w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Biodiversity and population dynamics of microorganisms in

a full-scale membrane bioreactor for municipal wastewater
treatment
Cai-Yun Wan a,1, Heleen De Wever b,1, Ludo Diels b, Chris Thoeye c, Jun-Bin Liang a,
Li-Nan Huang a,b,*
a

Key Laboratory of Gene Engineering of the Ministry of Education, School of Life Sciences, Sun Yat-Sen University, Guangzhou 510275,
PR China
b
Unit of Separation and Conversion Technology, Flemish Institute for Technological Research (VITO), B-2400 Mol, Belgium
c
Research and Product Development Department, Aquafin NV, B-2630 Aartselaar, Belgium

article info

abstract

Article history:

The total, ammonia-oxidizing, and denitrifying Bacteria in a full-scale membrane biore-

Received 5 June 2010

actor (MBR) were evaluated monthly for over one year. Microbial communities were

Received in revised form

analyzed by denaturing gradient gel electrophoresis (DGGE) and clone library analysis of

30 September 2010

the 16S rRNA and ammonia monooxygenase (amoA) and nitrous oxide reductase (nosZ )

Accepted 4 November 2010

genes. The community fingerprints obtained were compared to those from a conventional

Available online 12 November 2010

activated sludge (CAS) process running in parallel treating the same domestic wastewater.
Distinct DGGE profiles for all three molecular markers were observed between the two

Keywords:

treatment systems, indicating the selection of specific bacterial populations by the con-

Membrane bioreactor

trasting environmental and operational conditions. Comparative 16S rRNA sequencing

Bacteria

indicated a diverse bacterial community in the MBR, with phylotypes from the a- and

Ammonia-oxidizing and denitrify-

b-Proteobacteria and Bacteroidetes dominating the gene library. The vast majority of

ing bacteria

sequences retrieved were not closely related to classified organisms or displayed relatively

Molecular diversity

low levels of similarity with any known 16S rRNA gene sequences and thus represent

Population dynamics

organisms that constitute new taxa. Similarly, the majority of the recovered nosZ
sequences were novel and only moderately related to known denitrifiers from the a- and bProteobacteria. In contrast, analysis of the amoA gene showed a remarkably simple
ammonia-oxidizing community with the detected members almost exclusively affiliated
with the Nitrosomonas oligotropha lineage. Major shifts in total bacteria and denitrifying
community were detected and these were associated with change in the external carbon
added for denitrification enhancement. In spite of this, the MBR was able to maintain
a stable process performance during that period. These results significantly expand our
knowledge of the biodiversity and population dynamics of microorganisms in MBRs for
wastewater treatment.
2010 Elsevier Ltd. All rights reserved.

* Corresponding author. Key Laboratory of Gene Engineering of the Ministry of Education, School of Life Sciences, Sun Yat-Sen
University, Guangzhou 510275, PR China. Tel./fax: 86 20 8411 2399.
E-mail address: eseshln@mail.sysu.edu.cn (L.-N. Huang).
1
These authors contributed equally to this work.
0043-1354/$ e see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2010.11.008

1130

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

Introduction

Submerged membrane bioreactors (MBR) combine efficient

biological treatment with membrane separation and are now
widely accepted as an advanced technology for obtaining
high-quality effluent. This process offers several important
advantages over conventional wastewater treatment systems,
including high biodegradation capacity and efficiency, excellent permeate quality and low sludge production. As more
stringent effluent standards are expected and the costs of
membrane and membrane process continue to fall, the
applications of MBR in municipal and industrial wastewater
treatment are becoming increasingly widespread around the
globe (Judd, 2007).
The bacterial communities present in active biomass in
activated sludge mixed liquor represent the core component
of every MBR for biological carbon and nutrient removal.
While bacteria in wastewater treatment plants (WWTP) have
been intensively studied by culture-dependent methods (Ueda
and Earle, 1972) and nucleic acid-based molecular approaches
(Bond et al., 1995; Nogueira et al., 2002; Wagner and Loy, 2002),
surprisingly little research has been conducted to explore the
influence of membrane separation and operating conditions
on the overall microbial community structure and diversity of
the MBR. Consequently, much of what is known or assumed
concerning biological processes in MBRs has primarily come
from investigations of conventional activated sludge (CAS)
systems, regardless of the fact that significant differences in
operating conditions exist between the two treatment
processes. Importantly, MBRs are typically operated at high
sludge concentrations and low food-to-microorganism (F/M)
ratios. Consequently, since energy supply is limited, the
microorganisms would preferentially use the carbon sources
to satisfy their maintenance energy demands as opposed to
biomass growth (Muller et al., 1995; Low and Chase, 1999). In
addition, other contrasting operational aspects of MBRs,
including longer sludge retention time (SRT), shorter
hydraulic retention time (HRT) and shear forces, would also
have an impact on the activated sludge communities. To date,
denaturing gradient gel electrophoresis (DGGE)-based
community structure analysis has indicated that the bacterial
communities in pilot-scale MBRs fed with raw sewage were
distinct from that in the CAS process (Luxmy et al., 2000),
while another investigation (through fluorescent in situ
hybridization, FISH) has revealed minor differences in the
nitrifying community composition between parallel-running
MBR and CAS pilot systems (Manser et al., 2005). In addition,
pilot studies monitoring long-term community structure
changes have demonstrated diverse and dynamic bacterial
populations in MBRs for graywater (Stamper et al., 2003) and
municipal wastewater (Miura et al., 2007) treatment even
during periods of stable operation. More recently, Huang et al.
(2008) have revealed by 16S rRNA clone library analysis that
novel members of the Bacteria domain are ecologically
significant in laboratory-scale municipal wastewater treatment MBRs operated under different conditions. These pioneering works highlight the need for exploring the microbial
community composition and diversity in these relatively new
biological wastewater treatment systems. To our knowledge,

