Nitrification in hybrid bioreactors treating simulated domestic wastewater

Authors


Correspondence

Micol Bellucci, School of Civil Engineering and Geosciences, Newcastle University, Newcastle upon Tyne, NE1 7RU, UK. E-mail: micol.bellucci@gmail.com

Abstract

Aim

To provide deeper insights into nitrification process within aerobic bioreactors containing supplemental physical support media (hybrid bioreactors).

Methods and Results

Three bench-scale hybrid bioreactors with different media size and one control bioreactor were operated to assess how biofilm integrity influences microbial community conditions and bioreactor performance. The systems were operated initially at a 5-day hydraulic retention time (HRT), and all reactors displayed efficient nitrification and chemical oxygen demand (COD) removal (>95%). However, when HRT was reduced to 2·5 days, COD removal rates remained high, but nitrification efficiencies declined in all reactors after 19 days. To explain reduced performance, nitrifying bacterial communities (ammonia-oxidizing bacteria, AOB; nitrite-oxidizing bacteria, NOB) were examined in the liquid phase and also on the beads using qPCR, FISH and DGGE. Overall, the presence of the beads in a reactor promoted bacterial abundances and diversity, but as bead size was increased, biofilms with active coupled AOB–NOB activity were less apparent, resulting in incomplete nitrification.

Conclusions

Hybrid bioreactors have potential to sustain effective nitrification at low HRTs, but support media size and configuration type must be optimized to ensure coupled AOB and NOB activity in nitrification.

Significance and Impact of the Study

This study shows that AOB and NOB coupling must be accomplished to minimize nitrification failure.

Introduction

Biological nitrification is used in the activated sludge waste treatment systems to reduce ammonia discharges to the environment. Nitrification consists of the sequential oxidation of ammonia to nitrite and then nitrite to nitrate by ammonia-oxidizing bacteria (AOB) and nitrite-oxidizing bacteria (NOB), respectively. Although this two-step process is quite well understood, nitrification treatment performance can vary because the associated bacteria grow slowly and tend to be sensitive to ancillary environmental factors such as temperature, sludge age, pH, dissolved oxygen and nitrogen load (Rittmann and McCarty 2001). To avoid nitrification failure, nitrifying systems are usually operated with long solid retention times (SRTs, often higher than 10 days), which requires large reactors and extensive aeration that increase capital and operating costs. Therefore, there is a general need for alternative treatment system designs that reduce reactor volumes, but maintain higher SRTs to sustain effective nitrification.

One option is hybrid bioreactors that are intended to maintain both suspended and attached growth in the same chamber. In such systems, buoyant inert carriers are added in the bioreactor to promote biofilm formation and longer SRTs. Several hybrid designs have been examined at laboratory (Hem et al. 1994; Oyanedel-Craver et al. 2009) and pilot scales (Pastorelli et al. 1997; Rouse et al. 2007; Di Trapani et al. 2008), and the effect of temperature; SRT; and carrier types on performance have been assessed (Odegaard et al. 1999; Christensson and Welander 2003; Dulkadiroglu et al. 2004; Rouse et al. 2007; Levstek and Plazl 2009). However, past work has primarily focused on carbon removal in hybrid systems, whereas less work has been carried out on biological nitrification, which is the focus here.

Nitrification performance is known to be dependent on the activity, composition and diversity of the AOB and NOB (Kowalchuk et al. 1997; Rowan et al. 2003; Wittebolle et al. 2005; Akarsubasi et al. 2009; Bellucci et al. 2011). Quantitative studies based on real-time PCR (qPCR) and fluorescence in situ hybridization (FISH) demonstrated that efficient nitrification can be achieved in WWTPs when the AOB concentration is higher than 10cells ml−1 (Coskuner et al. 2005; Graham et al. 2007; Knapp and Graham 2007; Pickering 2008; Wittebolle et al. 2008). However, process stability also seems to be positively correlated to microbial species diversity and the composition of the nitrifying community in activated sludge and biofilm reactor systems as shown by polymerase chain reaction (PCR), denaturing gradient gel electrophoresis (DGGE) and/or construction of AOB 16S rRNA clone libraries (Egli et al. 2003; Rowan et al. 2003). In this light, hybrid systems should perform better that suspended (only) systems because they can simultaneously sustain higher biomass and provide more diverse niches, on supporting material and liquid phases, which should promote greater diversity and system resilience.

