ARTICLE IN PRESS

WAT E R R E S E A R C H 41 (2007) 134– 144

Available at www.sciencedirect.com

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

Methanogenic population dynamics and performance of an

anaerobic membrane bioreactor (AnMBR) treating swine
manure under high shear conditions

Sudini I. Padmasiria, Jiangzhao Zhanga, Mark Fitchc, Birgir Norddahld,

Eberhard Morgenrotha,b, Lutgarde Raskine,
a
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 205 N Mathews Ave, IL 61801, USA
b
Department of Animal Sciences, University of Illinois at Urbana-Champaign, 205 N Mathews Ave, IL 61801, USA
c
Department of Civil, Architectural, and Environmental Engineering, University of Missouri-Rolla, 222 Butler Carlton Civil Engineering Hall,
MO 65409, USA
d
TUFCO, Teknisk Udviklings og Forsknings Center Odense, Technical Development and Research Centre Odense, University of Southern
Denmark, Campusvej 55, DK 5230 Odense M, Denmark
e
Department of Civil and Environmental Engineering, The University of Michigan, 1351 Beal Ave, Ann Arbor, MI 48109 2125, USA

ar t ic l e i n f o A B S T R A C T

Article history: A 6-L, completely mixed anaerobic bioreactor with an external ultrafiltration membrane
Received 25 February 2006 module was operated for 300 days to evaluate the startup and performance of an anaerobic
Received in revised form membrane bioreactor (AnMBR) treating swine manure. The reactor had a successful
17 August 2006 startup at the initial loading rate of 1 g volatile solids (VS)/L/day. After a two-fold increase in
Accepted 22 September 2006 loading rate followed by a sudden, two-fold increase in flow velocity through the
Available online 15 November 2006 membrane module on day 75, the performance of the AnMBR deteriorated as measured
Keywords: by volatile fatty acid (VFA) accumulation, decrease in pH, and decrease in biogas
Anaerobic membrane bioreactor production. The methanogenic population dynamics in the reactor were monitored with
AnMBR terminal restriction fragment length polymorphism (T-RFLP). Changes in the relative levels
Shear of Methanosarcinaceae and Methanosaetaceae were consistent with changes in VFA
Swine manure concentrations, i.e., high and low levels of acetate corresponded to a high abundance of
Methanogens Methanosarcinaceae and Methanosaetaceae, respectively. The levels of hydrogenotrophic
T-RFLP methanogens of the order of Methanomicrobiales increased during decreased reactor
performance suggesting that syntrophic interactions involving hydrogenotrophic metha-
nogens remained intact regardless of the degree of shear in the AnMBR.
& 2006 Elsevier Ltd. All rights reserved.

1. Introduction effluent quality, and can be smaller in size than conventional

Anaerobic digesters with membrane separation units (anae-
robic membrane bioreactors [AnMBRs]) facilitate retention of
microorganisms and allow operation with high biomass
concentrations. Thus, AnMBRs are expected to provide more
efficient digestion, higher methane production, and better
anaerobic digesters. These advantages, together with in-
creased stringency in waste disposal for animal production
facilities, have led to pilot scale testing of AnMBRs to treat
swine waste (du Preez et al., 2005; Lee et al., 2001). The main
challenge for AnMBRs has been fouling of membrane units
(Choo and Lee, 1996; Elmaleh and Abdelmoumni, 1997; Choo

