Bioresource Technology 267 (2018) 650–656

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Mixed culture fermentation of synthesis gas in the microfiltration and T

ultrafiltration hollow-fiber membrane biofilm reactors
Hua-Jie Wanga,c,1, Kun Daib,1, Yun-Qi Wanga, Hou-Feng Wangb, Fang Zhangb,
⁎
Raymond Jianxiong Zenga,b,
a
CAS Key Laboratory for Urban Pollutant Conversion, Department of Chemistry, University of Science and Technology of China, Hefei 230026, PR China
b
Centre of Wastewater Resource Recovery, College of Resources and Environment, Fujian Agriculture and Forestry University, Fuzhou, Fujian 350002, PR China
c
School of Environmental and Chemical Engineering, Anhui Vocational and Technical College, Hefei, Anhui 230011, PR China

G R A P H I C A L A B S T R A C T

A R T I C LE I N FO A B S T R A C T

Keywords: The effects of pore sizes on the in-situ utilization of synthesis gas (syngas, H2 and CO) mixed culture fermen-
Synthesis gas tation (MCF) in the hollow-fiber membrane biofilm reactor (HfMBR) are not clear. Thus, the ultrafiltration (R1)
Mixed culture fermentation and microfiltration (R2) HfMBRs were constructed. Syngas was totally consumed within the formed biofilm in
Hollow fiber membrane biofilm reactor R1; contrarily, it accumulated notably in R2. In the batch mode of R1 and R2, volatile fatty acids (VFAs) of
(HfMBR)
acetate, butyrate and caproate were the main metabolites, but the production rate of total VFA in R1
Microfiltration
(61.9 mmol-C/(L·d)) was higher than that of R2 (27.6 mmol-C/(L·d)). In the continuous mode, the R1 perfor-
Ultrafiltration
Volatile fatty acids mance was much better than that of R2, and the biofilm in R2 was even washed out. Furthermore, Clostridium
(30.0%) was the main genus in the enriched biofilm of R1, which converted syngas to VFAs. Thus, the ultra-
filtration membrane shall be the suitable candidate for syngas MCF.

1. Introduction wastes of biomass and sludge are gaining more and more attentions in
the last decades (He et al., 2014; Kan et al., 2016). But, the direct
Exploring the environmental friendly technologies to produce bio- conversion of organic wastes by the biological processes is difficult and
chemicals and biofuels, such as acetate and ethanol, from the organic a significant amount of non-biodegradable material remains in the

⁎
Corresponding author at: CAS Key Laboratory for Urban Pollutant Conversion, Department of Chemistry, University of Science and Technology of China, Hefei
230026, PR China.
E-mail address: rzeng@ustc.edu.cn (R.J. Zeng).
1
These authors contributed equally to this work.