The aim of this study was to characterize and contrast nitrification process performance and the diversity and abundance of nitrifying bacterial guilds as a function of media size in different hybrid bioreactor systems. Reactors were comprised of a conventional completely mixed chamber that contained different sizes of buoyant, spherical polystyrene supporting materials (beads). Four bioreactors were operated at 5·0- then 2·5-day HRT to assess the effect of bead size on AOB and NOB abundance and nitrification performance under different operating conditions. Suspended and biofilm solids, chemical oxygen demand (COD), nitrification rate and nitrifying bacterial communities, analysed using qPCR, DGGE and FISH, were monitored in the bioreactors.

Materials and methods

Bioreactor set-up

Four continuous-flow custom-made laboratory-scale bioreactors were set up as follows. Each unit was inoculated with 3 l of return activated sludge (RAS) from the Municipal Wastewater Plant in Spennymoor, UK. The bioreactors were glass cylinders (height 90 cm, diameter 12·5 cm, water height 32) with a conical-funnel bottom (height = 12 cm) in which an air stone (15 mm × 25 mm; Interpret, Dorking, UK) had been placed in the centre. The bioreactor contents were constantly mixed (~100 rev min−1) with stainless steel paddle stirrers (stirrer diameter 5 cm, shaft 0·8 cm) situated at the level of the bottom of the cylindrical part. Aeration was provided at a constant flow rate of 1 l per minute (l min−1).The bioreactors were provided with synthetic ‘domestic’ wastewater comprising 66 mg l−1 (NH4)2SO4, 320 mg l−1 bacterial peptone (Lab M Ltd., Heywood, UK), 190 mg l−1 meat extract (Lab-Lemco Powder; Oxoid, Basingstoke, UK), 30 mg l−1 yeast extract granulated (Merck chemicals, Nottingham, UK), 30 mg l−1 urea, 28 mg l−1 K2HPO4, 2 mg l−1 CaCl2·2H2O and 2 mg l−1 of MgSO4·7H2O. Typically, this medium was autoclaved prior to use and, after sterilization, 1 ml l−1 of filtered sterile trace element solution (0·75 g l−1 FeCl3·6H2O, 0·075 g l−1 H3BO3, 0·015 g l−1 CuSO4· 5H2O, 0·09 g l−1 KI, 0·06 g l−1 MnCl2·4H2O, 0·03 g l−1 NaMoO4·2H2O, 0·06 g l−1 ZnSO4·7H2O, 0·075 g l−1 CoCl2·6H2O, 0·5 g l−1 EDTA and 1 ml l−1 concentrated hydrochloric acid) and 0·7 mg l−1 NaHCO3 were aseptically added to solution prior to use. This synthetic substrate had been used successfully in previous nitrification bioreactor studies (Graham et al. 2007; Bellucci et al. 2011) and resulted in ammonia, total Kjeldahl nitrogen (TKN) and COD levels of 18·6 ± 4·2, 96 ± 6 and 522 ± 50·5 mg l−1 (mean ± standard deviation, N = 9), respectively. The medium was stored at 4°C prior to use and was pumped to the bioreactors at a rate consistent with targeted nominal operating HRTs, that is, 5-day HRT for the first month and 2·5-day HRT after the first month of operation. The organic loading rate (ORL) was 105·1 and 210·5 mg COD l−1 day in the first and second phase of the experiment, respectively. The nitrogen loading rate (NLR) was 19·35 mg TKN l−1 day during the first month and increased to 38·70 mg TKN l−1 day during the second half of the experiment.

Three ‘hybrid’ bioreactors were provided with beads with different diameters, designated 3-BR, 6-BR and 12-BR (diameter of the bead in millimetre bioreactor), whereas one bioreactor was operated as a ‘no-bead’ control (CU); details are provided in Table 1. The beads were spherical and made of crystal polystyrene (density of 1·07 g l−1; Precision Plastic Ball Company Ltd, Ilkley, UK). The same total bead volume (i.e. 0·3 l; approximately 10% of the total reactor working volume) was used for the three hybrid units; therefore, the total bead-surface area increased in steps by a factor of two from the largest to smallest size bead reactors.