Corresponding author. Tel.: +1 734 647 6920; fax: +1 734 763 2275.
E-mail address: raskin@umich.edu (L. Raskin).
0043-1354/$ - see front matter & 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2006.09.021
 ARTICLE IN PRESS
WAT E R R E S E A R C H
et al., 2000; He et al., 2005). Membrane fouling is the result of
adsorption of organic matter, precipitation of inorganic
matter, and adhesion of microbial cells to the membrane
surface (Choo and Lee, 1996). High fluid flow velocities
resulting in high shear rates at the surface of the membrane
can be used in AnMBRs with external membrane units to help
reduce membrane fouling caused by adhesion of biomass and
colloidal organic matter to the membrane surface (Stephen-
son et al., 2000). However, the digestion efficiency in AnMBRs
may be negatively affected by exposure of biomass to high
shear conditions.
High mixing and shear intensities can negatively influence
anaerobic digestion performance. Brockmann and Seyfried
(1996) demonstrated a 50% loss in the specific activity of
anaerobic biomass treating potato starch wastewater in an
AnMBR with cross-flow microfiltration following re-circula-
tion of the entire reactor content through a membrane unit 20
times. Similar results were presented in a study by Stroot et
al. (2001) and McMahon et al. (2001), who demonstrated that
vigorously and continuously mixed solid waste digesters
(power input of 1.5 W/L) exhibited unstable performance,
whereas minimally mixed digesters (thoroughly shaken by
hand for 2 min each day) performed very well. They also
demonstrated that continuously mixed unstable digesters
were quickly (within 3 weeks) stabilized by reducing the
degree of mixing. Kim et al. (2002) obtained similar perfor-
mance results in anaerobic digesters under minimal and
vigorous mixing conditions. They compared the stability of
mesophilic and thermophilic anaerobic digesters under
different loading rates and mixing conditions. During startup,
both the minimally mixed mesophilic and thermophilic
reactors showed superior performance and achieved stable
operation in shorter time periods compared to continuously
mixed reactors. Under steady-state conditions with a feed of
4% solids and a hydraulic retention time (HRT) of 20 days,
these two minimally mixed reactors produced more biogas
and less volatile fatty acid (VFA) in each temperature category
than the corresponding continuously mixed reactors. When
the loading rate was increased, the thermophilic minimally
mixed reactors were the last to fail at an organic loading rate
of 20 g volatile solids (VS)/L/day.
While it is clear from the above discussion that high mixing
and shear intensities have a negative impact on anaerobic
digestion performance, the reasons behind such observations
are less obvious. Brockmann and Seyfried (1996) suggested
that the observed reduced performance in their AnMBR may
be the result of a physical interruption of the syntrophic
association between proton reducing acetogenic bacteria and
their methanogenic partners. This suggestion seems logical
since high shear rates may result in break-up of microbial
aggregates and a disruption of the juxtaposition of relevant
microbes. Similarly, Kim et al. (2002) concluded that their
results imply that minimal mixing resulted in microbial
consortia proximity, and that this strategy can be used to
alleviate the problem of poor effluent quality in thermophilic
systems. Likewise, Stroot et al. (2001) and McMahon et al.
(2001) hypothesized that vigorous mixing may have inter-
rupted syntrophic associations. In contrast to the other
studies, their hypothesis was supported by microbial popula-
tion dynamics results, which indicated that the relative
41 (20 07) 13 4 – 144 135
abundance of syntrophic propionate oxidizing bacteria,
saturated fatty acid-b-oxidizing syntrophs, and various
groups of methanogens was lower under vigorous mixing
conditions (McMahon et al., 2001, 2004). Alternately, their
observations may be explained by differences in the rates of
hydrolysis and fermentation between vigorously and mini-
mally mixed systems. Vigorous and continuous mixing may
have promoted rapid hydrolysis and fermentation, resulting
in production of fermentation products at rates greater than
the utilization rates of such products by syntrophs, whereas
minimally mixed conditions may have resulted in slower
hydrolysis and fermentation, allowing immediate consump-
tion of fermentation products by syntrophic bacteria and
methanogens (Stroot et al., 2001). This second hypothesis
assumes interactions between syntrophic bacteria and
methanogens are not disrupted, but implies these micro-
organisms are not able to keep up with the high production of
fermentation products. Their population dynamics data (i.e.,
lower abundance of syntrophic bacteria and methanogens
under vigorously mixed conditions) also support this second
hypothesis.
From the above discussion, it can be concluded that the
degree of shear to which the biomass in an AnMBR is exposed
can impact the performance and stability of anaerobic
digestion. This may be even more critical when complex
waste streams containing substantial levels of suspended
solids, such as swine manure, are to be treated. The overall
objective of this study was to evaluate whether high shear in
AnMBRs with external membrane units can decrease process
efficiencies and whether these changes are reflected in the
microbial community structure. This paper presents results
collected with a laboratory-scale AnMBR and focuses on
reactor and membrane performance and characterization of
methanogenic population dynamics with terminal restriction
fragment length polymorphism (T-RFLP).
2. Materials and methods
2.1. AnMBR operation
A 6-L, completely mixed anaerobic digester (supplied by
Odense College of Engineering, Odense M, Denmark) was
operated to treat swine manure (Fig. 1). Solid–liquid separa-
tion was achieved using a tubular ultrafiltration membrane of
1-m length and 12-mm diameter with a total surface area of
0.0377 m2. The membrane was made of polyethersulfone
(Weir Envig, Paarl, South Africa) with a molecular weight cut-
off of 20,000 Da. The reactor content was pumped through the
membrane module using a progressing cavity pump (NM021,
Netzsch, Germany) controlled with a variable frequency
converter (Danfoss, Loves Park, IL, USA). The trans-membrane
pressure (TMP) was varied using a back pressure valve (Tru-
Trol BBCA-CAT2, Tru-tech, Mars, PA, USA). Reactor operation
and data acquisition were controlled using Lab View (National
Instruments, Austin, TX, USA) on a PC with a FieldPoint
external input/output card (National Instruments, Austin, TX,
USA). The reactor temperature was kept constant at 3771 1C
with a water jacket.
 ARTICLE IN PRESS
136 WA T E R R E S E A R C H
Back Pressure
Mixer
Valve
To Gas Meter
Feed
4°C P3
Anaerobic
Digester
37°C
P2
Waste/ Check
Sampling Valve
Fig. 1 – Schematic diagram of anaerobic membrane bioreactor (AnMBR). P1 is a progressing cavity pump and P2, P3, and P4
are peristaltic pumps.
The AnMBR was inoculated with a 1:1:1 (v:v:v) mixture of
sludge dredged from a swine lagoon (near Vichy, MO, USA),
anaerobic granules from an upflow anaerobic sludge blanket
(UASB) reactor treating brewery wastewater (Anheuser-Busch,
Baldwinsville, NY, USA), and anaerobic sludge from the
primary anaerobic sludge digester of the Urbana Champaign
Sanitary District North-East wastewater treatment plant
(Urbana, IL, USA). Swine manure was collected from finisher
buildings (University of Illinois at Urbana-Champaign, Swine
Research Farm, Urbana, IL, USA) and blended to ensure break
up of large particles. The homogenized waste was stored in
containers (one for each 24-h period of operation) at 20 1C.
Every day, a new batch of feed was thawed overnight and
diluted with tap water to the desired concentration of 6 g VS/L
before addition to an influent storage tank kept at 4 1C which
was continuously mixed. The reactor was fed 125 mL every 3 h
from the influent storage tank, which corresponds to a
loading rate of 1 g VS/L/day (1 g VS corresponds to approx.
1 g COD). The HRT of the system was controlled at 6 days by
returning excess permeate back to the digester. The startup-
loading rate was doubled to 2 g VS/L/day on day 53 and tripled
to 3 g VS/L/day on day 186.
To characterize membrane fouling, the initial water flux of the
clean membrane unit was measured. The permeate flux (J)
through the membrane was measured with a graduated
cylinder and a timer under different values of TMP. The TMP
was controlled by adjusting the backpressure valve (Fig. 1) and
was measured online with a pressure sensor (Red valve series
42, Red Valve Co., Inc., Carnegie, PA, USA). During operation, the
permeate flow rate was measured with a variable area flow
meter (Gilmont Instruments, Racine, WI, USA). The cross flow
velocity through the membrane was measured with a magnetic
flow meter (Endress+Hauser Promag 50 H, Greenwood, IN, USA).
The total membrane resistance (Rtotal) was calculated as follows:
TMP
Rtotal ¼ , (1)
mJ
where m is the viscosity of the permeate estimated using the
viscosity of pure water.
41 (2007) 134– 144
Flow Pressure Flow
UF
Tubular
Membrane Effluent
P1
P1
Frequency Converter
2.2. Chemical analyses
The biogas production was measured with a wet-test gas
meter (Schlumberger Industries, Dordrecht, The Netherlands)
and the methane composition in the biogas was determined
using a gas chromatograph (Perkin Elmer AutoSystem gas
chromatograph, Norwalk, CT, USA) equipped with a GS-Q
column and a flame ionization detector operated isother-
mally at 40 1C using helium at a flow rate of 26 mL/min. The
suspended solids (SS), volatile suspended solids (VSS), and
soluble chemical oxygen demand (SCOD) in the reactor, and
total COD (TCOD) and SCOD in the effluent were determined
according to Standard Methods (APHA, WEF, AWWA, 1998). The
total VFA concentration and bicarbonate alkalinity in the
reactor were measured by a simple titration technique
(Anderson and Yang, 1992). A parameter a, defined as the
ratio of VFA concentration over bicarbonate alkalinity, was
also calculated. This parameter has been found useful to
monitor reactor upsets or failures (Poggi-Varaldo and Olesz-
kiewicz, 1992). The individual VFAs were analyzed using a
high-pressure liquid chromatograph (Waters 486 Tunable
Absorbance detector, Millipore Corporation, Milford, MA,
USA) with an ion exclusion column of 300 mm  7.8 mm
(BIO-RAD, Aminex HPX-87 H, Hercules, CA, USA) the mobile
phase was 0.005 M H2SO4 at flow rate of 0.6 mL/min. The
ammonia+ammonium (NH3–N) concentration in the effluent
was determined using a HACH test kit (Method 8038, Hach,
Loveland, CO, USA).
2.3. DNA extraction and PCR
DNA was extracted using a combined detergent and bead
beating procedure (UltraClean Soil DNA Kit, Mo Bio Labora-
tories Inc., Solana Beach, CA, USA) according to the manu-
facturer’s instructions. The polymerase chain reaction (PCR)
targeting 16S ribosomal RNA (rRNA) genes of Archaea was
performed using the A109f (S-D-Arch-0109-a-S-17) and A912r
(S-D-Arch-0912-a-A-20) archaeal primer pair described pre-
 ARTICLE IN PRESS
WAT E R R E S E A R C H 41 (20 07) 13 4 – 144 137

