMM) of The University of Queensland for help with TEM procedure and imaging, Amanda Nouwens from the Australian Institute of Bioengineering and Nanotechnology (AIBN) of The University of Queensland for help with proteomic analysis and Florent Angly from the Advanced Water Management Centre (AWMC) of The University of Queensland for useful discussions. This project is supported by International Science Linkages established under the Australian Government's innovation statement, ‘Backing Australia's Ability,’ and was linked to the European Union INNOWATECH programme.