Table 1. Properties of the polystyrene supporting material
 3-BR6-BR12-BR
Diameter (mm)3612
Surface area per bead (mm2)28·3113452
Volume per bead (mm3)14·1113905
Surface/Volume ratio210·5
Total number of beads21 2162653332
Total surface provided (m2)0·60·30·15
Specific surface area (mm2 mm−3_ liquid-phase volume)0·20·10·05

Sampling programme

Dissolved oxygen (DO), temperature and pH were monitored semicontinuously in the bioreactors using dedicated probes (Broadley Technologies Ltd., Bedford, UK), whereas N-compounds, COD and suspended solids were quantified every 3 days. To characterize microbial community abundances and composition, liquid phase and bead samples were collected in duplicate every 9 days according to the following schedule: 16 beads per sample day from 3-BR, 4 beads per sample day from 6-BR and 1 bead per sample day from 12-BR, which ensured that the total bead volume was constant among bioreactors over time. The low number of beads collected per sampling event provided adequate biomass for molecular biological analyses without consequentially reducing bead habitat in the reactors over time. Further, it allowed direct comparison of biofilm data because bead-surface areas were constant among samples and within reactors.

Chemical–physical analytical methods

Nitrification performance was quantified by measuring inline image-N, inline image-N and inline image-N concentrations in the liquid phase. Ammonia concentrations were determined using the Ammonium Cell Test kit (MERCK KGaA, Darmstadt, Germany), whereas inline image-N and inline image-N were determined using ion chromatography (Dionex DW- 100 Ion Chromatography; Dionex Corp., Sunnyvale, CA, USA). The ion chromatograph, which was fitted with an IonPac AS14A analytical column (Dionex Corp., Sunnyvale, CA, USA) and a 25-μl injection loop, was operated at a flow rate of 1 ml min−1 with 8·0 mmol l−1 Na2CO3/1·0 mmol l−1 NaHCO3 as the eluent. COD was determined using a COD Cell Test kit (MERCK KGaA) according to the manufacturer's instructions. Total suspended solids (TSS) and volatile suspended solids (VSS) were quantified in triplicate according to Standard Methods (Clesceri et al. 1998). A correlation between VSS concentrations of the liquid phase and total number of bacteria assessed by qPCR was used to estimate the VSS concentrations associated with bead samples.

DNA extraction

Samples for DNA extraction were collected from both the beads and liquid phase, although the bead samples needed to be preprocessed to obtain biomass prior to extraction. Typically, 2 ml of liquid phase was collected and immediately centrifuged at 12 000 g for 5 min; the resulting pellet was stored at −20°C prior to DNA extraction. At the end of the experiment, all pellets were thawed and resuspended and vortexed in 2 ml of filtered (0·2−μm acetate filters) and autoclaved water. From this slurry, 250 μl were transferred to a Lysing Matrix E tube (MP Biomedicals, Solon, OH, USA) for mechanical lysis of the cells using a Ribolyser (Hybaid Ltd., Ashford, UK). After cell lysis, DNA was extracted using the FastDNA SPIN For Soil Extraction kit (MP Biomedicals, Solon, OH, USA) in accordance with manufacturer's instructions.

To extract the DNA from the beads, the predefined number of beads was collected and directly placed into 20-ml sterile tubes that contained all reagents needed for cell lysis. Lysis was then performed by aggressively agitating the tubes for 3 min at 3000 min−1 shaking frequency in a vibrating grinding mill (Mikro-Dismembrator; B. Braun Biotech International, Melsungen, Germany). The resulting biofilm suspension was transferred to a sterile 2-ml centrifuge tube, and DNA extraction was performed using similar methods as for the liquid phase. Resulting DNA from the beads and also the liquid phase were eluted with 100 μl of sterilized molecular biology grade water and stored at −20°C until further analyses.