viously (Lueders and Friedrich, 2000). The 50 end of the

Specificity
forward primer was labeled with 6-carboxyfluorescein (FAM)

90.2
94.7
(%)

100
for terminal restriction fragment length polymorphism
(T-RFLP) analysis. The PCR reaction mixture contained 1  PCR
buffer, 2 mM MgCl2, 0.2 mM of each deoxynucleoside tripho-
sphate (dNTP), 0.2 mM each of forward and reverse primers,

Enzyme MseI

Coverage
approximately 10 ng of the DNA template and 2.5 U of Ex Taq

70.3
69.2
(%)
DNA polymerase (TaKaRa Biomedicals, Otsu, Shiga, Japan) in

47
a final volume of 50 ml. The PCR was performed in a thermal
cycler (PTC-200 DNA Engine, MJ Research Inc., Reno, NV, USA).

Table 1 – Size of T-RFs, coveragea, and specificityb of T-RFLP with primer pair A109f–A912r for three restriction enzymes AluI, MaeI, and MseI
The amplification was done with one denaturation step at
94 1C for 2 min, followed by 30 cycles of denaturation at 94 1C

Size (bp)
for 30 s, annealing at 55 1C for 30 s, and extension at 72 1C for

103
459
420
1 min with a final extension of 8.5 min.

2.4. T-RFLP and data analysis

Specificity
Amplicons from duplicate PCR reaction mixtures were pooled

86.2
71.6
(%)

100

1.7
using a PCR purification kit (UltraClean PCR clean-up Kit, Mo
Bio Laboratories Inc., Carlsbad, CA, USA). The purified
fluorescently labeled PCR products were digested with

Enzyme MaeI
restriction enzymes AluI (New England Biolabs Inc., Beverly,

Coverage
CA, USA), MaeI (Fermentas Inc., Hanover, MD, USA), and MseI

70.9

30.7
(%)

100
95

Specificity ¼ number of target species with specific fragment size/total number of species with specific fragment size.
(New England Biolabs Inc.) for 3 h at 37 1C followed by an
enzyme inactivation step at 65 1C for 20 min. The digested
samples were treated by ethanol precipitation to remove
excess salt. The fluorescently labeled terminal restriction

Size (bp)
fragments (T-RFs) obtained in this manner were separated by

400

502
169

169
capillary electrophoresis (ABI Prism 3730  l Analyzer, Applied
Biosystems, Foster City, CA, USA) at the W.M. Keck Center for

Coverage ¼ number of target species with specific fragment size/total number of target species.
Comparative and Functional Genomics at the University of
Illinois at Urbana-Champaign Biotechnology Center (Urbana,
Specificity

IL, USA) to determine the number and size of T-RFs obtained

51.8

94.3

22.7
(%)

100
8.7
from each sample. Fragment analysis was conducted using
GeneMapperTM Version 3.7 software (Applied Biosystems,
Foster City, CA, USA). The three restriction enzymes were
selected using a model developed by Klein (2003) to allow
Enzyme AluI