https://doi.org/10.1016/j.biortech.2018.07.098
Received 16 March 2018; Received in revised form 18 July 2018; Accepted 19 July 2018
Available online 21 July 2018
0960-8524/ © 2018 Elsevier Ltd. All rights reserved.
H.-J. Wang et al.
effluent. For example, most of the degradable cellulose and hemi-
cellulose in the biomass are packed with lignin that are resistant to
microbial degradation (Abubackar et al., 2011). And several methods
including acid-based methods and hydrothermal processing are pro-
posed to remove lignin and hemicelluloses (Dai et al., 2018; Jönsson
and Martín, 2016). Gasification to synthesis gas (syngas, mainly CO and
H2) provides an alternative technology for the utilization of these re-
calcitrant organic wastes (Latif et al., 2014). The produced syngas could
be furthermore converted to volatile fatty acids (VFAs, including
acetate, butyrate and caproate, etc.) and biofuels (such as ethanol) by
mixed culture fermentation (MCF) (Ge et al., 2015; He et al., 2018; Latif
et al., 2014).
The poor aqueous solubility of H2 and CO is one major limiting
factor in syngas fermentation (Esquivel-Elizondo et al., 2017; Lee et al.,
2016). High impeller speed is commonly suggested to provide high gas/
liquid mass transfer coefficient, but it can also lead to high-power
consumption and may inhibit bacterial activity (Henstra et al., 2007;
Zhao et al., 2014). Recently, He et al. (2018) used porous sponge
scouring pad to promote CO transferring to the liquid and biofilm for-
mation in the reactor, and the maximal concentrations of caproate,
heptylate, and caprylate were 0.22, 0.21, and 0.14 g/L, respectively.
Increasing the specific gas-liquid interfacial area can diminish poor gas
solubility. In the hollow-fiber membrane biofilm reactor (HfMBR), H2
permeates from inside of the membrane lumen and is directly con-
sumed by biofilms naturally attached on the outer surface of the
hollow-fiber membrane (Zhang et al., 2013b). Zhang et al. demon-
strated the in-situ H2 consumption (100%) in a mesophilic HfMBR, in
which the concentrations of caproate (0.98 g/L) and caprylate (0.42 g/
L) were higher (Zhang et al., 2013b). Several researchers also show that
the microporous or nonporous membranes are useful for reduction or
oxidation NO3−, ClO4− and other contaminants with H2, O2 and CH4-
based membrane biofilm reactors (Luo et al., 2015; Xie et al., 2017).
Meanwhile, the membrane reactor also presents other advantages, such
as small reactor footprint and easy scale-up (Martin and Nerenberg,
2012). However, research on the biochemicals production from syngas
in HfMBR is rarely reported.
On the other hand, microfiltration (the pore size between 0.1 and
10 μm) and ultrafiltration (the pore size between 0.01 and 0.1 μm)
membranes are the common materials in membrane bioreactors for
wastewater treatment (Holloway et al., 2015; Peter-Varbanets et al.,
2009). Till now, most of the studies focus on the microfiltration
membrane with the pore size of 0.1–0.2 μm in HfMBRs (Lai et al., 2016;
Wang et al., 2017). Several workers found that membrane pore size
could affect the gas utilization, for example, poor CO solubility in the
aqueous phase occurred as the reactors using large pore sizes (20 μm)
such as column diffuser and sparger (Munasinghe and Khanal, 2010). In
bacteria-free reactors, Orgill et al. (2013) used O2 as the gaseous mass
transfer agent and found that the nonporous polydimethylsiloxane
hollow-fiber membrane provided the highest volumetric mass transfer
coefficient (1062 1/h), followed by the trickle-bed reactor (421 1/h)
and stirred tank reactor (114 1/h). Yasin et al. (2014) determined the
CO mass transfer using hollow fiber membrane with pore size of 0.1 μm
and found that mass transfer coefficient (kLa) increased from 63.7 1/h
to 135.7 1/h as the inlet pressure increased from 37.2 kPa and 93.8 kPa.
Consequently, the pore size in MF and UF membranes shall be an
important factor for biofilm attachment and metabolites production in
syngas fermentation. In our former works, the membrane pore size was
mainly 0.01 μm in HfMBR (Wang et al., 2017; Wang et al., 2018b). But,
to the best of our knowledge, such studies of higher pore sizes (above
0.01 μm) are rarely reported on syngas (CO and H2) fermentation in
HfMBR. Thus, the aims in this work were to 1) construct the hollow-
fiber membranes (MF of 8–10 μm and UF of 0.02–0.05 μm) biofilm re-
actors and compare the metabolites production; 2) reveal the dominant
bacteria in HfMBR biofilm by the Illumina MiSeq high-throughput se-
quencing. Consequently, it is expected that the outcomes would benefit
to choose the suitable membrane in HfMBR for syngas MCF in future.
651
Bioresource Technology 267 (2018) 650–656
2. Materials and methods
2.1. HfMBR setup and operational conditions
Two HfMBRs were configured with the ultrafiltration (R1, pore size
0.02–0.05 μm) and microfiltration (R2, pore size 8–10 μm) hollow-fiber
membranes (PureSea Spring Membrane Technology Co. Ltd, China),
respectively, in which the working volume and total membrane surface
area were same, and were 420 mL and 0.016 m2, respectively. Syngas of
60% H2 and 40% CO were fed into HfMBR. The inlet gas pressures of R1
and R2 were manually adjusted by the regulator and were monitored by
a gas-pressure gauge (Hua Yitong Electromechanical Equipment Co.
Ltd, Nanjing, China). Recirculation pump provided an inner loop within
the flow of 60 L/h. pH was controlled at 6 ± 0.1 by a pH controller
with 1 M NaOH and temperature was controlled at 35 ± 1 °C by a
water bath. The medium was fed to the reactor with a peristaltic pump
(Longer pump, Baoding, China).
Two HfMBRs were inoculated with anaerobic sludge (60 mL) col-
lected from an anaerobic digester, which was the same as that of Wang
et al. (2017). The reactors were initially supplemented with 10 mmol/L
Bromoethane sulfonate (BES) to inhibit methanogenesis. The whole
experimental processes of R1 and R2 were divided into batch and
continuous modes. R1 and R2 were initially run in the batch mode, in
which 90% medium was removed from the reactor at day 53, and the
fresh medium without BES was added. At day 90, the continuous mode
began, the medium with no BES added was fed to the reactors. The
reactor performance of R1 was investigated under different HRTs (5.5,
3.3, and 1 days), in which 4 cycles were operated for each experiment.
Whereas, the biofilm in R2 was washed out under the HRT of 5.5 days,
and the experiment was stopped at day 107. The composition of
medium was the same as that of Zhang et al. (2013b).
2.2. Chemical analysis and scanning electron microscopy (SEM) image of
biofilm
The gas and liquid samples were determined every day. The con-
centrations of VFAs, ethanol and butanol were measured using a gas
chromatograph (Agilent 7890, CA). The samples were filtered with