there have been no molecular microbial diversity surveys of

full-scale MBR systems for municipal wastewater treatment,
and seasonal variations of these diverse communities in
relation to process performance remain little known.
We hypothesized that the contrasting operational and
environmental conditions of full-scale MBRs as opposed to
CAS systems will have great impact on the physiological state
and bacterial community structure and population dynamics
of mixed liquor. Our study site (in a European country)
represents a unique and ideal WWTP at which to examine and
compare the microbial community composition, since there is
an MBR and a CAS system running in parallel treating the
same municipal wastewater and the MBR was originally
inoculated with activated sludge from the CAS process. We
used DGGE fingerprinting technique to examine how the
structure of bacterial populations varied seasonally in both
environments. Clone library analysis of phylogenetic and
functional markers provided the first detailed molecular look
at the composition and diversity of total community and
ammonia- oxidizing and denitrifying Bacteria in a full-scale
municipal wastewater treatment MBR. Additionally, we
examined the impact of changes in MBR operating conditions
on the key microbial groups when external carbon sources for
the stimulation of denitrification were switched from one to
another during the study period. Significant shifts in total
bacteria and denitrifying community were observed with
some of the major shifting bands in the 16S rRNA gene DGGE
profiles corresponding to microorganisms capable of
denitrification.

2.

Materials and methods

2.1.

Full-scale MBR

The submerged MBR (6520 m3/d in capacity) was constructed

in 2003 as an extension of the existing CAS system to comply
with a more stringent effluent regulation and an increase in
load. It consists of a denitrification compartment where pretreated municipal wastewater is introduced, a nitrification
compartment, and a filtration compartment where activated
sludge is retained by submerged hollow-fiber microfiltration
membrane modules (Zenon, ZeeWeed, total membrane
surface area 10,160 m2 and nominal pore diameter 0.03 mm).
The compartments are well-mixed either by continuous stirring (denitrification) or aeration (nitrification and filtration).
The MBR is operated at a constant flow of 230 m3/h, with the
remaining flow (ranging between 0 and 1440 m3/h, with an
average value of 418 m3/h) being treated in the parallel CAS
system. Detailed set-up and operational information of the
full-scale MBR was given elsewhere (Fenu et al., 2010).
Initially, acetate was added to the denitrification compartment to enhance denitrifying activity. This carbon source was
then changed to butyrate on December 5 2006. Subsequent
switches of carbon source between butyrate and acetate
occurred on April 4 and in the end of May 2007 (Fig. 1).
In contrast to the MBR, the CAS process has no denitrification zones. Pretreated wastewater is introduced directly to
the aeration compartments. Separation of treated water from

1131

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

A S O

N D J

F M A

M J J

Time (m)
J

A S O N

F M

A M J

MAR 2007
FEB 2007
APR 2007
JUN 2007
MAY 2007
SEP 2006
AUG 2006
NOV 2006
DEC 2006
OCT 2006
JUL 2006
JAN 2007
JUL 2007
MAY 2007
APR 2007
JAN 2007
FEB 2007
MAR 2007
DEC 2006
JUL 2007
JUN 2007
NOV 2006
OCT 2006
SEP 2006
AUG 2006
JUL 2006

Marker

Marker

Marker

Time (m)

MBR

CAS

Similarity (%)

CAS

MBR

75

80

85

90

95

100

Fig. 1 e Comparison and Pearson correlation analysis of 16S rRNA gene DGGE fingerprint profiles from the MBR and CAS over
the course of 13 months (July 2006 through July 2007). Clustering is based on the unweighted pair group method using
arithmetic average (UPGMA) algorithm. Arrows indicate the time points when switches of carbon source occurred: from acetate
to butyrate on December 5 2006 (one week before the December 2006 sludge sampling), back to acetate on April 4 2007 (three
weeks before the April 2007 sampling), and to butyrate again in the end of May (one week after the May 2007 sampling).

the biomass is accomplished by secondary sedimentation.

Consequently, sludge concentration (mixed liquor suspended
solid, MLSS) in the bioreactors is maintained at a much lower
level than that in the MBR (2.49 g/l versus 9.08 g/l, see Table S1
in the Supplementary Materials). Average sludge load (or F/M)
during the study period was 0.18 and 0.025 kg BOD/kg MLSS.d
for the CAS and MBR systems, respectively.

2.2.

Analytical procedures

All samples were kept in cooler boxes during sampling and

transportation to the laboratory (within 1 h). 24-h composite
samples of feed wastewater and effluents (membrane
permeate and CAS effluent) were collected monthly for over
one year (July 2006-July 2007), taking into account the HRT of
both treatment systems. Samples were analyzed for a full-set
of conventional performance parameters (Supplementary
Materials text). Mixed liquor was sampled (in triplicate)
directly from the aeration compartments of the two treatment
systems. To indicate biological activity, specific oxygen uptake
rate (SOUR) and specific nitrification rate (SNR) were determined as described previously (Huang et al., 2008). Additional
sludge samples were taken from the denitrification compartment of the MBR for the measurement of specific denitrification rate (SDNR) (Supplementary Materials text) and
molecular microbial analysis of the denitrifying community
(as described below).

2.3.