Coverage

coverage of the majority of the known groups of methano-

76.2
84.6
72.6
54.1

71.4
(%)

gens and to target individual groups of methanogens. The

model allows in silico screening of a collection of downloaded
16S rRNA sequences for populations corresponding to T-RFs
obtained with specified restriction enzymes for a selected
primer pair. The 16S rRNA sequences of the domain Archaea
Size (bp)

used in the in silico analysis were downloaded from the

341
341
434
474

108

National Center for Biotechnology Information (NCBI) data-

base (Bethesda, MD, USA) in September 2004. The coverage
and specificity of each of the three restriction enzymes
selected are presented in Table 1.
species in

database

Each peak was given an allowance of 75 nucleotides during

No of

NCBI

172
26
73
61

14

T-RF assignment and only T-RFs resulting in peaks greater

than 50 fluorescence units were considered as operational
taxonomic units (OTUs). The relative abundance of each OTU
was calculated as the ratio of the peak area of that OTU over
the sum of the peak areas of all OTUs. Only the OTUs that
Methanosarcinaceae
Methanomicrobiales
Methanosarcinales-

Methanosarcinales-
Methanobacteriales

Methanosaetaceae

were found to have a relative abundance greater than 10% at

Methanococcales
Target group

least once during reactor operation are presented in the

results.
An experiment was carried out to check the precision
and reproducibility associated with the T-RFLP analysis.
b
a

The experiment consisted of two separate DNA extractions

 ARTICLE IN PRESS
138 WA T E R R E S E A R C H 41 (2007) 134– 144

of the same reactor sample (collected on day 291), triplicate

PCR for each extraction, triplicate enzyme digestion for each 3. Results and discussion
PCR, followed by duplicate runs by the capillary electrophor-
esis unit. The de-salting step following enzyme digestion 3.1. Reactor performance
was not replicated and, as a result, the variation associated
with restriction enzyme digestion also includes the variation 3.1.1. Day 0–52
associated with de-salting. The reproducibility was calculated The AnMBR was started with a loading rate of 1 g VS/L/day. A
as the portion of total variance that could be attributed few days after startup, the specific biogas production
to each T-RFLP step in a particular fragment size. For example, increased to 2–3 L/g VS/day (Fig. 2a). The decreases in biogas
the % total variance of the PCR step ¼ s2PCR/(sDNA extrac- production observed on day 19 and days 27–38 were due to
2 2 2 2
tion +sPCR+sDigestion+sCapilary electrophoresis). Variance compo- problems with the gas collection system. After an initial
nents were estimated with the MIVQUEO option of the increase in effluent SCOD soon after startup, the effluent
VARCOMP procedure of SAS (Version 8.1, SAS Institute Inc.). SCOD decreased and remained stable between 200 and

Days 0-52 53-90 91-115 116-300

6
(a) Specific gas production
Gas production (L /gVS/day)
Loading rate (gVSS / L/day)

5 Loading rate

0
100

SCOD removal efficiency (%)

80
6000 (b) SCOD -effluent
SCOD (mg/L)