0.45 mm microfiltration membrane and then acidified with 3% (v/v)
formic acid before analysis. The contents of H2, CO2, CO, and CH4 in the
headspace were analyzed using a gas chromatograph (Lunan model
SP7890, CN). The biofilms of R1 (day 153) and R2 (day 108) after the
continuous experiments were analyzed by SEM (SIRION 200, FEI, USA)
and the details of the method were described by Zhang et al. (2013b).
2.3. Data analysis
Statistical analysis was carried out with the SPSS 20.0 package. The
bivariate correlation method was used to check the influences of mi-
crofiltration and ultrafiltration membrane on VFAs production. The
one-way ANOVA method was used to check the effect of HRT on the CO
consumption rate in R1.
2.4. DNA extraction and high throughout sequencing
The biofilms of R1 (day 153) and R2 (day 108) after the continuous
experiments were first detached from the outer surface of the hollow-
fiber membrane and then collected in phosphate-buffered saline (PBS).
DNA samples of inoculum, R1 biofilm and R2 biofilm were then ex-
tracted from PBS using the PowerSoil DNA Isolation Kit (MO BIO, USA).
The primers used to target the V3-V4 regions of both bacterial and
archaeal 16S rRNA genes were identified as 341F and 806R (Sundberg
et al., 2013). The entire sequencing process was performed at the No-
vogene Institute (Beijing, China). The sequencing data of inoculum, R1
biofilm and R2 biofilm in the current study were archived in NCBI
Sequence Read Archive with accession numbers of SRR6370522,
H.-J. Wang et al.
Fig. 1. Performances of ultrafiltration (R1) and microfiltration (R2) HfMBRs in
batch mode, (A) R1, (B) R2. Notes: ↘ indicates the liquid discharge and fresh
medium addition in R1 and R2.
SRR6370523, and SRR6370524, respectively.
3. Results and discussion
3.1. Performances of ultrafiltration and microfiltration HfMBRs in the
batch mode
Without biofilm formed, the H2 flow rates and KLa of ultrafiltration
and microfiltration membranes were initially determined according to
Yasin et al. (2014). The MF membrane offered higher gas flow rates, for
example, under the inlet pressure of 0.01 MPa, the flow rates of H2 in
R1 was only 3.75 L/(h*m2), but, it increased to 187.5 L/(h*m2) in R2.
Whereas, the UF membrane offered higher KLa, for example, under the
inlet pressure of 0.01 MPa, the flow rates of H2 in R1 was 83.1 1/h, but,
it decreased to 70.7 1/h in R2.
After R1 and R2 were inoculated with anaerobic sludge, Fig. 1
shows the evolution of acetate (C2), butyrate (C4), caproate (C6), while
other products such as caprylate, ethanol and butanol were not de-
tected. The inlet pressure was controlled at 0.01–0.015 MPa in R1 to
ensure no syngas detected in the headspace. And in R2, the inlet
pressure was always below 0.005 MPa since the biofilm was not formed
well. The inlet CO and H2 contents were 40% and 60%, respectively; the
CO and H2 contents in the headspace of R1 were lower than 0.5%, but
these values were around 30% and 50% respectively in R2 since the
biofilm was not formed well. Thus, the gas consumption percentage
reached to 99% in R1, but, it was rather low in R2.
As the common metabolite of CO and H2 via the Wood-Ljungdahl
652
Bioresource Technology 267 (2018) 650–656
pathway by acetogens, acetate accumulated notably in both R1 and R2.
The maximum concentration of acetate in phase 1 was 19.6 (R1) and
29.4 (R2) g/L, respectively. Butyrate and caproate were also detected in
R1, the max concentrations were 4.6 and 2.9 g/L, respectively, whereas,
only butyrate was detected in R2 and the final concentration was just
0.6 g/L. The production rate of total VFAs according to the C numbers
of each metabolites (Zhang et al., 2013b) in R1 (61.9 mmol-C/(L·d))
was higher than that of R2 (27.6 mmol-C/(L·d)). Because the gas flow
rates of R2 was higher than that of R1 and CO toxicity was considered
to be the resistance for its utilization (Esquivel-Elizondo et al., 2017),
the hollow-fiber membrane with 0.02–0.05 μm pore size was con-
sidered to be a better membrane than the microfiltration membrane
(8–10 μm) to in-situ consume CO in HfMBRs.
However, at the end of phase 1, the metabolites of syngas did not
increase notably that was due to the inhibition of high concentration
metabolites, especially acetate. According to our recent work (Zhang
et al., 2018), the inhibition on hydrogenotrophic methanogenesis
reached 52% when the acetate concentration was 4.5 g/L and pH was
6.0 in a HfMBR. Niel et al. (2003) observed that when the acetate
concentration was over 21.9 g/L at pH 7.0, it started to inhibit substrate
degradation of Caldicellulosiruptor saccharolyticus. Zhang et al. also
found that glucose could not be consumed by mixed culture when
acetate concentration was 34 g/L at pH 7.5 (Zhang et al., 2014). And, in
this work, the acetate concentration reached to 29 g/L, the inhibition
could reach 80% according to the equation of Zhang et al. (2018). Thus,
the high VFAs concentration was a main inhibition factor in HfMBRs.
Therefore, 90% of the broth was replaced with the new medium at day
53.
After replacing with the fresh medium, the metabolites increased
again (Fig. 1B), which indicated that the acetogens biofilms were
formed on the outer surface of hollow-fiber membranes in R1 and R2. In
R1, the butyrate concentration increased faster than that of acetate and
caproate, other metabolites were still not detected. The max con-
centration of acetate was 10.2 g/L, which was lower than that in phase
1. Contrarily, the max concentrations of butyrate and caproate were
higher than that in phase 1, and were 13.5 and 3.5 g/L, respectively. In
phase 2 of R2, the final acetate concentration was only 15.0 g/L, but the
butyrate concentration reached 3.2 g/L and caproate was also detected
with the max concentration of 1.6 g/L. In phase 2, no CO and H2 were
detected in R1, however, these two gases were all detected in R2 and
the partial pressures were higher than 0.005 MPa. Since CO and H2
emission from the reactor may result in the toxicity to human and even
lead to explosion, the microfiltration membrane with 8–10 μm pore size
was not suitable for syngas utilization in HfMBR.
On the other hand, the correlation analysis method was used to
reveal the membrane types on metabolites production. In phase 1, pore
size had a direct positive effect on the acetate concentration and their
correlation coefficient was 0.915 (p < 0.01); and there was a negative
correlation between the pore size and the butyrate concentration, and
their correlation coefficient was −0.965 (p < 0.01). After the medium
was replaced in phase 2, there was a negative correlation between the
acetate concentration and the butyrate concentration, and their corre-
lation coefficient was −0.792 (p < 0.01). The butyrate concentration
had a direct positive effect on the caproate concentration and their
correlation coefficient was 0.954 (p < 0.01), because the butyrate