Quantitative PCR

Quantitative PCR (qPCR) was employed to measure the abundance of AOB and NOB in the liquid phase and on the beads, using specific primer sets (Sigma-Genosys, The Woodland, TX, USA) targeting the 16S rRNA gene (Table 2). Quantitative PCR was performed using a Bio-Rad iCycler equipped with an iCycler iQ fluorescence detector and associated software (Bio-Rad Version 2.3, Bio-Rad Laboratories Ltd., Hemel Hempstead, UK). PCRs used a 15-μl reaction mixture, which contained 2 μl of DNA template, 1 μl of primer mixture (7·5 pmol each) and 12 μl of PCR Precision Mastermix PCR reagent (Primer Design Ltd., Southampton, UK). For fluorescence detection, SYBR green I (10 000x, Sigma-Aldrich Ltd., Gillingham, UK) was first diluted 1/100 in sterile and autoclaved molecular water, and the resulting solution was added to the PCR reaction mixture as 1% of the total volume of the reaction mix (v/v). The thermal cycling was carried out as previously reported by Kowalchuk et al. (1997) and Graham et al. (2007).

Table 2. Primers used in the experiment for qPCR
Target organismPrimer/probeSequence (5′-3′)References
  1. AOB, ammonia-oxidizing bacteria.

Total bacteria1055fATGGCTGTCGTCAGCTHarms et al. (2003)
1392rACGGGCGGTGTGTAC
AOBCTO 189fA/BGGAGRAAAGCAGGGGATCGKowalchuk et al. (1997)
CTO 189fCGGAGGAAAGTAGGGGATCGHermansson and Lindgren (2001)
RT1rCGTCCTCTCAGACCARCTACTG
NitrobacterNitro-1198fACCCCTAGCAAATCTCAAAAAACCGGraham et al. (2007)
Nitro-1423rCTTCACCCCAGTCGCTGACC
NitrospiraNspra-675fGCGGTGAAATGCGTAGAKATCGGraham et al. (2007)
Nspra-746rTCAGCGTCAGRWAYGTTCCAGAG

To compare bacterial abundances between the hybrid bioreactors containing beads and the control unit, the 16S rRNA gene copy numbers were converted to equivalent cell numbers assuming that one and 4·2 rRNA operons exist for nitrifying and total bacterial cells, respectively (Klappenbach et al. 2001). Bioreactor cell numbers were estimated by extrapolating unit concentrations to the whole bioreactor volume for liquid-phase data and by calculating the number of cells per bead to the whole bioreactor based on the total number of beads in each system.

Fish

Visualization of AOB and NOB was performed using FISH on floc samples collected from the liquid phase at the end of the experiment. Specifically, floc biomass was collected and then stored in ethanol (sample/ethanol ratio 1 : 1 [v/v]) at −20°C prior to fixation, which then was performed using 4% paraformaldehyde fixative solution (PFA) as described elsewhere (Amann et al. 1990; Bellucci and Curtis 2011). Hybridization was performed according to Coskuner et al. (2005) with probes targeting the whole bacterial community, AOB and NOB (Table 3). However, due to differing stringencies among probes, consecutive hybridization/washing steps were required. The procedure was first performed using probes with the highest melting temperature and then ones with lower melting temperature. FISH images were captured randomly by confocal laser scanning microscope (CLSM; Leica Microsystems Ltd., Milton Keynes, UK).

Table 3. Probes used for FISH analyses
TitleTargetSequence (5′-3′)FormamideReference
Eub338iMost eubacteriaGCT GCC TCC CGT AGG AGTVariesAmann et al. (1990)
Eub338iiPlanctomycetalesGCA GCC ACC CGT AGG TGTVariesDaims et al. (1999)
Eub338iiiVerrucomicrobialesGCT GCC ACC CGT AGG TGTVariesDaims et al. (1999)
Nso1225β-Proteobacterial AOBCGC CAT TGT ATT ACG TGT GA35%Mobarry et al. (1996)
NeuHalophilic Nitrosomonas sp.CCC CTC TGC TGC ACT CTA35%Wagner et al. (1995)
CTECompetitor to NeuTTC CAT CCC CCT CTG CCG35%Wagner et al. (1995)
6a192Nitrosomonas oligotropha lineageCTT TCG ATC CCC TAC TTT CC35%Purkhold et al. (2003)
c6a192Competitor to 6a192CTT TCG ATC CCC GAC TTT CC35%Purkhold et al. (2003)
Nit3Nitrobacter sp.CCT GTG CTC CAT GCT CCG40%Wagner et al. (1996)
CNit3Competitor to Nit3CCT GTG CTC CAG GCT CCG40%Wagner et al. (1996)
Ntspa662Nitrospira sp.GGA ATT CCG CGC TCC TCT35%Daims et al. (2001)
CNtspa662Competitor to Ntspa662GGA ATT CCG CTC TCC TCT35%Daims et al. (2001)