SCOD - reactor
COD removal % 60
4000
40

2000
20

0 0

(c)
VFA concentration

3000 VFA's 6
(mg/ L as HAc)

pH
pH and α

alpha
2000 4

1000 2

0 0
Formic
(d)
Acetic
2000
Propionic
(mg/L as HAc)
Concentration

Butyric
1500

1000

500

0
0 25 50 75 100 125 150 175 200 225 250 275 300
Days

Fig. 2 – AnMBR performance data. (a) Specific biogas production and loading rate, (b) soluble COD in the reactor and the
effluent and soluble COD removal efficiency, (c) pH, a, and total VFA concentration, (d) individual VFA concentrations.
 ARTICLE IN PRESS
WAT E R R E S E A R C H
250 mg/L (Fig. 2b). The effluent TCOD was very similar to the
effluent SCOD (data not shown) and the average TCOD and
SCOD removal efficiencies were over 96% and 86%, respec-
tively, during this period. The total VFA concentrations in the
reactor were also low and averaged 250 mg/L as acetic acid
without significant variations in the individual VFA concen-
trations (Fig. 2c and d). The excellent performance of the
reactor during startup also was reflected in stable pH (average
of 7.5) and a values (average of 0.2) (Fig. 2c). Poggi-Varaldo and
Oleszkiewicz (1992) determined that an a value of 1.0
corresponded to the threshold of stability for anaerobic
codigestion of the organic fraction of municipal solid waste
(OFMSW) and sewage sludge and Ripley et al. (1986) observed
that an a value less than 0.3 was required for successful
digestion of poultry manure.
The observed performance suggests that the inoculum,
which consisted of biomass collected from three different
anaerobic environments, provided sufficient levels of the
relevant populations for the anaerobic degradation of swine
manure and that these populations exhibited suitable levels
of activity. In addition, the performance results indicate that
anaerobic digestion, at least for the loading rate of 1 g VS/L/
day, was not affected by the high shear conditions in the
membrane recirculation loop (the average cross flow velocity
was 1.6 m/s).
3.1.2. Day 53–90
The loading rate of the system was doubled to 2 g VS/L/day on
day 53 and a two-fold increase in biogas production was
observed soon thereafter. The biogas production and the
methane content in the biogas remained high for a period of
21 days (up to day 75) (Fig. 2a) following the increase in
loading rate. The SCOD removal was more than 96% during
this 21-day period and the reactor VFA concentrations were
below 1000 mg/L as acetic acid, indicating successful treat-
ment of swine manure. These results indicate that anaerobic
digestion of swine manure in an AnMBR remained feasible for
a loading rate of 2 g VS/L/day and shear conditions in the
membrane recirculation loop corresponding to an average
cross flow velocity of 1.1 m/s.
The stator in the progressing cavity pump (P1 in Fig. 1) was
replaced on day 75 due to declining pump performance. As a
result of the stator change, the cross flow velocity in the
membrane recirculation loop increased from approximately
0.9–1.9 m/s. The system performance began to deteriorate
soon after these changes. VFAs accumulated, the a value
increased, and the SCOD in the reactor and the effluent
increased to over 4000 mg/L (Fig. 2). The data in Fig. 2(d) show
that the decrease in performance was accompanied by
increases in reactor acetate, propionate, and butyrate con-
centrations. VFA concentrations can increase in response to
high NH3–N concentrations. However, the NH3–N concentra-
tion was measured to be below 400 mg-N/L, which is well
below the inhibitory level of 1.1 g NH3–N/L reported by Hansen
et al. (1998).
3.1.3. Day 91–115
On day 91, the loading rate was reduced to 1 g VS/L/day since
the performance continued to deteriorate (Fig. 2). To increase
the buffering capacity of the system, 13.1 g of NaHCO3 were
41 (20 07) 13 4 – 144 139
added to the system on day 95. Since the addition of NaHCO3
did not improve the performance, feeding was terminated on
day 98 and a second equivalent dose of NaHCO3 was added on
day 101. The concentration of total sodium added to the
reactor following the second dose was calculated to be
approximately 1030 mg/L, which is well below reported
inhibitory levels of 3500–5500 mg/L (McCarty, 1964).
Following the second addition of NaHCO3, the reactor
started recovering. The a value decreased to below 1.0 and
the pH increased to above 7.0 (Fig. 2c). The methane content
in the biogas also increased and the SCOD and total VFAs in
the reactor decreased. Acetate, propionate, and butyrate
concentrations decreased from day 103 to 114, but the
decrease in propionate concentrations was considerably
slower (approximately 34% reduction from day 103 to 110)
compared to reductions in acetate (approximately 81%
reduction from day 103 to 110) and butyrate (100% reduction
from day 103 to 110) concentrations.
3.1.4. Day 116–300
After the system recovered (i.e., returned to low a and VFA
values), the feeding was resumed on day 116 at a loading rate
of 1 g VS/L/day. On day 142, the loading rate was increased to
2 g VS/L/day and on day 186–3 g VS/L/day. We observed an
increase in VFA concentrations approximately 2 weeks after
each increase in loading rate (Fig. 2c). However, by reducing
the loading rate for a short period of time, the VFA levels
returned to low values. The reactor performance was rela-
tively stable for the remainder of the operational period at a
loading rate of 3 g VS/L/day. However, continuous operation
beyond day 300 at this loading rate of 3 g VS/L/day proved
difficult (data not reported) and it did not seem possible to
increase the loading rate beyond 3 g VS/L/day without affect-
ing the performance of the AnMBR. This could be because of
the high C:N ratio present in the raw swine manure. Similar
problems were encountered in pilot plant AnMBRs treating
swine manure for which the loading rate could not be
increased beyond 2.4 g VS/L/day (unpublished data).
3.2. Membrane performance
Membrane performance was characterized using membrane
flux and total membrane resistance (Fig. 3). The target TMP
for membrane operation was 0.4 bar, but during reactor
operation measured TMPs varied between 0.3 and 0.7 bar
during the first 135 days. Membrane flux using a clean
membrane rapidly declined from a clean water flux of 100 L/
m2/h to values in the range of 5–10 L/m2/h with a correspond-
ing increase in total membrane resistance. Details of mem-
brane fouling and cleaning have been described elsewhere
(Zhang et al., 2007).
During reactor operation, the stator of the progressing cavity
pump in the membrane recirculation loop (Fig. 1) was
gradually wearing out and needed to be replaced on days 22,
75, 209, and 269. The relatively fast deterioration of the stator
could be due to abrasion caused by sand particles found in the
swine waste. The initial stator change, with resulting rapid
increase in the cross flow velocity from 0.9 to 1.9 m/s, did not
have any effect on the total membrane resistance. The second
stator was replaced on day 75, which resulted in a significant