production rate was higher than the consumption rate for caproate
production. Thus, the correlation analysis method could not reveal the
suitable membrane materials for HfMBR in this work.
3.2. Performances of ultrafiltration and microfiltration HfMBRs in the
continuous mode
R1 and R2 were changed to the continuous mode at Day 90, and
consequently the concentrations of acetate, butyrate and caproate all
decreased notably within the initial days. The metabolite evolution is
shown in Fig. 2. HRT of R1 was sequentially set at 5.5, 3.3 and 1 days,
H.-J. Wang et al.
Fig. 2. Performances of ultrafiltration (R1, A) and microfiltration (R2, B)
HfMBRs in the continuous mode.
whereas, R2 was stopped at day 107 because the biofilm was notably
detached and washed out. The duplicated experiments in R2 also ver-
ified this result. And the OD600 values in the bulk liquid of R1 and R2
under HRT of 5.5 days was determined and these values in R1 did not
change much (around 0.15), on the contrary, it decreased from 0.46 to
0.05 in R2.
During the continuous experiments, CO and H2 was always un-
detected in R1, but both of them accumulated in R2 and even reached
40% and 60% after day 105, which also demonstrated that the micro-
filtration membrane with pore size of 8–10 μm was not suitable for
syngas fermentation. Acetate and butyrate were the main metabolites of
syngas fermentation, whereas, caproate, ethanol and butanol were just
detected under HRT 5.5 days. On the other hand, because BES was not
added, methane was detected at the end of experiments and the content
was around 10% in R1.
At the end of each HRT experiment, the production rates of acetate,
butyrate and caproate are summarized in Table 1. Under HRT 5.5 days,
the total metabolites production rate of R1 (38.7 mmol-C/(L·d)) was
higher than that of R2 (9.7 mmol-C/(L·d)). In R1, the production rates
of caproate, ethanol and butanol under HRT 5.5 days were 0.15, 0.02
and 0.03 g/(L·d), respectively. However, as HRT was reduced to
3.3 days, only acetate and butyrate were detected in R1, and their
production rates were 0.3 and 0.1 g/(L·d), respectively. The total me-
tabolites production rates of R1 under HRT 3.3 and 1.0 days were si-
milar and were 14.5 and 14.6 mmol-C/(L·d), respectively, which were
lower than that of HRT 5.5 days. On the other hand, the ANOVA results
demonstrated that HRT caused statistical significant differences
(p < 0.05) on the CO consumption rate (Fig. 3), in which the pairwise
comparison of the CO consumption rate showed significant differences
(p < 0.05) between HRT of 5.5 days and other two HRTs (3.3 and
1 days); but no significant differences (p > 0.05) between HRT of
653
Bioresource Technology 267 (2018) 650–656
3.3 days and 1 day (Fig. 3).
3.3. The microbial community analysis
The ultrafiltration and microfiltration hollow-fiber membranes in
R1 and R2 changed to black and yellow after the continuous experi-
ments, which suggested that biofilms were attached on the outer sur-
face of membranes. The SEM results confirmed that the main micro-
organisms were mainly rod and cocci shaped in R1 and R2.
The Illumina Miseq high-throughput sequencing technology was
used to reveal the inoculum and biofilms in R1 and R2, and the main
indices are summarized in Table 2. Over 63,000, 66,000 and 68,000
effective sequences were obtained for three samples, respectively. Be-
cause the 341F-806R primer was used in this work, the mean sequen-
cing length was round 400 bp. After the whole experiments, the OTU
number of biofilm in R1 and R2 was 373 and 526, respectively, which
was lower than that of the initial inoculum (787). Meanwhile, when the
sequence number of each sample was greater than about 8000, Shannon
diversity for the isolated microorganism did not increase clearly. The
Alpha diversity indices including Shannon, ACE, and Chao1, of R1 and
R2 biofilms were all lower than that of the inoculum, which indicated
that the enrichment microorganisms in the biofilm changed notably. On
the other hand, the coverages of three samples were all approaching to
1.0 (Table 2). Therefore, the sequencing depth in this work was enough
to analyze the whole community.
The microbial communities of the inoculum and enriched biofilms
of R1 are shown in Fig. 4, which demonstrated that the microbial
community changed notably in the enriched biofilm in HfMBR. As
shown in Fig. 4A, at the phylum level, Proteobacteria (29.4%), Eur-
yarchaeota (26.6%) and Firmicutes (12.5%) were the main phyla in the
seed sludge. However, after the enrichment, the main phyla shifted to
Firmicutes (67.6%) and Euryarchaeota (15.8%) in R1, and to Firmicutes
(57.1%) and Euryarchaeota (18.3%) in R2. Therefore, after the en-
richment, Firmicutes were dominant bacteria in HfMBR of R1 and R2.
Moreover, at the genus level (Fig. 4B), the dominant genera were
Methanosaeta (23.3%) and Syntrophobacter (13.6%) in the inoculum
while Clostridium (30.0%) and Syntrophomonas (20.0%) were the main
genera in enriched biofilms of R1. As known, Clostridium such as C.
ljungdahlii and C. drakei, the well-known genus that can grow under pH
from 4.0 to 7.0, is the spore-forming rod, and produces acetate, buty-
rate and caproate from CO and H2, etc. (Liou et al., 2005; Tanner et al.,
1993; Zhang et al., 2013a). Therefore, changes of microbial community
supported the HfMBR performances for the production of acetate, bu-
tyrate and caproate from CO and H2. And the notably different domi-
nant bacteria also might cause the long enrichment time (10 days) as
shown in Fig. 1. On the other hand, Syntrophomonas is the common
genus adapted for syntrophic growth with methanogens to degrade
VFAs such as acetate and butyrate to methane (Sieber et al., 2010). But,
the methane production was not notable in this work, thus the function
of Syntrophomonas was still not clear.
However, the main genus in R2 was Oscillospira, and its percentage
was only 12.9%. Oscillospira was reported to be curved rods shaped and
similar to the closest neighbor of strain Clostridium (Iino et al., 2007; Ye
et al., 2016), which was consistent with the SEM results. The percen-
tage of Clostridium was only 3.2%. The washed out of biofilm in R2
might result in notable differences of sequencing results between R1
and R2, since the same syngas and membrane material were utilized in
this work. Consequently, based on the results in this work and our
former works of HfMBRs (Wang et al., 2018a; Wang et al., 2017; Wang
et al., 2018b; Zhang et al., 2013b; Zhang et al., 2013a), the dominant
bacteria mainly related to temperature, not the pore size, thus, the
enriched bacteria in UF and MF packed HfMBRs were similar at 35 °C,
i.e. Clostridia. Whereas, Wang et al. (2017) found that dominant bac-
teria were Thermoanaerobacteria at 55 °C.
H.-J. Wang et al. Bioresource Technology 267 (2018) 650–656