DGGE

A nested PCR approach was employed for DGGE analyses to approximate general community diversity in the bioreactors. A ca. 465-bp fragment of 16S rRNA gene of AOB was amplified using CTO189F and CTO654R (Kowalchuk et al. 1997), while the primer sets 27f-Nbac-1050r and 27f-Nspira-705r (Freitag et al. 2005) were used to analyse Nitrobacter and Nitrospira, respectively. The PCR products were then used as a template for a second PCR with primers primer 3 and primer 2 (Muyzer et al. 1993). Primer 3 has an additional 40 nucleotide GC-rich sequence at its 5′ end. All the PCRs were carried out in 25-μl reactions containing 23·5 μl of PCR buffer (MegaMix-BLUE, Microzone Ltd., Haywards Heath, UK), 0·5 μl of each primer (10 pmol) and 0·5 μl of DNA template. The PCR conditions for CTO, 27f-Nbac-1050r, 27f-Nspira-705r and primer2–primer3 sets were previously described (Muyzer et al. 1993; Kowalchuk et al. 1997; Freitag et al. 2005). The nested PCR-amplified fragments were separated using a D-Code DGGE system (Bio-Rad Laboratories Ltd., UK).

Each sample was loaded onto 10% (w/v) polyacrylamide gel using a denaturing gradient ranging from 35% to 55% (100% of denaturant comprises 7 mol l−1 Urea and 40% v/v of deionized formamide). The gel was prepared with 10% (w/v) acrylamide stock solution (acrylamide-N,N0-methylenebisacrylamide, 37 : 1) in 1x TAE buffer (Tris-acetate-EDTA; 40 mmol l−1 Tris-acetate, 1 mmol l−1 EDTA, pH 8·0) with an appropriate amount of 100% denaturant solution. Electrophoresis was conducted in 1x TAE buffer at 60°C for about 4·5 h at a constant voltage (200 V). The gels were stained in a solution of SYBR green I (Sigma) diluted1/10 000 in 1x TAE buffer for 30 min and examined with an ultraviolet transilluminator (Quantity-one; Bio-Rad, UK). Comparisons of the DGGE band patterns were made with Bionumerics 4.0 (Applied Maths BUBA, Sint-Martens-Latem, Belgium) and the similarities between pairs of DGGE profiles were calculated using the Dice coefficient.

Assessment of AOB diversity and statistical analyses

Assuming that each band in the DGGE corresponds to a unique bacterial strain, AOB species richness (S) was estimated by counting the number of bands per DGGE profile. The diversity of the AOB community (Shannon index) and its evenness (Pielou's index) was calculated and compared between bioreactors, and between the bulk solution and the beads from the same bioreactor per sampling. The Shannon diversity index is defined by the following equation:

display math

where pi is the proportion of individuals found in the ith species (Shannon and Weaver 1949). As this index considers both the abundance and evenness of species present, the intensity of the DGGE bands relative to the summed intensity of all the bands in the profile was used as a measure of the proportion of individuals for calculation of the Shannon index. Pielou's index (Pielou 1969) was used as a measure of evenness and was calculated as follows:

display math

Both indices were calculated using PAST ver. 1.92 (http://folk.uio.no/ohammer/past/), and all statistical analyses were performed using Minitab 15 statistical software (Minitab Inc., State College, PA, USA).