DNA Extraction and DGGE analysis

Aliquots (1e3 mL) of mixed liquor samples were centrifuged

for 10 min at 12 000 g, 4  C. The cell pellets were rinsed twice
with sodium phosphate buffer (120 mM, pH 8.0) and total
community genomic DNA was extracted using an UltraClean
Soil DNA kit (MoBio, Solana Beach, Calif.). The 16S rRNA gene,

and ammonia monooxygenase and nitrous oxide reductase

genes (amoA and nosZ, responsible for the key steps in the
ammonia oxidation and denitrification pathways, respectively) were chosen as molecular markers for community
fingerprint analysis of total bacteria and ammonia-oxidizing
and denitrifying communities, respectively. For total bacterial
community, a 496-bp 16S rRNA gene fragment was amplified
with primer set GC-63F/518R as described previously (Huang
et al., 2008), and DGGE of the PCR products was performed
using an INGENYphorU-2 apparatus (INGENY International
BV, The Netherlands). Fragments of amoA (about 490 bp) were
amplified with primers amoA-1F-Clamp and amoA-2R-TC
(Supplementary Materials text) and resolved by DGGE
following the procedures of Nicolaisen and Ramsing (2002)
with some modifications (Supplementary Materials text).
The nosZ fragments (414 bp in length) were obtained using
primers nosZeF and nosZ1622R-GC (Throback et al., 2004) and
DGGE was performed as detailed in the Supplementary
Materials text. DGGE profiles were analyzed using Bionumerics 4.0 software (Applied Maths). Similarity matrices
between tracks were calculated from the intensity data with
band-independent, whole-densitometric curve-based Pearson
correlation coefficients and then subjected to UPGMA
(unweighted pair group method using arithmetic average)
clustering.

2.4.

Clone library, rarefaction, and statistical analysis

Clone libraries of 16S rRNA, amoA and nosZ genes were constructed for DNA extracts from MBR sludges obtained in
November 2006. Nearly complete 16S rRNA gene fragments
were amplified in triplicate with primers 27F and 1492R (Dojka
et al., 1998) as described previously (Huang et al., 2004). The
amplicons were pooled, purified with a QIAquick PCR purification kit (Qiagen), and cloned into pMD18-T vector by TA

1132

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

Table 1 e Bioreactor performance of the full-scale MBR and CAS process.a

COD
Influent
MBR effluent
CAS effluent

220  90
24  6
31  7

TOC
54  20
ND
ND

DOC
18  5
8.6  2.1
9.2  1.7

NH
4 eN

TeN
26  8
7.7  1.9 b
14  6

19  8
0.2  0.2
2.9  2.3

NO
3 eN
0.9  0.6
4.8  2.4
8.5  7.0

NO
2 eN
0.1  0.1
0.1  0.2
0.3  0.2

b
b

TeP
3.9  1.6
0.3  0.2
0.5  0.2

a Concentrations (mean  standard deviation) in mg/l. Data are based on the monthly samples collected during the whole study period (n 13).
b Some values are below detection limit and thus not included in the statistical analysis. ND, not determined.

Cloning Kit (Takara). Randomly selected clones from the

library were determined by colony PCR (with primer combinations 27F/M13-47 and 27F/RV-M) for their rRNA gene inserts
orientation and then sequenced forwardly using primers M1347 or RV-M on Applied Biosystems 3730xl capillary
sequencers. Chimeric sequences were identified as described
(Huang et al., 2004) and excluded from subsequent analysis.
For representatives of the operational taxonomic units (OTU,
defined as groups in which sequences differed by 3%) that
comprised two or more clones, nearly complete 16S rRNA
gene sequences were obtained by a second sequencing run
starting from the opposite side of the vector with the corresponding vector-primers. Rarefaction, richness (nonparametric richness estimators ACE and Chao 1) and diversity
statistics (the Shannon diversity index) were calculated using
DOTUR software (Schloss and Handelsman, 2005) and phylogenetic analyses were performed using the ARB software
package (Ludwig et al., 2004) as described previously (Huang
et al., 2008). Neighbor-joining (NJ) trees that differed in the
reference sequences, the set of alignment positions and the
outgroup sequences used, were generated and compared.
For functional gene cloning, non-GC-clamp primers were
used to obtain the amoA and nosZ fragments. PCR conditions were
the same as those used for DGGE analysis. Randomly selected
clones from each library were screened by restriction fragment
length polymorphism (RFLP) analysis (Supplementary Materials
text). Then one to three representatives of each unique RFLP
type were fully sequenced using vector specific primers. OTUs
were defined at the 95% sequence similarity threshold using
DOTUR (Schloss and Handelsman, 2005). Statistical and rarefaction analysis of the amoA and nosZ sequences were conducted as
for the 16S rRNA sequences. Phylogenetic trees were calculated
using ARB software (Ludwig et al., 2004).
To compare microbial community composition before and
after the carbon source change, additional clone libraries of
16S rRNA and nosZ genes were established for MBR sludge
sampled in February 2007.
The 16S rRNA, amoA and nosZ gene sequences obtained in
this study have been deposited in the EMBL/GenBank/DDBJ
database under accession numbers FN827166-FN827320 and
FN868486-FN868537, respectively.

3.

Results

3.1.

Biological activities and performance

Both treatment systems were able to maintain a good and

stable carbon (COD) removal over the course of 13 months
(Table 1, Fig. S1). Complete nitrification was consistently

achieved in the MBR as it was operated at a low food

(ammonium) to microorganisms ratio. In contrast, although
ammonium concentrations in the CAS effluent were mostly
below 3.0 mg/l, ammonium removal was relatively low
(29e77%) from December 2006 through March 2007 (Fig. S1)
likely due to the relatively low temperature and SNR during
that period. Statistical analysis revealed a correlation between
ammonium removal and bioreactor temperature (Pearsons
correlation coefficient r 0.702, P < 0.001) and SNR (r 0.560,
P < 0.05) for the CAS system. Although average SOUR value
was relatively higher for the CAS process (Table S1), total
community activity (as indicated by the SOUR values) was not
significantly different between the two treatment systems
(P 0.269, paired t tests). In contrast, the ammonia-oxidizing
Bacteria (AOB) community was more active in the CAS reactor
as evidenced by the higher SNR values (Table S1 and P 0.004,
paired t tests). In spite of these, the MBR process constantly
provided a better treatment in terms of treated water quality
and carbon and nutrient removal from sewage (Table 1,
Fig. S1) probably due to the higher biomass concentration and
membrane rejection of suspended solids and organic
compounds. An additional denitrification zone and the
constant flow would also have contributed to the increased
process efficiency. The substantial fluctuations in total and
denitrifying bacteria associated with changes in carbon
source (as described below) did not significantly affect the
overall bioreactor performance of the MBR.