Table 1
Summary of metabolites production rates in the continuous mode of R1 and R2.
metabolites Unit HRT-5.5 days HRT-3.3 days HRT-1 day

R1 R2 R1 R2 R1 R2

Acetate g/(L·d) 0.6 ± 0.03 0.1 ± 0.02 0.3 ± 0.04 – 0.3 ± 0.03 –
mmol-C/(L·d) 20 ± 1.0 4.3 ± 0.57 10 ± 1.3 – 10 ± 1.0 –
Butyrate g/(L·d) 0.2 ± 0.02 0.05 ± 0.003 0.1 ± 0.004 – 0.1 ± 0.01 –
mmol-C/(L·d) 9.1 ± 0.9 2.3 ± 0.13 4.5 ± 0.2 – 4.6 ± 0.5 –
Caproate g/(L·d) 0.14 ± 0.01 0.06 ± 0.006 – – – –
mmol-C/(L·d) 7.2 ± 0.52 2.8 ± 0.3 – – – –
Ethanol g/(L·d) 0.02 ± 0.001 – – – – –
mmol-C/(L·d) 0.9 ± 0.04 – – – – –
Butanol g/(L·d) 0.02 ± 0.003 – – – – –
mmol-C/(L·d) 1.5 ± 0.16 – – – – –
Total g/(L·d) 1.0 ± 0.04 0.2 ± 0.03 0.4 ± 0.04 – 0.4 ± 0.04 –
mmol-C/(L·d) 38.7 ± 1.62 9.7 ± 1.0 14.5 ± 1.5 – 14.6 ± 1.5 –

Note: “–” means that the metabolites were undetected.

“ ± ” means standard error, and n = 3.