Results

Bioreactor performance

The four bioreactors were operated for 1 month at an HRT of 5 days to determine baseline nitrifying operating and performance patterns for the different hybrid systems. All bioreactors performed similarly at a 5-day HRT with complete ammonia and nitrite oxidation being observed (see day 0 in Fig. 1) and COD removal efficiencies between 91 and 95% in all units. Furthermore, pH ranged between 7·1 and 7·5, DO between 4·9 and 5·9 mg l−1 and temperature averaged 20 ± 1°C with no observed differences among hybrid and control bioreactors (all P-values > 0·05; anova). In the liquid phase, the TSS and VSS concentrations were higher in the control unit (773 ± 13 and 557 ± 10 mg l−1, respectively) relative to the three hybrid bioreactors (3-BR, TSS = 335 ± 13 and VSS = 302 ± 13 mg l−1; 6-BR, TSS = 262 ± 18 and VSS = 222 ± 11 mg l−1; and 12-BR TSS = 280 ± 10 and VSS = 230 ± 5·0 mg l−1) (all P-value < 0·05; anova), which is consistent with significant biomass being associated with beads in the hybrid units.

Figure 1.

Concentrations of inorganic ammonia (♦), nitrite (■) and nitrate (▲) over time in the control (CU) and hybrid bioreactors (3-BR, 6-BR and 12-BR). At day 0, the reactors were operated at 5-day hydraulic retention time (HRT). The dash line represents the time when the HRT was decreased from 5 to 2·5 days (from day 1 to day 27).

When the reactors were switched to 2·5-day HRT, previous treatment performance continued for 13 and 19 days in the 12-BR and 6-BR units, respectively. However, performance then declined, as evidenced by increased inline image-N and inline image-N levels in the effluents. Ammonia removal was more consistent in the control and 3-BR units, although complete nitrification in 3-BR was only seen at the very end of the experiment.

In contrast, COD removal was always high in the units operated at an HRT of 2·5 days, suggesting that the presence of the beads did not influence carbon removal efficiency. However, Table 4 shows that total VSS in the hybrid bioreactors was consistently lower than in the control unit (i.e. bead biomass plus liquid-phase biomass) (P-value < 0·05; anova), which is contrary to the hypothesis that the inclusion of supporting material increases the amount of biomass in the systems. Furthermore, the biomass on the beads relative to the total biomass in the hybrid bioreactors ranged from only 22% to 38%, indicating that there was less overall biomass in the hybrid units when HRT was 2·5 days.

Table 4. Average of the bioreactors performance during 2·5-day hydraulic retention time (HRT) operation
 2·5-day HRT Operations
CU3-BR6-BR12-BR
  1. AOB, ammonia-oxidizing bacteria; CU, control unit; COD, chemical oxygen demand; DO, dissolved oxygen; VSS, volatile suspended solids; TSS, total suspended solids.

pH7·2 ± 0·47·2 ± 0·47·6 ± 037·5 ± 0·3
DO (mg l−1)5·4 ± 0·94·1 ± 0·55·6 ± 0·34·6 ± 0·5
Temperature (°C)21 ± 0·721 ± 0·721 ± 0·821 ± 0·7
COD (mg l−1)17 ± 1126 ± 1521 ± 1024 ± 6
inline image-N (mg l−1)73 ± 1160 ± 1059 ± 2252 ± 23
inline image-N (mg l−1)0·04 ± 0·18·8 ± 6·31·7 ± 1·37·6 ± 11·5
inline image-N (mg l−1)3·6 ± 7·60·6 ± 1·39·2 ± 1612 ± 17
TSS (mg l−1)
Bulk376 ± 74228 ± 47195 ± 39215 ± 32
Beadsn/andndnd
VSS (mg l−1)
Bulk335 ± 50204 ± 42176 ± 30190 ± 31
Beadsn/a46 ± 9·168 ± 1362 ± 12
Proportion on beads (%)
AOBn/a13 ± 2114 ± 9·053 ± 52
Nitrobactern/a0·005 ± 0·0080·018 ± 0·0190·16 ± 0·31
Nitrospiran/a6·5 ± 5·425 ± 2411 ± 12

Quantitative PCR and FISH analysis of nitrifying bacteria

Proportional AOB and NOB abundances in the liquid phase and on the beads were quantified over time using qPCR in the 2·5-day HRT bioreactors. With the exception of AOB in 12-BR, the large majority of AOB and NOB were not associated with the beads (Table 4). However, the hybrid bioreactors did tend to have higher absolute AOB levels than the control unit (all P-values < 0·05; anova), and AOB abundances generally increased with bead-surface area in the three hybrid reactors (Fig. 2).