140
2.5
2.0
Cross flow velocity (m/s) 1.5
1.0
0.5
0.0
0 25
Fig. 3 – Total resistance across the membrane, the permeate flux, and cross flow velocity of the membrane over time. The
stator of the progressing cavity pump was changed on days 22, 75, 209, and 269 (Zhang et al., 2007).
reduction in total resistance (Fig. 3). This reduction in total
membrane resistance can be explained by the removal of a
fouling layer, which had accumulated during the first 75 days
of operation and was removed by the sudden increase in shear.
This decrease in total membrane resistance was followed by a
steady increase in total membrane resistance during subse-
quent operation. For subsequent stator changes (days 209 and
269), the flow velocity was gradually increased after each stator
change to avoid negative impacts on biological conversion
processes as observed with the second stator change.
The lack of influence on membrane resistance due to the
initial stator change on day 22 can be explained by the shorter
operational period (there was less build up of fouling layer)
and the low solids content in the reactor due to the low
loading rate: approximately 20 g SS/L on day 22 compared to
40 g SS/L on day 75. The initial membrane was used for over
100 days without any form of cleaning and was replaced on
day 134 for fouling analysis. The satisfactory long-term
membrane performance suggests that maintaining high
cross-flow velocities was critical to reduce the rate of
membrane fouling. However, sudden changes in shear can
have a negative effect on biological processes in the AnMBR,
although it had a positive effect on membrane performance.
The membrane flux of the second membrane was not
measured due to problems related to partial clogging of the
flow meter by microbial growth in the effluent tubing
resulting in unreliable data. Microbial growth in the effluent
tubing was not surprising since the effluent SCOD during this
period averaged 1272 mg/L.
3.3. Microbial community analysis
3.3.1. Archaeal community dynamics with T-RFLP
Using T-RFLP analysis, four archaeal populations belonging to
the orders Methanomicrobiales and Methanosarcinales were
found to be abundant in the AnMBR throughout the 300-day
operational period. A fifth group, belonging to the family of
ARTICLE IN PRESS
WA T E R R E S E A R C H 41 (2007) 134– 144
Permeate flux (L /m /h) and Total resistance (1012/m)
150
Cross flow velocity
Permeate flux
Total resistance 120
90
60
2
30
0
50 75 100 125 150 175 200 225 250 275 300
Days
Methanobacteriales, was only present at substantial levels
during the first 60 days of the reactor run (Fig. 4, Table 2).
At the start of operation, Methanosaeta spp. (T-RF MseI 103),
which have a high affinity for acetate (i.e., able to grow under
low acetate concentrations) (Jetten et al., 1992), were abun-
dant (Fig. 4). Methanosaeta spp. were present in the reactor at
startup due to their presence in the inoclum, which consisted
of 1/3 (v/v) of anaerobic granules. Methanosaeta spp. have been
found to act as nuclei for granulation in UASB reactors (Zheng
et al., 2006) and anaerobic migrating blanket reactor systems
(Angenent et al., 2004). However, the granules used to
inoculate the AnMBR in the current study were crushed at
the time of inoculation and conditions in the AnMBR did not
promote granule formation. Thus, the increase in the levels of
Methanosaeta spp. during the initial operational period was
likely due to the low acetate concentration during this period
(Fig. 2d). With the increase in loading rate (day 53), the acetate
concentration in the reactor began to increase and the
abundance of Methanosaeta spp. decreased (Fig. 4). Following
the increase in cross flow velocity through the membrane
module (day 75, Fig. 3), the Methanosaeta spp. remained low,
but did not decrease further (Fig. 4), suggesting that the
increase in shear did not contribute to the reduction in
Methanosaeta spp. levels. Consistent with these observations,
the abundance of Methanosarcina spp. (T-RFs AluI 474 and MseI
474), which are able to proliferate under high acetate
concentrations (Jetten et al., 1992), increased from day 93
onwards; the acetate concentration in the reactor was at its
maximum on day 92. The Methanosarcina spp. continued to
dominate in the reactor for the rest of the operational period.
Methanogens belonging to two hydrogen utilizing metha-
nogenic groups (Methanobacteriales and Methanomicrobiales)
were present in the AnMBR. The levels of Methanobacteriales
were high soon after startup, but decreased to below 5% after
approximately 30 days of operation. Methanomicrobiales be-
came the dominant hydrogen utilizers after the levels of
Methanobacteriales had decreased and their relative abundance
 ARTICLE IN PRESS
WAT E R R E S E A R C H 41 (20 07) 13 4 – 144 141

40
Methanobacteriales - AluI 341 (a)
35 Methanomicrobiales - AluI 434

Relative abundance (% area)

30

25

20

15

10

0
Methanosaetaceae - MseI 103 (b)
35 Methanosarcinaceae - MseI 474
Methanosarcinaceae - AluI 474
Relative abundance (% area)

30

25

20

15

10

0
0 25 50 75 100 125 150 175 200 225 250 275 300
Days

Fig. 4 – Methanogenic population dynamics in the AnMBR monitored with T-RFLP. (a) Hydrogenotrophic methanogens, (b)
aceticlastic methanogens.

Table 2 – Groups of Archaea identified in the AnMBR using T-RFLP following PCR with A109f-A912r primers and digestion
with restriction enzymes AluI, MaeI, and MseI

Archaea population present Terminal restriction fragment (T-RF) Specific genera identified Primary
in the AnMBR and the T-RF size (bp) by combining information substrate
used for relative obtained from digestions
abundance calculations with three restriction
enzymes (number of
species in each genus)
AluI MaeI MseI

Methanosaetaceae - MseI 103 108 169 103a Methanosaeta (7), Methanolobus Acetate
(1)
Methanosarcinaceae - MseI 474 108 169 474a Methanomethylovorans (4), Acetate, C-1
uncultured Methanosarcina (4), compounds
Thermoplasma volcanium (2),
uncultured Methanosaeta sp. (1)
Methanobacteriales – AluI 341 341 400 103 Methanobrevibacter (79) Hydrogen
Methanomicrobiales – AluI 434 434 502 420 Methanospirillum (4), Hydrogen
Methanocalculus (2)
Methanosarcinaceae - AluI 474 474 169 773a Methanosarcina (20), Acetate, C-1
Methanolobus (2), compounds
Methanococcoides (3)