3.2.
Population dynamics of total bacteria and N-cycling
communities: MBR versus CAS
PCR-based DGGE fingerprints were used to compare microbial
community composition between the MBR and CAS, and to
follow temporal fluctuations in bacterial populations in each
treatment system. Profoundly different pattern types by DGGE
analysis of the 16S rRNA gene were distinguished between the
two bacterial communities, and this was confirmed by the
statistical analysis which formed two separate major clusters
according to the environment (Fig. 1). While the MBR 16S rRNA
fingerprints were dominated by a limited number of bands,
the CAS fingerprints displayed complex banding patterns and
thus indicated a more diverse bacterial community. For DGGE
analysis of N-cycling groups, the amoA gene profiles revealed
a simple AOB community in the MBR and CAS, while fingerprints of the nosZ gene exhibited complex banding patterns for
the denitrifying communities in both systems (Fig. S2 and
Fig. 2). Similar to those of the 16S rRNA gene, DGGE profiles of
both functional markers formed system-specific clusters in
the respective UPGMA dendrograms. These results indicated
that the membrane separation process and different

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

operating conditions had led to the selection of different total

bacterial and N-cycling communities in the MBR.
Temporally, the MBR bacterial community showed
considerable stability before December 12 2006 (Fig. 1).
However, a clear shift in community composition was evident
since the samples of January 2007 which were taken one and
a half months after the carbon source was changed from
acetate to butyrate. Thereafter, the MBR populations gradually
evolved to another fluctuating community with the DGGE
fingerprints forming a separate cluster in the UPGMA dendrogram. Fluctuations were mainly due to shifts of the dominant
bands and variations in relative abundance of common bands
among samples from that period. Matching of clones from the
16S rRNA gene library back to the community band patterns (by
running GC-63F/518R-amplicons of randomly selected clones
together with amplicons of the original total community
genomic DNA on the same DGGE gels) revealed that some of the
important shifting bands corresponded to microorganisms
capable of denitrification (Table S2). Highly similar results were
obtained by DGGE analysis of the denitrification coding gene
(Fig. 2). The nosZ fingerprints of samples collected one and
a half months after the first carbon source change formed
a separate and fluctuating clade in the clustering analysis. In
contrast, little variations were found in AOB populations
between the sludges taken before and after the carbon source
change (Fig. S2). Generally, the clustering of amoA fingerprints
more followed the time course, reflecting a gradual evolution of
this functional group over time in the MBR.
DGGE analysis revealed fluctuations over time in community composition of total bacteria and N-cycling groups in the
CAS process (Figs. 1 and 2 and Fig. S2). Fingerprints of 16S rRNA
gene from December 2006 to March 2007 were highly similar,
and these communities formed a tight cluster separating from
the prior and subsequent samples in the UPGMA dendrogram.
This trend was even more profound in DGGE analysis of the
nosZ gene. The amoA fingerprint analysis also showed that AOB
communities from the cold months differed considerably from

1133

those of other seasons. These findings suggested that factors

related to seasonal forces had an influence on the CAS bacterial
populations. In addition, samples collected in July 2006 clustered apart from those of July 2007 in all three UPGMA
dendrograms, suggesting no seasonal reproducibility in
temporal evolution of microbial communities.

3.3.

Microbial diversity in the full-scale MBR

To investigate the biodiversity of total bacterial community

and N-cycling groups in the MBR, clone libraries of the 16S
rRNA and amoA and nosZ genes were constructed for sludge
samples collected in November 2006. Comparative sequence
analysis of 146 randomly selected clones from the 16S rRNA
gene library revealed 99 unique OTUs distributed among at
least 14 phylogenetic divisions (Table 2 and Fig. 3). The
minimum numbers of bacterial species in the MBR were predicted at approximately 300 according to both nonparametric
estimators (Chao 1 and ACE) (Table 2). The vast majority of the
OTUs comprised a single clone, indicating a diverse bacterial
community. This was further supported by the steep rarefaction curve (Fig. S3). Clone distribution within the library
was the following: b-Proteobacteria (27%), Bacteroidetes (25%),
and a-Proteobacteria (14%). The remaining phylogenetic
groups, such as the d-Proteobacteria, Actinobacteria, Acidobacteria, Chorobi, and Firmicutes, made up a total of 34% of
the bacterial library. Only seven OTUs were identified as
already known species, and five OTUs affiliated with not yet
cultured microorganisms in public databases with 97%
sequence similarity (Table 2). The remaining 87 OTUs (90% of
the total clones) showed <97% similarity with the most closely
related bacterial 16S rRNA gene sequences and thus represent
novel phylotypes not described in previous analyses of activated sludge communities. Of these, some displayed very low
levels of identity to any known 16S rRNA sequences and thus
could not be affiliated (Fig. 3).

Fig. 2 e Comparison and Pearson correlation analysis of nosZ gene DGGE fingerprints from the MBR and CAS over the course
of 13 months (July 2006 through July 2007). Clustering is based on the unweighted pair group method using arithmetic
average (UPGMA) algorithm. Arrows indicate the time points when switches of carbon source occurred: from acetate to
butyrate on December 5 2006 (one week before the December 2006 sludge sampling), back to acetate on April 4 2007 (three
weeks before the April 2007 sampling), and to butyrate again in the end of May (one week after the May 2007 sampling).