Figure 2.

Correlation between the ammonia-oxidizing bacteria abundance detected by qPCR at the end of the experiment and the surface area available for the biofilm growing in the reactors.

To further investigate these incongruous observations, FISH was used to examine physical relationships between AOB and NOB as a function of floc integrity and size in the different units. CLSM images of flocs from different bioreactors operated at 2·5-day HRT were inspected (Fig. 3), and only the control unit had well-defined flocs with visible AOB microcolonies. Whereas, abundant AOB were not apparent in any hybrid flocs, which were always smaller and more diffuse relative to flocs in the control reactor. NOB were not observed in any floc samples. Therefore, although AOB and NOB were detected by qPCR in all reactors and AOB abundances correlated with bead-surface area, AOB and NOB signals were not closely associated in any of the hybrid units.

Figure 3.

Typical flocs observed in the control unit (a) and in the 16-BR (b) reactor, using confocal laser scanning microscope after in situ hybridization with fluorescent probes. Heterotrophic bacteria are labelled in green, whereas ammonia-oxidizing bacteria are in blue; nitrite-oxidizing bacteria in yellow were not detected.

Diversity of AOB and NOB

Nested PCR of the 16S rRNA gene followed by DGGE was used to assess how the presence of beads influenced AOB and NOB composition and species richness in the different systems. Clear community differences can be seen between the liquid phase and bead samples among units and over time (see Figs 4 and 5). Parallel DGGE on NOB communities showed no detectable differences for either Nitrobacter or Nitrospira subgroups in any of the bioreactors. Relative to AOB and assuming each band in the DGGE corresponds to a single strain, one can approximate species richness by counting the number of bands. The average number of bands detected in liquid phase and bead samples was 10·1 ± 2·0 and 9·75 ± 3·0 bands line−1 (mean ± standard deviation; P-value > 0·05, anova), respectively.

Figure 4.

AOB 16S rRNA gene community fingerprint detected both in the bulk solution and on the beads in hybrid and control reactors at day 1 and 10 by denaturing gradient gel electrophoresis.

Figure 5.

AOB 16S rRNA gene community fingerprint detected both in the bulk solution and on the beads in hybrid and control reactors at day 19 and 27 by denaturing gradient gel electrophoresis.

Overall, significantly greater AOB species richness (number of bands) was observed in the hybrid reactors relative to the control unit (P-value = 0·026, anova). Whereas, evenness of the AOB guild was not significantly affected by the presence of beads (P-value = 0·077, anova). Further, if one compares the relative AOB evenness in the suspended and biofilm microbial communities, no significant differences were seen in 3-BR and 12-BR (P-value > 0·05, anova), whereas AOB evenness on beads from 6-BR was significantly higher than AOB in the bulk solution (P-value < 0·05, anova). A comparison of Shannon indices among all bioreactors showed that the overall bacterial diversity did not differ significantly (P-value = 0·153, anova), nor were there significant differences between the suspended and biofilm communities within individual bioreactors (all P-values > 0·05, anova).

Discussion

The aim of this study was to develop a bioreactor system that retained efficient nitrification at low HRT. As such, the bioreactors were first operated at 5-day HRT to permit nitrifying guilds to develop in flocs and biofilms and also to determine baseline nitrification treatment efficiencies. Complete nitrification (and also high rates of COD removal) was seen for all reactors at 5-day HRT. However, upon reducing HRT to 2·5 days, nitrification efficiencies progressively declined both in the control unit and in two of the three hybrid reactors. It had been hypothesized that including beads in the reactor chambers should sustain efficient nitrification, even at lower HRTs by retaining more attached cells in the hybrid units, but this was not seen and the question is why.

Greater apparent biodiversity existed in the bioreactors with a higher surface area (3-BR and 6-BR; Fig. 2), but it was also shown that only 3-BR sustained efficient nitrification when HRT was reduced to 2·5 days (Fig. 1 and Table 4). Furthermore, despite better performance in 3-BR, attached biomass concentrations (based on VSS) were consistently low in all three hybrid reactors relative to the control and liquid phases. Finally, FISH data indicate no close associations between AOB and NOB floc signals in the liquid phase.