a
Theses fragment sizes were not included in Table 1 since the specificity and coverage are very low when considered by themselves. However,
when the use of MseI was combined with information obtained by using AluI and MaeI, specific groups of species were identified.
 ARTICLE IN PRESS
142 WA T E R R E S E A R C H
remained above 10% throughout the operational period. With
the increase in loading rate and increase in shear in the
system, the abundance of the Methanomicrobiales population
increased (days 59–95). During this time period, the reactor
performance deteriorated with increased VFA concentrations
and decreased methane production. The increase in the
abundance of the Methanomicrobiales population suggests that
this population became more active due to an increase in
hydrogen production during this period. Although hydrogen
levels were not be measured in the biogas, the observed
increase in propionate concentration after the increase in
loading rate on day 52 (Fig. 2d) is consistent with an increase
in hydrogen production. With the change in stator on day 75,
the propionate concentration rose even further. The concur-
rent increases in propionate and hydrogen production could
be the result of a faster hydrolysis rate due to increased shear
conditions, leading to higher production of fermentation
products, including hydrogen, as well as an increased
production of hydrogen by syntrophic bacteria through
conversion of other fermentation products. This would
indicate that the increase in shear was not sufficient to break
up the syntrophic interactions between the syntrophic
bacteria and their methanogenic partners, as suggested by
Brockmann and Seyfried (1996). However, the syntrophic
bacteria were not able to convert fermentation products at
the rate they were produced during this time period, which led
to the acidification of the reactor. Reactor performance
recovered after addition of NaHCO3 and termination of
feeding for 2 weeks. The abundance of the Methanomicrobiales
population decreased with the termination of feeding (days
98–115), and subsequently increased when feeding was
resumed. It should be noted that some strains of Methano-
sarcina also can utilize hydrogen. However, the importance of
hydrogen as a growth substrate for Methanosarcina spp. in
complex microbial communities has not been studied exten-
sively. A few studies have suggested that Methanosarcina spp.
are not able to compete for hydrogen with other hydrogen
utilizing methanogens in most anaerobic digesters (Ahring
et al., 1991; Zinder, 1993).
The reactor VFA concentration again increased each time
the loading rate was increased. The first incident occurred
between days 200 and 205 (Fig. 2c and 2d). During this short
time period and a few days thereafter, an increase in the
levels of Methanosarcina spp. was observed (Fig. 4b), which
Table 3 – Percent variance associated with different T-RFLP steps in AnMBR sample collected on day 291 using restriction
enzyme AluI, and coefficient of variation (CV) for relative peak area and absolute peak area
Fragment size % of total variance in absolute peak area associated with different
(bp) T-RFLP steps
DNA PCR Digestion
Extraction (+Desalting)
104 0.0 0.0 88.8
355 20.0 0.0 79.6
432 17.1 0.0 68.0
472 0.0 0.0 89.2
41 (2007) 134– 144
could be the result of the increase in acetate concentration.
With decreasing acetate concentrations, the Methanosarcina
spp. relative abundance also decreased until day 244. VFAs
accumulated again between days 246 and 253, which again
resulted in an increase in the relative abundance of Methano-
sarcina spp. Their relative abundance remained high for the
rest of the operational period (except for days 291 and 295)
consistent with the relatively high VFA concentration (except
for days 274–291) observed in the reactor. As the relative
abundance of Methanosarcina spp. decreased between days
205 and 244, a corresponding increase in the abundance of
the Methanomicrobiales population was observed (Fig. 4a
and b). These changes could be the result of high turnover
of propionate leading to increased hydrogen production in the
reactor, and a corresponding increase in the levels of
hydrogen utilizers. It should be noted that the observed
changes in population levels may be the result of changes in
the levels of populations not targeted by the T-RFLP analysis
since relative population levels, rather than absolute popula-
tion levels, were determined.
3.3.2. Precision and reproducibility associated with T-RFLP for
AnMBR samples
T-RFLP has become a powerful tool for the rapid analysis of
microbial community structure in multiple samples. It is
often considered a semi-quantitative technique due to biases
associated with DNA extraction and PCR amplification.
However, the output can be used to evaluate microbial
diversity and, in combination with sequence information, to
monitor changes in the abundance of specific populations.
The experiment conducted to gain confidence in the data
consisted of a variance component analysis performed on four
dominant peak areas (peak area 410%) observed in
the AnMBR sample of day 291 with restriction enzyme AluI
(Table 3). Peaks 104 and 472 had relative peak areas of
approximately 25%, whereas peaks 355 and 432 exhibited
relative peak areas of approximately 10%. The variance
component analysis indicated that most of the variance
(70–90%) was associated with the enzyme digestion step, which
also includes the de-salting step. The coefficient of variation
(CV) involved when considering the absolute peak area was also
high (434%, Table 3). However, when relative peak area (i.e.,
relative abundance) was considered, less variation (CV o20%)
was observed among the samples (Table 3).
CV (%) for CV (%) for
absolute relative
peak area peak area
Capillary
electrophoresis
11.2 34 20
0.4 45 16
14.8 53 16
10.8 37 16
 ARTICLE IN PRESS
WAT E R R E S E A R C H
The reported relative population levels (Fig. 4) are based on
single analyses. However, the variances in peak area asso-
ciated with different T-RFLP steps indicate that a single
sample analysis is sufficient to represent a particular data
point. The low CV of relative peak area also supports this
observation. In addition, a single analysis allowed us to
process a time series of data in a fast and efficient manner.
Osborn et al. (2000) also presented an evaluation of T-RFLP
profiling using environmental DNA samples isolated directly
from polluted and pristine soils. Some of the factors