1134

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

Table 2 e Diversity, predicted richness and novelty of the environmental clone sequences retrieved from the MBR.a
No. of
clones
analyzed

No. of
OTUs

146

99

47

amoA

80

13

93.8

nosZ

78

22

89.7

16S rRNA

% Coverage

Chao 1
valueb

ACE
valueb

Shannon
index

OTUs 97%
cultivatedc

OTUs 97%
not yet cultivatedc

Novel OTUs
(<97%)d

303
(203, 497)
16
(13, 30)
29
(23, 56)

310
(214, 487)
17
(14, 32)
28
(24, 47)

4.38

7
(7.1, 6.8)
2
(15.4, 6.3)
0

5
(5.1, 3.4)
6
(46.2, 75.0)
0

87
(87.9, 89.7)
5
(38.5, 18.8)
22
(100, 100)

2.35
2.70

a Data are based on the November 2006 sludge clone libraries. Operational taxonomic units (OTU) were defined by a 3% and 5% difference in the
nucleic acid sequence alignment for the 16S rRNA and the amoA and nosZ genes, respectively. Accordingly, Goods coverage estimates were
calculated using a 3% and 5% cutoff for the 16S rRNA and functional genes, respectively.
b Numbers in parentheses are lower and upper 95% confidence intervals for the nonparametric richness estimators Chao1 and ACE.
c Sharing 97% (95% for the amoA and nosZ genes) sequence similarity to the closest sequence with a cultivated representative.
d Sharing 97% (95% for the amoA and nosZ genes) sequence similarity to the closest sequence without a cultivated representative.

Screening and comparative sequencing of 80 amoA clones

resulted in a total of 13 OTUs (Table 2). Phylogenetic analysis
revealed a low diversity of AOB in the MBR (Fig. S4). All
retrieved amoA sequences belonged to the b-Proteobacteria
AOB and, with only one exception, affiliated with the

Nitrosomonas oligotropha-like cluster. The vast majority of

these sequences (53% of the total OTUs and 72% of the clone
library) formed two individual clades separating apart from
well-described nitrifier species but with amoA clones previously recovered from other activated sludge systems as their

Fig. 3 e Evolutionary distance dendrogram constructed using the neighbor-joining method and showing the phylogenetic
affiliation of the 16S rRNA gene sequences retrieved from the MBR (activated sludge collected in November 2006). A region
from position 28 to position 927 (Escherichia coli numbering) was included in the phylogenetic analysis. The number of OTUs
and percentage of total clones (numbers shown to the left of the clades) belonging to the corresponding division in the clone
library are indicated in parentheses. The scale bar represents the substitution per nucleotide position. The tree was
calculated using the ARB software package. Members of the Archaea domain were used as outgroups (not shown).

1135

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

Flavobacterium
Terrimonas

Ferribacterium
Zoogloea

Niastella
Unclassified

Flavobacterium

Haliscomenobacter

Rhodoferax
Unclassified

Unclassified
Aquabacterium

Rhodoferax

Ferribacterium

Unclassified

Acidovorax
Thiobacter
Diaphorobacter
Propionivibrio
Zoogloea

Comamonas
Derxia
Dechloromonas

-Proteobacteria November 2006

Curvibacter

Aquabacterium
Hydrogenophaga

Terrimonas

-Proteobacteria February 2007

Bacteroidetes November 2006

Bacteroidetes February 2007

Fig. 4 e Distribution of genera within the dominant groups (b-Proteobacteria and Bacteroidetes) in the November 2006 and
February 2007 MBR 16S rRNA clone libraries. The total number of clones used for each gene library analysis was 146 and
113, respectively. Activated sludge samples were obtained approximately three weeks before and 11 weeks after the first
carbon source change (from acetate to butyrate on December 5 2006). Bacterial 16S rRNA gene sequences are assigned to
genera by using the Ribosomal Database Project (RDP) Classifier. Percentages are based on the number of clones of bProteobacteria and Bacteroidetes in each library.

closest relatives. No Nitrosospira-related sequences were

detected in the amoA library.
Sequence analysis of 78 randomly selected clones from the
nosZ library revealed a total of 22 OTUs (Table 2). Of these, 20
sequences exhibited <90% similarity with their closest neighbors in public databases and were not closely related to any
cultured denitrifiers (71e84% similarity) (Fig. S5). In marked
contrast to the 16S rRNA and amoA gene sequences, the vast
majority of the nosZ sequences (20 OTUs and 93% of the gene
library) do not have their database matches from activated
sludge processes; instead, they are most closely related to
environmental clones recovered from soils and sediment.
Phylogenetic analysis grouped all nosZ sequences into two
major clusters whose cultured members are a- and b-Proteobacteria, although in some cases they formed subclades clearly
separated from other known bacterial species (Fig. S5).
Additional clone libraries of the 16S rRNA and nosZ gene
were generated for MBR sludge collected in February 2007
when stabilization of microbial community after the disturbance (change in external carbon) was evidenced by DGGE
analysis (Figs. 1 and 2). The data were then compared with
those from the November 2006 sample. Although comparative
16S rRNA sequencing revealed that the b-Proteobacteria,
Bacteroidetes and a-Proteobacteria remained the predominant groups within the February 2007 community, relative
abundance of clones in the gene library increased by 44 and
40% for the first two divisions and decreased by 42% for the
latter (Fig. S6). Large differences between the two 16S rRNA
gene libraries were also detected at finer phylogenetic levels of
resolution. Of the 13 b-Proteobacteria-affiliated genera identified, only four were found common in both communities and
shifts in relative abundance of these genera were observed,
particularly for Zoogloea, Rhodoferax, and the unclassified
b-Proteobacteria that were only most closely related to environmental clones (Fig. 4). Similarly, significant shifts in
abundance distribution of genera affiliated with the Bacteroidetes were found between the two clone libraries, with
Flavobacterium clearly enriched and the unclassified Bacteroidetes less favored in the February 2007 community (Fig. 4).
At the phylotype level, only 17 OTUs were detected in both
bacterial communities, and they were disproportionately
distributed in the respective clone libraries. Most notably, of