notably detached and the reactor was even broken down in the con-
tinuous mode in this work. Till now, the pore sizes of the mostly used
hollow-fiber membranes for methane and H2 utilization were
0.1–0.2 μm (Lai et al., 2016; Wang et al., 2017). And the microporous
and nonporous membranes are also suggested for reduction or oxida-
tion NO3−, ClO4− and other contaminants (Xie et al., 2017). In HfMBR,
the sufficient microorganisms attached on the membrane surface are
necessary for the efficient and stable operation. Thus, membranes with
the low pore size (< 0.2 μm) is more suitable for syngas fermentation in
HfMBR. On the other hand, it shall be noted that membranes can also
provide a better gas/liquid mass transfer making smaller gas bubble
(i.e. acting as a diffuser). For example, Munasinghe et al. demonstrated
that reactors having larger pore sizes such as column diffuser and
sparger only showed a limited CO solubility in the aqueous phase
(Munasinghe and Khanal, 2010). Yasin et al. (2014) found that mass
transfer coefficient (kLa) increased from 63.7 1/h to 135.7 1/h as the
inlet pressure increased from 37.2 kPa and 93.8 kPa. Thus, in this work
we found the membrane pore size of 0.02–0.05 μm was better for
HfMBR because it could in-situ consume syngas. But, higher pore size
may also be better for gas delivery and biofilm resistance in membrane
Fig. 3. Effect of HRT on the CO consumption rate in the continuous mode of R1.
(The CO consumption rate is assumed to be equal to the sum of metabolites
reactors.
production rates in Table 2.) Lastly, since the metabolites in syngas MCF are a mixture, the fol-
lowing processes are also necessary to recover and concentrate the
metabolites from the fermentation broth. Electrodialysis (ED) is a sui-
3.4. Perspectives of HfMBR for syngas fermentation
table technology for acids recovery, however, the broth in syngas MCF
also includes several inorganic salts that decrease the real separating
The hollow-fiber membrane provides an alternation to enhance the
factors and increase the energy consumption. So coupling syngas fer-
transfer of poor solubility gas such as CO and H2 in syngas fermentation
mentation with ED is also needed furthermore analysis. Recently, Khor
process (Lai et al., 2016; Wang et al., 2017; Zhang et al., 2013a; Zhang
et al. proposed the conversion of medium chain fatty acids to decane via
et al., 2013b). However, the membrane parameters such as pore size
Kolbe electrolysis, which is an important chemicals and biofuel (Khor
that could affect the biofilm formation and gas transfer are not de-
et al., 2017). And the removal of metabolites is also benefit for in-
monstrated in detail. In this work, two different hollow-fiber mem-
hibiting the methanogenesis. Additionally, in this work, we also found
branes of ultrafiltration (0.02–0.05 μm) and microfiltration (8–10 μm)
that the CO2 notably accumulated due to the low content of H2 (60%).
were used for syngas fermentation in HfMBR. We found that in ultra-
To reduce CO2 emission, extra H2 should be added in the inlet syngas.
filtration HfMBR, the dominant metabolites included acetate, butyrate
While other technologies, such as microbial electrosynthesis cells that
and caproate, their concentrations were 25.5 g/L, 13.5 g/L and 3.5 g/L,
could convert CO2 to methane, acetate and other biochemicals, are also
respectively, which were higher than that of Zhang et al. (acetate 7.4 g/
the potential coupling technologies (Schroder et al., 2015). Thus,
L, butyrate 1.8 g/L and caproate 0.98 g/L) (Zhang et al., 2013b).
syngas MCF and the coupling with other technologies seem more
On the other hand, the biofilm in microfiltration HfMBR was

Table 2
The main indices of the seed sludge and biofilms in R1 and R2 by Illumina high-throughput sequencing technology.
Sample name Effective Sequence number Average length (bp) OTU number Shannon ACE Chao1 Coverage

Inoculum 63,326 414 787 6.146 816.5 814.4 0.999

R1 biofilm 66,777 410 373 4.054 417.6 421.8 0.999
R2 biofilm 68,009 408 526 5.465 560.6 555.7 0.999