Overall, these data imply that incomplete nitrification in units operating at 2·5-day HRT, especially in the reactors with larger beads, resulted from poor floc development, possibly resulting from greater physical shear and/or floc disaggregation due to the presence of the beads (Fig. 3a). Visual analyses showed that bulk flocs were smaller when larger beads were present, and FISH data imply that the AOB and NOB were not consistently present together in the bulk floc in any of the hybrid units. These data suggest that the polystyrene beads, especially larger beads, must be damaging the physical integrity of bacterial clusters needed for complete nitrification. In fact, if one calculates the approximate shear forces on the bead surfaces (Vijaya Lakshmi et al. 2000), one finds that frictional effects increase by a factor of 1·1 with increasing diameter, suggesting that surface shear is four times greater on 12-BR than 3-BR, which could explain poorer biofilm formation with increasing bead diameter.

Interestingly, qPCR and DGGE data show that AOB and NOB were actually present in all reactors, but evidence further suggests that these guilds were present in the hybrid units as single cells or unassociated colonies in smaller flocs. Consistent with Larsen et al. (2008), the data imply that the shear forces on the bead surface eroded nitrifiers from the colonies, without affecting cell integrity, and this floc fragmentation (<40 μm) uncoupled the AOB and NOB, resulting in poor nitrification performance (Wittebolle et al. 2008). Therefore, although hybrid bioreactors have the potential for efficient nitrification at lower HRTs, careful consideration of media size and mixing conditions is needed to ensure appropriate floc development. It is interesting to note that COD removal was consistent among all hybrid bioreactors, which implies that nitrification was more affected by floc issues than carbon removal, which suggests spatial orientation is less important to carbon-consuming bacteria relative to nitrifiers.

One other possible explanation for destabilization of nitrification in the hybrid units is the reputed sensitivity of NOB to sudden changes in system conditions, that is, the loss of specific NOB subguilds can trigger the collapse of ammonia removal due to a fragile mutualism of AOB and NOB (Graham et al. 2007). It is not clear whether this happened here, although it is clearly possible that elevated physical shear and poor floc formation might have destabilized the NOB. Cluster analyses on DGGE band patterns showed that the AOB guild composition and, to a lesser extent, the NOB guild did fluctuate over time, but it was not possible to relate observed shifts to reactor performance.

Despite the negative impact of large beads on complete nitrification, the general presence of beads did positively affect species richness, especially AOB, with the beads representing a potential refuge for some species. However, increased richness did not enhance performance, except possibly in the bioreactors with the smallest bead (BR-3). It is suspected that factors other than species richness are more important to complete nitrification. Specifically, the results highlight the importance of spatial orientation, floc aggregation and AOB and NOB mutualism rather than abundance and diversity.

The work shows the possible importance of floc/biofilm development and structure on nitrification. Previous evidence indicates that nitrifier abundance and the coupling of AOB and NOB guilds within colonies are essential for complete ammonia and nitrite oxidation (Knapp and Graham 2007), which our results confirm. Therefore, reactor designers must consider floc conditions in their bioreactor designs, that is, conditions that facilitate larger floc formation, and/or suitable attached growth should be promoted. Although this was not considered here, we suspect effective nitrification is possible with hybrid reactors. Specifically, we recommend that hybrid bioreactors should have suspended support media with specific surface areas >0·2 mmmm−3 to promote biofilm formation, reduce biofilm shear and erosion, and allow adequate oxygen transfer through the biofilms. However, more work is needed to refine this recommendation, especially at the pilot plant and full scale, because correctly designed hybrid systems will ultimately reduce nitrification treatment costs by reducing reactor sizes and potentially aeration needs.

Acknowledgements

We thank María Roel Fernández who assisted in the laboratory work, Charles Knapp who provided advice on molecular methods and ECOSERV for financial support under the EU Marie Curie Excellence Programme (MEXT-CT-2006-023469). The authors declare to have no conflict of interest.

Ancillary