the three most abundant OTUs in either clone library, two

were not detected or could only be occasionally recovered in
the other library. Phylogenetic analysis identified in the
evolutionary distance trees sample-specific clusters or clusters where clones were disproportionately distributed in the
two 16S rRNA gene libraries (data not shown). In addition,
rarefaction (Fig. S3) and diversity index analysis (data not
shown) revealed significantly different relative species richness in the two MBR sludges, the February 2007 sample having
lower diversity.
Clone library analysis of the nosZ gene also indicated
a significant deviation in denitrifying bacteria from the
November 2006 community. The library constructed from the
February 2007 sample was dominated (80%) by a diverse group
of sequences clustered with cultured members from the
b-Proteobacteria (data not shown). Of the 18 OTUs identified,
11 were not detected in the November 2006 sample. Significant shifts in some of the dominant OTUs were also observed
between the two denitrifying communities. Rarefaction
analysis indicated that the November 2006 nosZ library had
a relative species richness significantly higher than that of the
February 2007 library (Fig. S3), similar results being obtained
by the nonparametric estimators of species richness,
including Chao 1and ACE values (29 versus 18 and 28 versus
20, respectively).

4.

Discussion

In a previous pilot-scale study, Luxmy et al. (2000) used

molecular fingerprinting technique to demonstrate that
bacterial community in the MBR fed with raw sewage was
different from that in the CAS process. In agreement with this,
our DGGE analysis revealed process type-specific microbial
communities, the full-scale MBR and CAS systems harboring
distinct total bacterial communities and N-cycling groups
with the corresponding community fingerprints separating
apart in the UPGMA dendrograms independent of seasons
(Figs. 1 and 2 and Fig. S2). Although these results need to be
confirmed by further research at other full-scale plants, we
believe that our findings are practically representative given
that samples spanning a period of over one year were

1136

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

analyzed for both systems. Our study site is particularly

interesting for such a comparison since the two processes are
operated in parallel and treat the same wastewater. Importantly, the MBR was originally inoculated with activated
sludge from the CAS plant. The differences in microbial
communities of the two environments are likely due to the
complete rejection of biomass by the membrane and the
contrasting operating conditions applied (e.g., sludge age, F/M
ratio, flow consistency) in the MBR which would greatly affect
microbial community structure and physiological state of
mixed liquor. The more complex 16S rRNA DGGE profiles
observed for the CAS samples are most probably related to the
bigger temporal fluctuations in substrate concentrations in
this system since it is operated at a variable flow mode, thus
allowing for diverse organisms with different growth kinetics
to be transiently competitive. Compared with the CAS
process, the MBR has an additional denitrification compartment and this may also have a significant influence on the
activated sludge community. Finally, the impact of addition of
readily biodegradable carbon substrates on the MBR microbial
functional groups might be important since less variable
substrate composition may lead to less complex microbial
diversity (Hagman et al., 2008).
Our comparative 16S rRNA sequence analysis has shown
that, similar to the conventional treatment systems (Wagner
and Loy, 2002), Proteobacteria- (b and a) and BacteroidetesOTUs were most frequently detected in the MBR community
(Fig. 3). However, the vast majority of retrieved sequences
represent putative phylotypes never obtained in culture or not
described in previous investigations of activated sludge
ecosystems. Similarly, novel bacterial sequences were abundantly recovered from lab-scale MBRs treating domestic
wastewater and operated under different conditions (Huang
et al., 2008). These results are consistent with molecular
characterizations of microbial communities in wastewater
treatment ecosystems which have revealed that previously
unrecognized, as yet uncultured bacteria are responsible for
most key processes in CAS and biofilm reactors (Wagner and
Loy, 2002). In-depth molecular inventories of the relatively
novel, microbially diverse MBR systems are needed, as this
more comprehensive picture would help reveal factors influencing process efficiency and stability and understand how
selection pressures imposed by membrane separation affect
community structure and diversity within bioreactors. Ultimately, a thorough knowledge of the microbial community
structure in the mixed liquor biomass will provide opportunities to elucidate the underlying mechanisms for biofilm
development and maturation on the membrane surface which
lead to severe irreversible fouling, a major obstacle for the
wider application of MBRs for wastewater treatment.
Ammonium is removed from wastewater via nitrification
and denitrification. Previous investigations have shown that
a wide variety of different b-proteobacterial ammonia
oxidizers occur in nitrifying WWTPs, with members affiliated
with the Nitrosomonas europaea/eutropha lineage, the Nitrosococcus mobilis lineage and the Nitrosomonas marina cluster
were most frequently detected (reviewed by Wagner and Loy,
2002). However, Manser et al. (2005) reported more recently
that bacteria related to the N. oligotropha lineage dominated
the AOB populations in a pilot MBR plant treating domestic