654
H.-J. Wang et al.
Fig. 4. Microbial communities of the inoculum and biofilms of R1 and R2 in HfMBR at the phylum level (A) and the genus level (B) Note: Phylogenetic groups
accounting for ≤2% are classified in the artificial group “others”.
appropriate from an economic perspective.
4. Conclusions
The ultrafiltration (R1) and microfiltration (R2) HfMBRs were
constructed to convert syngas (H2/CO) to acetate, butyrate and
caproate. It was found that R1 could totally consume syngas within the
formed biofilm, contrarily, syngas was accumulated notably in R2. In
the batch mode, the production rate of total VFA in R1 (61.9 mmol-C/
(L·d)) was higher than that of R2 (27.6 mmol-C/(L·d)). In the con-
tinuous mode, the R1 performance was better than that of R2, and the
biofilm in R2 was even washed out. Microbial analysis showed that
Clostridium (30.0%) was the main genus in the biofilm of R1.
Acknowledgments
The authors would like to acknowledge the financial support from
National Natural Science Foundation of China (51478447) and Central
Guidance on the Development of Local Science and Technology (Fujian,
2017L3003).
655
Bioresource Technology 267 (2018) 650–656
Appendix A. Supplementary data
Supplementary data for this work including the additional experi-
ments of determining gas flow rates and gas to liquid factor (KLa),
metabolites production in R2 and effluent OD600 in R1 and R2, and SEM
pictures can be found in the e-version of this paper online.
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.biortech.2018.07.098.
References
Abubackar, H.N., Veiga, M.C., Kennes, C., 2011. Biological conversion of carbon mon-
oxide: rich syngas or waste gases to bioethanol. Biofuel Bioprod. Bior. 5 (1), 93–114.
Dai, K., Zhang, F., Zhang, Y., Zeng, R.J., 2018. The chemostat metabolite spectra of al-
kaline mixed culture fermentation under mesophilic, thermophilic, and extreme-
thermophilic conditions. Bioresour. Technol. 249, 322–327.
Esquivel-Elizondo, S., Delgado, A.G., Rittmann, B.E., Krajmalnik-Brown, R., 2017. The
effects of CO2 and H2 on CO metabolism by pure and mixed microbial cultures.
Biotechnol. Biofuels 10 (1), 220.
Ge, S., Usack, J.G., Spirito, C.M., Angenent, L.T., 2015. Long-term n-caproic acid pro-
duction from yeast-fermentation beer in an anaerobic bioreactor with continuous
product extraction. Environ. Sci. Technol. 49 (13), 8012–8021.
He, C., Chen, C.L., Giannis, A., Yang, Y.H., Wang, J.Y., 2014. Hydrothermal gasification of
sewage sludge and model compounds for renewable hydrogen production: a review.
Renew. Sust. Energy Rev. 39, 1127–1142.
He, P., Han, W., Shao, L., Lü, F., 2018. One-step production of C6–C8 carboxylates by
H.-J. Wang et al.
mixed culture solely grown on CO. Biotechnol. Biofuels 11 (1), 4.
Henstra, A.M., Sipma, J., Rinzema, A., Stams, A.J.M., 2007. Microbiology of synthesis gas
fermentation for biofuel production. Curr. Opin. Biotechnol. 18 (3), 200–206.
Holloway, R.W., Wait, A.S., Fernandes da Silva, A., Herron, J., Schutter, M.D., Lampi, K.,
Cath, T.Y., 2015. Long-term pilot scale investigation of novel hybrid ultrafiltration-
osmotic membrane bioreactors. Desalination 363, 64–74.
Iino, T., Mori, K., Tanaka, K., Suzuki, K.-I., Harayama, S., 2007. Oscillibacter valericigenes
gen. nov., sp. nov., a valerate-producing anaerobic bacterium isolated from the ali-
mentary canal of a Japanese corbicula clam. Int. J. Syst. Evol. Micr. 57 (8),
1840–1845.
Jönsson, L.J., Martín, C., 2016. Pretreatment of lignocellulose: Formation of inhibitory
by-products and strategies for minimizing their effects. Bioresour. Technol. 199,
103–112.
Kan, T., Strezov, V., Evans, T.J., 2016. Lignocellulosic biomass pyrolysis: A review of
product properties and effects of pyrolysis parameters. Renew. Sust. Energy Rev. 57,
1126–1140.
Khor, W.C., Andersen, S., Vervaeren, H., Rabaey, K., 2017. Electricity-assisted production
of caproic acid from grass. Biotechnol. Biofuels 10 (1), 180.
Lai, C.-Y., Wen, L.-L., Shi, L.-D., Zhao, K.-K., Wang, Y.-Q., Yang, X., Rittmann, B.E., Zhou,
C., Tang, Y., Zheng, P., Zhao, H.-P., 2016. Selenate and nitrate bioreductions using
methane as the electron donor in a membrane biofilm reactor. Environ. Sci. Technol.
50 (18), 10179–10186.
Latif, H., Zeidan, A.A., Nielsen, A.T., Zengler, K., 2014. Trash to treasure: production of
biofuels and commodity chemicals via syngas fermenting microorganisms. Curr.
Opin. Biotechnol. 27, 79–87.
Lee, J., Jang, N., Yasin, M., Lee, E.Y., Chang, I.S., Kim, C., 2016. Enhanced mass transfer
rate of methane via hollow fiber membrane modules for Methylosinus trichosporium
OB3b fermentation. J. Ind. Eng. Chem. 39, 149–152.
Liou, J.S.-C., Balkwill, D.L., Drake, G.R., Tanner, R.S., 2005. Clostridium carboxidivorans
sp. nov., a solvent-producing clostridium isolated from an agricultural settling la-
goon, and reclassification of the acetogen Clostridium scatologenes strain SL1 as
Clostridium drakei sp. nov. Int. J. Syst. Evol. Micr. 55 (5), 2085–2091.