investigated included the use of replicates, pooling of
extracted DNA samples, choice of polymerase, PCR annealing
temperature, and dilution effects. They obtained a coefficient
of variation less than 11% in the reproducibility of peak
heights in three replicates (the de-salting step was not
included in their protocol). These results obtained demon-
strate that, once standardized, T-RFLP is a robust and
reproducible methodology for the rapid analysis of microbial
community structure in different samples.
4. Conclusions
 Stable anaerobic digestion of swine manure can be
achieved in an AnMBR with external membrane filtration
using cross flow velocities of up to 2 m/s.
 An increase in cross flow velocity was found to benefit
membrane performance, but resulted in poor anaerobic
digestion performance.
 T-RFLP results indicated that hydrogen utilizing methano-
gens increased in abundance during a period of poor
reactor performance, caused by a sudden increase in shear
conditions.
 The decrease in reactor performance likely can be
explained by an increase in the rate of hydrolysis caused
by an increase in shear leading to a buildup of fermenta-
tion products. It is unlikely that the decrease in reactor
performance was due to breakup of microbial aggregates
and interruption of syntrophic interactions.
Acknowledgements
The authors would like to thank Aurelio Briones for helpful
discussions and Mike Tenuto, Andrew White, and Tara
Jackson for their help with sampling. This work was partially
supported by The Water CAMPWS, a Science and Technology
Center of Advanced Materials for the Purification of Water
with Systems under the National Science Foundation agree-
ment number CTS-0120978.
R E F E R E N C E S
Ahring, B.K., Westerman, P., Mah, R., 1991. Hydrogen inhibition of
acetate metabolism and kinetics of hydrogen consumption by
Methanosarcina thermophilia TM-1. Arch. Microbiol. 157,
38–42.
Anderson, G.K., Yang, G., 1992. Determination of bicarbonate and
total volatile acid concentration in anaerobic digesters using a
simple titration. Water Environ. Res. 64, 53–59.
41 (20 07) 13 4 – 144 143
Angenent, L.T., Sung, S., Raskin, L., 2004. Formation of granules
and Methanosaeta fibres in an anaerobic migrating blanket
reactor (AMBR). Environ. Microbiol. 6 (4), 315–322.
APHA, WEF, AWWA, 1998. Standard methods for the examination
of water and wastewater. In: Clesceri, L. S., Greenberg, A. E.,
Eaton, A.D. (Eds.), 20th ed. American Public Health Associa-
tion, Washington, DC.
Brockmann, M., Seyfried, C.F., 1996. Sludge activity and cross-flow
microfiltration–A non beneficial relationship. Water Sci.
Technol. 34 (9), 205–213.
Choo, K.-H., Lee, C.-H., 1996. Membrane fouling mechanisms in
the membrane-coupled anaerobic bioreactor. Water Res. 30 (8),
1771–1780.
Choo, K.H., Kang, I.J., Yoon, S.H., Park, H., Kim, J.H., Adiya, S., Lee,
C.H., 2000. Approaches to membrane fouling control in
anaerobic membrane bioreactors. Water Sci. Technol. 41,
363–371.
du Preez, J., Norddahl, B., Christensen, K., 2005. The BIOREKs
concept: a hybrid membrane bioreactor concept for very
strong wastewater. Desalination 183 (1–3), 407–415.
Elmaleh, S., Abdelmoumni, L., 1997. Cross-flow filtration of an
anaerobic methanogenic suspension. J. Membrane Sci. 131,
261–274.
Hansen, K.H., Angelidaki, I., Ahring, B.K., 1998. Anaerobic diges-
tion of swine manure: inhibition by ammonia. Water Res. 32
(1), 5–12.
He, Y., Xu, P., Li, C., Zhang, B., 2005. High-concentration food
wastewater treatment by an anaerobic membrane bioreactor.
Water Res. 39, 4110–4118.
Jetten, M.S.M., Stams, A.J.M., Zehnder, A.J.B., 1992. Methanogen-
esis from acetate—a comparison of the acetate metabolism in
Methanothrix soehngenii and Methanosarcina spp. FEMS Micro-
biol. Rev. 88, 181–197.
Kim, M., Ahn, Y.H., Speece, R.E., 2002. Comparative process
stability and efficiency of anaerobic digestion; mesophilic vs.
thermophilic. Water Res. 36 (17), 4369–4385.
Klein, A.N., 2003. MS Thesis, University of Illinois, Urbana-
Champaign, USA.
Lee, S.M., Jung, J.Y., Chung, Y.C., 2001. Novel method for
enhancing permeate flux of submerged membrane system in
two-phase anaerobic reactor. Water Res. 35 (2), 471–477.
Lueders, T., Friedrich, M., 2000. Archaeal population dynamics
during sequential reduction processes in rice field soil. Appl.
Environ. Microbiol. 66, 2732–2742.
McCarty, P.L., 1964. Anaerobic waste treatment fundamentals.
Part III, toxic materials and their control. Public Works 95
(November), 91–94.
McMahon, K.D., Stroot, P.G., Mackie, R.I., Raskin, L., 2001.
Anaerobic codigestion of municipal solid waste and biosolids
under various mixing conditions—II: microbial population
dynamics. Water Res. 35 (7), 1817–1827.
McMahon, K.D., Zheng, D., Stams, A.J., Mackie, R.I., Raskin, L.,
2004. Microbial population dynamics during start-up and
overload conditions of anaerobic digesters treating municipal
solid waste and sewage sludge. Biotechnol. Bioeng. 87,
823–834.
Osborn, A.M., Moore, E.R.B., Timmis, K.N., 2000. An evaluation of
terminal-restriction fragment length polymorphism (T-RFLP)
analysis for the study of microbial community structure and
dynamics. Environ. Microbiol. 2, 39–50.
Poggi-Varaldo, H.M., Oleszkiewicz, J.A., 1992. Anaerobic cocom-
posting of municipal solid-waste and waste sludge at high
total solids levels. Environ. Technol. 13, 409–421.
Ripley, L.E., Boyle, W.C., Converse, J.C., 1986. Improved alkali-
metric monitoring for anaerobic-digestion of high-strength
wastes. J. Water Pollut. Control Fed. 58 (5), 406–411.
Stephenson, K.B., Jud, S., Jefferson, B., 2000. Membrane Bioreac-
tors for Wastewater Treatment. IWA Publishing, London.
 ARTICLE IN PRESS
144 WA T E R R E S E A R C H
Stroot, P.G., McMahon, K.D., Mackie, R.I., Raskin, L., 2001.
Anaerobic codigestion of municipal solid waste and biosolids
under various mixing conditions—I. Digester performance.
Water Res. 35 (7), 1804–1816.
Zhang, J., Padmasiri, S.I., Fitch, M., Norddahl, B., Raskin, L.,
Morgenroth, E., 2007. Influence of cleaning frequency and
membrane history on fouling in an anaerobic membrane
bioreactor. Desalination 207, 153–166.
41 (2007) 134– 144
Zheng, D., Angenent, L.T., Raskin, L., 2006. Monitoring granule
formation in anaerobic upflow bioreactors using oligonu-
cleotide hybridization probes. Biotechnol. Bioeng. 94 (3),
458–472.
Zinder, S.H., 1993. Physiological ecology of methanogens. In:
Ferry, J.G. (Ed.), Methanogenesis, Ecology, Physiology, Bio-
chemistry & Genetics. Chapman & Hall, Inc., London, pp.
128–206.