wastewater. Similarly, our comparative amoA sequencing

revealed that the AOB community in the full-scale MBR is
simple with the detected OTUs almost exclusively affiliated
with the N. oligotropha cluster (Fig. S4). Since members of the
N. oligotropha lineage have relatively high substrate affinity
(Koops and Pommerening-Roser, 2001), they may gain
ecological advantages in MBRs for domestic wastewater
treatment where the food (ammonia) to microorganisms
(ammonium oxidizers) ratios are generally low. The presence
of an AOB community low in species richness in the MBR
might render its nitrification more susceptible to perturbation.
Microbial population dynamics have been well-documented in traditional treatment systems (Lee et al., 2002;
Wells et al., 2009). Although substantial fluctuations in
bacterial community structure have been demonstrated
previously in laboratory- or pilot-scale MBRs (Stamper et al.,
2003; Miura et al., 2007; Huang et al., 2008), the bacterial population in the full-scale MBR was rather stable during the
period before external carbon change (Figs. 1 and 2). Temporal
fluctuations in influent wastewater characteristics and
temperature in the bioreactors are generally thought to affect
microbial community structure in WWTPs. The impact of feed
strength should be less significant for MBRs since they are
typically operated at lower F/M ratios. Likewise, the influence
of invasion by exotic species via influent and extinction of
bacteria from the indigenous community are less important
since MBRs tend to maintain high biomass concentrations and
limited sludge wasting by membrane rejection. In addition,
the limited variations in bioreactor temperature at our study
site would also favor a stable bacterial community in the MBR.
Readily biodegradable carbon sources have been intermittently added to the anoxic zones of conventional treatment
plants to enhance denitrification (Hallin and Pell, 1998;
Hasselblad and Hallin, 1998). Although the effectiveness of
this approach to increase the denitrification capacity is well
documented, little is known about the impacts of external
carbon on the microbial community structure and other biological processes of activated sludge. Our DGGE analysis
revealed profound different banding patterns of the 16S rRNA
and nosZ gene after the carbon source was changed in
December 2006 from acetate to butyrate in the full-scale MBR,
whereas similar shifts in microbial community profile were
not detected in the parallel-running CAS process during the
same period (Figs. 1 and 2). Clone library analysis further
supported that total bacterial and denitrifying community
composition differed significantly between MBR sludge
obtained before and after the external carbon change, the
later sample having lower 16S rRNA and nosZ gene diversity.
Hagman et al. (2008) have recently reported substrate
(external carbon source) preferences by involved microbial
consortia, with acetate being used by more diverse bacterial
populations which are among the dominant denitrifying
groups typically present in activated sludge ecosystems.
Similarly, Ginige et al. (2005) have demonstrated that the
addition of acetate (to enhance denitrification) caused
a dramatic shift in the microbial community structure, with
denitrifiers from the families Comamonadaceae and Rhodocyclaceae in the b-Proteobacteria being rapidly enriched in the
laboratory-scale bioreactor originally seeded with activated
sludge from a full-scale plant (which had not been exposed to

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

any form of external carbon source augmentation). Interestingly, our matching clone screening showed that 16S rRNA
sequences corresponding to some of the major shifting bands
in the total-community DGGE profiles of the MBR could be
assigned to as yet uncultured Ferribacterium/Dechloromonasrelated bacteria and Zoogloea species in the family Rhodocyclaceae (Table S2), and members of genera Dechloromonas and
Zoogloea have previously been found consistently present in
a denitrifying fluidized bed bioreactor (Gentile et al., 2006).
These indicate a possible impact of the change in carbon
source on the denitrifier community in the full-scale bioreactor. The additional fluctuations in community profiles from
April through July 2007 might reflect response of the mixed
liquor bacterial populations to the subsequent switches of
carbon source between butyrate and acetate. In spite of this,
the overall process performance of the MBR remained relatively stable during the study period. Other studies have
demonstrated that dynamic bacterial populations may still
maintain stable performance in bioreactor treatment systems
(Stamper et al., 2003; Fernandez et al., 1999). It is also possible
that both microbial community structure and function were
affected by the perturbation but function returned to its
original state shortly, and our monthly sampling was not
frequent enough to detect such changes. In a previous investigation of a pilot-scale denitrifying bioreactor, both function
and community structure were disrupted by disturbances, but
nitrogen removal recovered within a few days, indicating
a high functional resilience (Gentile et al., 2006). It was suggested that flexible community structure and potentially the
activity of minor members under nonperturbation conditions
promotes system recovery. It remains unknown whether the
adaptation and shift in the MBR microbial community influence other biological processes and stability of activated
sludge at our study site.

5.

Conclusions

This represents the first report using cultivation-independent

molecular approaches to elucidate the phylogenetic composition and diversity of the microbial communities in a fullscale MBR for municipal wastewater treatment. Our data have
shown that total community bacteria and N-cycling groups in
the MBR were constantly distinct from those in the parallelrunning CAS bioreactor, reflecting the contrasting environmental and operational conditions in the two treatment
systems. The frequent detection of novel 16S rRNA and nosZ
gene sequences in the clone libraries indicate the predominance and thus the potentially significant role of previously
unrecognized bacteria in the carbon and nutrient removal
processes. This study is an important step toward exploring
the relationship between bacterial diversity and biogeochemical function within the MBR ecosystem.

Acknowledgements
Li-Nan Huang was partially supported by the European
Commission through a Marie Curie Fellowship (MIF1-CT-2005-

1137

021768). We thank G. Borgmans for technical supports and H.

Paar for the DGGE analysis, and the two anonymous reviewers
for providing thoughtful and constructive comments on the
manuscript.

Appendix. Supplementary material

Supplementary data related to this article can be found online
at doi:10.1016/j.watres.2010.11.008.