Luo, Y.-H., Chen, R., Wen, L.-L., Meng, F., Zhang, Y., Lai, C.-Y., Rittmann, B.E., Zhao, H.-
P., Zheng, P., 2015. Complete perchlorate reduction using methane as the sole
electron donor and carbon source. Environ. Sci. Technol. 49 (4), 2341–2349.
Martin, K.J., Nerenberg, R., 2012. The membrane biofilm reactor (MBfR) for water and
wastewater treatment: principles, applications, and recent developments. Bioresour.
Technol. 122, 83–94.
Munasinghe, P.C., Khanal, S.K., 2010. Syngas fermentation to biofuel: evaluation of
carbon monoxide mass transfer coefficient (kLa) in different reactor configurations.
Biotechnol. Prog. 26 (6), 1616–1621.
Niel, E.W.J.V., Claassen, P.A.M., Stams, A.J.M., 2003. Substrate and product inhibition of
hydrogen production by the extreme thermophile, Caldicellulosiruptor saccharolyticus.
Biotechnol. Bioeng. 81 (3), 255–262.
Orgill, J.J., Atiyeh, H.K., Devarapalli, M., Phillips, J.R., Lewis, R.S., Huhnke, R.L., 2013. A
comparison of mass transfer coefficients between trickle-bed, hollow fiber membrane
and stirred tank reactors. Bioresour. Technol. 133, 340–346.
Peter-Varbanets, M., Zurbrügg, C., Swartz, C., Pronk, W., 2009. Decentralized systems for
potable water and the potential of membrane technology. Water Res. 43 (2),
656
Bioresource Technology 267 (2018) 650–656
245–265.
Schroder, U., Harnisch, F., Angenent, L.T., 2015. Microbial electrochemistry and tech-
nology: terminology and classification. Energy Environ. Sci. 8 (2), 513–519.
Sieber, J.R., Sims, D.R., Han, C., Kim, E., Lykidis, A., Lapidus, A.L., McDonnald, E.,
Rohlin, L., Culley, D.E., Gunsalus, R., McInerney, M.J., 2010. The genome of
Syntrophomonas wolfei: new insights into syntrophic metabolism and biohydrogen
production. Environ. Microbiol. 12 (8), 2289–2301.
Sundberg, C., Al-Soud, W.A., Larsson, M., Alm, E., Yekta, S.S., Svensson, B.H., Sørensen,
S.J., Karlsson, A., 2013. 454 pyrosequencing analyses of bacterial and archaeal
richness in 21 full-scale biogas digesters. FEMS Microbiol. Ecol. 85 (3), 612–626.
Tanner, R.S., Miller, L.M., Yang, D., 1993. Clostridium ljungdahlii sp. nov., an acetogenic
species in clostridial rRNA homology group I. Int. J. Syst. Bacteriol. 43 (2), 232–236.
Wang, H.-J., Dai, K., Xia, X.-Y., Wang, Y.-Q., Zeng, R.J., Zhang, F., 2018a. Tunable pro-
duction of ethanol and acetate from synthesis gas by mesophilic mixed culture fer-
mentation in a hollow fiber membrane biofilm reactor. J. Clean. Prod. 187, 165–170.
Wang, Y.-Q., Yu, S.-J., Zhang, F., Xia, X.-Y., Zeng, R.J., 2017. Enhancement of acetate
productivity in a thermophilic (55 °C) hollow-fiber membrane biofilm reactor with
mixed culture syngas (H2/CO2) fermentation. Appl. Microbiol. Biotechnol. 101 (6),
2619–2627.
Wang, Y.-Q., Zhang, F., Zhang, W., Dai, K., Wang, H.-J., Li, X., Zeng, R.J., 2018b.
Hydrogen and carbon dioxide mixed culture fermentation in a hollow-fiber mem-
brane biofilm reactor at 25°C. Bioresour. Technol. 249, 659–665.
Xie, G.-J., Cai, C., Hu, S., Yuan, Z., 2017. Complete nitrogen removal from synthetic
anaerobic sludge digestion liquor through integrating anammox and denitrifying
anaerobic methane oxidation in a membrane biofilm reactor. Environ. Sci. Technol.
51 (2), 819–827.
Yasin, M., Park, S., Jeong, Y., Lee, E.Y., Lee, J., Chang, I.S., 2014. Effect of internal
pressure and gas/liquid interface area on the CO mass transfer coefficient using
hollow fibre membranes as a high mass transfer gas diffusing system for microbial
syngas fermentation. Bioresour. Technol. 169, 637–643.
Ye, T., Cai, H., Liu, X., Jiang, H.-L., 2016. Dominance of Oscillospira and Bacteroides in
the bacterial community associated with the degradation of high-concentration di-
methyl sulfide under iron-reducing condition. Ann. Microbiol. 66 (3), 1199–1206.
Zhang, F., Ding, J., Shen, N., Zhang, Y., Ding, Z.-W., Dai, K., Zeng, R.J., 2013a. In situ
hydrogen utilization for high fraction acetate production in mixed culture hollow-
fiber membrane biofilm reactor. Appl. Microbiol. Biotechnol. 97 (23), 10233–10240.
Zhang, F., Ding, J., Zhang, Y., Chen, M., Ding, Z.-W., van Loosdrecht, M.C.M., Zeng, R.J.,
2013b. Fatty acids production from hydrogen and carbon dioxide by mixed culture in
the membrane biofilm reactor. Water Res. 47 (16), 6122–6129.
Zhang, F., Zhang, Y., Ding, J., Dai, K., van Loosdrecht, M.C.M., Zeng, R.J., 2014. Stable
acetate production in extreme-thermophilic (70°C) mixed culture fermentation by
selective enrichment of hydrogenotrophic methanogens. Sci. Rep. UK 4, 5268.
Zhang, W., Dai, K., Xia, X.-Y., Wang, H.-J., Chen, Y., Lu, Y.-Z., Zhang, F., Zeng, R.J., 2018.
Free acetic acid as the key factor for the inhibition of hydrogenotrophic methano-
genesis in mesophilic mixed culture fermentation. Bioresour. Technol. 264, 17–23.
Zhao, H.-P., Ontiveros-Valencia, A., Tang, Y., Kim, B.-O., VanGinkel, S., Friese, D.,
Overstreet, R., Smith, J., Evans, P., Krajmalnik-Brown, R., Rittmann, B., 2014.
Removal of multiple electron acceptors by pilot-scale, two-stage membrane biofilm
reactors. Water Res. 54, 115–122.