references

Bond, P.L., Hugenholtz, P., Keller, J., Blackall, L.L., 1995. Bacterial
community structures of phosphate-removing and
nonphosphate-removing activated sludges from sequencing
batch reactors. Applied and Environmental Microbiology 61,
1910e1916.
Dojka, M.A., Hugenholtz, P., Haack, S.K., Pace, N.R., 1998.
Microbial diversity in a hydrocarbon- and chlorinated-solvent
contaminated aquifer undergoing intrinsic bioremediation.
Applied and Environmental Microbiology 64, 3869e3877.
Fenu, A., Roels, J., Wambecq, T., De Gussem, K., Thoeye, C., De
Gueldre, G., Van De Steene, B., 2010. Energy audit of a full scale
MBR system. Desalination 262, 121e128.
Fernandez, A., Huang, S.Y., Seston, S., Xing, J., Hickey, R.,
Criddle, C., Tiedje, J., 1999. How stable is stable? Function
versus community composition. Applied and Environmental
Microbiology 65, 3697e3704.
Gentile, M., Yan, T., Tiquia, S.M., Fields, M.W., Nyman, J., Zhou, J.,
Criddle, C.S., 2006. Stability in a denitrifying fluidized bed
reactor. Microbial Ecology 52, 311e321.
Ginige, M.P., Keller, J., Blackall, L.L., 2005. Investigation of an
acetate-fed denitrifying microbial community by stable
isotope probing, full-cycle rRNA analysis, and fluorescent in
situ hybridization-microautoradiography. Applied and
Environmental Microbiology 71, 8683e8691.
Hagman, M., Nielsen, J.L., Nielsen, P.H., Jansen, J.la C, 2008. Mixed
carbon sources for nitrate reduction in activated sludgeidentification of bacteria and process activity studies. Water
Research 42, 1539e1546.
Hallin, S., Pell, M., 1998. Metabolic properties of denitrifying
bacteria adapting to methanol and ethanol in activated
sludge. Water Research 32, 13e18.
Hasselblad, S., Hallin, S., 1998. Intermittent addition of external
carbon to enhance denitrification in activated sludge. Water
Science and Technology 37, 227e233.
Huang, L.N., De Wever, H., Diels, L., 2008. Diverse and distinct
bacterial communities induced biofilm fouling in membrane
bioreactors operated under different conditions.
Environmental Science and Technology 42, 8360e8366.
Huang, L.N., Zhou, H., Zhu, S., Qu, L.H., 2004. Phylogenetic
diversity of bacteria in the leachate of a full-scale recirculating
landfill. FEMS Microbiology Ecology 50, 175e183.
Judd, S., 2007. The status of membrane bioreactor technology.
Trends in Biotechnology 26, 109e116.
Koops, H.P., Pommerening-Roser, A., 2001. Distribution and
ecophysiology of the nitrifying bacteria emphasizing cultured
species. FEMS Microbiology Ecology 37, 1e9.
Lee, N., Jansen, J., Aspegren, H., Dircks, K., Henze, M., Schleifer, K.
H., Wagner, M., 2002. Population dynamics and in situ
physiology of phosphorus-accumulating bacteria in
wastewater treatment plants with enhanced biological
phoshorus removal operated with and without nitrogen
removal. Water Science and Technology 46, 163e170.

1138

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 2 9 e1 1 3 8

Low, E.W., Chase, H.A., 1999. The effect of maintenance energy

requirements on biomass production during wastewater
treatment. Water Research 33, 847e853.
Ludwig, W., Strunk, O., Westram, R., Richter, L., Meier, H.,
Yadhukumar Buchner, A., Lai, T., Steppi, S., Jobb, G., 2004.
ARB: a software environment for sequence data. Nucleic Acids
Research 32, 1363e1371.
Luxmy, B.S., Nakajima, F., Yamamoto, K., 2000. Analysis of
bacterial community in membrane separation bioreactors by
fluorescent in situ hybridization (FISH) and denaturing
gradient gel electrophoresis (DGGE) techniques. Water Science
and Technology 41, 259e268.
Manser, R., Gujer, W., Siegrist, H., 2005. Membrane bioreactor versus
conventional activated sludge system: population dynamics of
nitrifiers. Water Science and Technology 52, 417e425.
Miura, Y., Hiraiwa, M.N., Ito, T., Itonaga, T., Watanabe, Y.,
Okabe, S., 2007. Bacterial community structures in MBRs
treating municipal wastewater: relationship between
community stability and reactor performance. Water
Research 41, 627e637.
Muller, E.B., Stouthamer, A.H., van Verseveld, H.W., Eikelboom, D.
H., 1995. Aerobic domestic wastewater treatment in a pilot
plant with complete sludge retention by cross-flow filtration.
Water Research 29, 1179e1189.
Nicolaisen, M.H., Ramsing, N.B., 2002. Denaturing gradient gel
electrophoresis (DGGE) approaches to study the diversity of
ammonia-oxidizing bacteria. Journal of Microbiological
Methods 50, 189e203.

Nogueira, R., Melo, L.F., Purkhold, U., Wuertz, S., Wagner, M.,
2002. Nitrifying and heterotrophic population dynamics in
biofilm reactors: effects of hydraulic retention time and the
presence of organic carbon. Water Research 36, 469e481.
Schloss, P.D., Handelsman, J., 2005. Introducing DOTUR,
a computer program for defining operational taxonomic units
and estimating species richness. Applied and Environmental
Microbiology 71, 1501e1506.
Stamper, D.M., Walch, M., Jacobs, R.N., 2003. Bacterial population
changes in a membrane bioreactor for graywater treatment
monitored by denaturing gradient gel electrophoretic analysis
of 16S rRNA gene fragments. Applied and Environmental
Microbiology 69, 852e860.
Throback, I.N., Enwall, K., Jarvis, A., Hallin, S., 2004. Reassessing
PCR primers targeting nirS, nirK and nosZ genes for
community surveys of denitrifying bacteria with DGGE. FEMS
Microbiology Ecology 49, 401e417.
Ueda, S., Earle, R.L., 1972. Microflora of activated sludge. Journal
of General and Applied Microbiology 18, 239e248.
Wagner, M., Loy, A., 2002. Bacterial community composition and
function in sewage treatment systems. Current Opinion in
Biotechnology 13, 218e227.
Wells, G.F., Park, H.D., Yeung, C.H., Eggleston, B., Francis, C.A.,
Criddle, C.S., 2009. Ammonia-oxidizing communities in
a highly aerated full-scale activated sludge bioreactor:
betaproteobacterial dynamics and low relative abundance
of Crenarchaea. Environmental Microbiology 11,
2310e2328.