Bioresource Technology 124 (2012) 68–76

Contents lists available at SciVerse ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Syntrophic interactions drive the hydrogen production from glucoseat low temperature in microbial electrolysis cells

Lu Lu a, Defeng Xing a,⇑, Nanqi Ren a, Bruce E. Logan a,b

a State Key Laboratory of Urban Water Resource and Environment, School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, Chinab Department of Civil and Environmental Engineering, 212 Sackett Building, The Pennsylvania State University, University Park, PA 16802, USA

h i g h l i g h t s

" H2 production from glucose at 4 �C in MECs overcomes the dark-fermentation bottleneck." H2 yield at 4 �C is comparable with that obtained at mesophilic temperatures in MECs." Combining pyrosequencing with CV reveal the syntrophic interactions in MECs at 4 �C." Psychrotolerant fermenters and exoelectrogens allowed current generation from glucose." Methanogenesis and homoacetogenesis were negligible in glucose-fed MECs at 4 �C.

a r t i c l e i n f o

Article history:Received 5 May 2012Received in revised form 9 August 2012Accepted 10 August 2012Available online 19 August 2012

Keywords:Hydrogen productionMicrobial electrolysis cell (MEC)Low temperatureSyntrophic interactionPyrosequencing

0960-8524/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.biortech.2012.08.040

⇑ Corresponding author. Address: School of Municineering, Harbin Institute of Technology, P.O. Box 2614District, Harbin, Heilongjiang Province 150090, China

E-mail addresses: dxing@hit.edu.cn (D. Xing), rnq@

a b s t r a c t

H2 can be obtained from glucose by fermentation at mesophilic temperatures, but here we demonstratethat hydrogen can also be obtained from glucose at low temperatures using microbial electrolysis cells(MECs). H2 was produced from glucose at 4 �C in single-chamber MECs at a yield of about 6 mol H2 mol�1

glucose, and at rates of 0.25 ± 0.03–0.37 ± 0.04 m3 H2 m�3 d�1. Pyrosequencing of 16S rRNA gene andelectrochemical analyses showed that syntrophic interactions combining glucose fermentation withthe oxidization of fermentation products by exoelectrogens was the predominant pathway for currentproduction at a low temperature other than direct glucose oxidization by exoelectrogens. Another syn-trophic interaction, methanogenesis and homoacetogenesis, which have been found in 25 �C reactors,were not detected in MECs at 4 �C. These results demonstrate the feasibility of H2 production from abun-dant biomass of carbohydrates at low temperature in MECs.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Microbial electrolysis cells (MECs) are a new method for elec-trochemically producing hydrogen using current generated by exo-electrogenic microorganisms. However, most MEC studies haveexamined systems at ambient temperatures using acetate as thefuel. It was recently shown that H2 gas could also be produced inan MEC at relatively low temperatures (e.g. 4 and 9 �C) using ace-tate, making this technology a promising method for biohydrogenproduction even in very cold climates (Lu et al., 2011). However, itis important to consider the utilization of fuels other than acetateat low temperatures because most biomass available for biofuelsproduction is primarily stored as fermentable carbohydrates suchas glucose and cellulose. In previous bioelectrochemical system

ll rights reserved.

pal and Environmental Engi-, 73 Huanghe Road, Nangang

. Tel./Fax: +86 451 86289195.hit.edu.cn (N. Ren).

(BES) studies, including both MECs and microbial fuel cells (MFCs),there have been large differences in substrate metabolism andreactor performance in mesophilic environments using ferment-able carbohydrates, such as glucose, compared to studies usingacetate (Freguia et al., 2008; Lee et al., 2008). It is therefore impor-tant to better understand how fermentable substrates are de-graded by microorganisms in MECs at low temperatures.

Glucose is a simple carbohydrate that can be converted to elec-trical current in BESs (Rabaey et al., 2003; Selembo et al., 2009). Afew exoelectrogenic bacteria can directly oxidize glucose (e.g.Rhodoferax ferrireducens, Klebsiella pneumoniae, and Aeromonashydrophila) and transfer electrons to electrodes (Logan, 2009). Inmixed-culture systems, previous studies imply that syntrophicinteraction between fermenters and exoelectrogens is the majorroute to metabolize glucose for current production under meso-philic conditions (Freguia et al., 2008; Zhang et al., 2011). Glucoseis first oxidized to organic acids or H2 by fermentation, followed byconsumption of fermentation products by the exoelectrogens,which eliminates feedback inhibition of glucose fermentation

L. Lu et al. / Bioresource Technology 124 (2012) 68–76 69

(Freguia et al., 2008). However, fermentation leading to currentgeneration is not invariably the dominant reaction in glucose-fedanode biofilms using mixed cultures, as some researchers have alsofound that Fe (III)-reducing bacteria, related to the Aeromonasgenus, that could directly oxidize glucose were dominant in anodebiofilms and therefore played an important role in generating cur-rent (Chung and Okahe, 2009; Park et al., 2008). Another importantsyntrophic process in glucose-fed BESs generally exists betweenfermenters and methanogens (Freguia et al., 2008) due to the com-petition of methanogens with exoelectrogens for fermentationproducts (acetate and formate or H2) under anaerobic conditions.Community profiles from fermentable substrate-fed MFCs alsosuggest a third possible syntrophic interaction. Homoacetogenicbacteria can channel electrons from H2 (produced by fermentation)to acetate that is re-oxidized by the exoelectrogens to produceelectricity (Parameswaran et al., 2011). In MECs, the H2 consumedby homoacetogens can come from the cathode.

All of these syntrophic interactions in BESs have previouslybeen studied at mesophilic temperatures, and there have been nostudies of microbial communities in systems fed carbohydratesat low temperatures (below 10 �C). The few studies conducted atlow temperatures (4–15 �C) with MFCs have mainly been focusedon a non-fermentable substrate (acetate) (Cheng et al., 2011; Patilet al., 2010). Since most fermentative bacteria and methanogensare generally considered to be inactive under psychrotolerant con-ditions (Scherer and Neuhaus, 2006), the dominant pathway ofcurrent production at low temperatures has yet to be established.It is not known whether current is produced by direct glucose oxi-dization by exoelectrogens, or through syntrophic process betweenfermentative and electrochemically active microorganisms. It maybe that there are different exoelectrogenic microorganisms thatfunction at lower, compared to higher, temperatures.

In this study, glucose was used as an electron donor in MECs inorder to determine if (i) H2 could be successfully produced from afermentable substrate at a low temperature (4 �C) and (ii) if theprimary pathway of glucose conversion to current was direct orindirectly through fermentation processes. To accomplish thesegoals, we examined the formation of the intermediates duringelectrohydrogenesis of glucose in concert with hydrogen gas pro-duction. Cyclic voltammetry (CV) was employed to evaluate theelectrochemical features of the anode biofilms, allowing identifica-tion of the midpoint potential and inference of possible extracellu-lar electron transfer mechanisms. Dominant microbial populationsin the biofilms of the MECs under psychrotolerant conditions wereinvestigated using two independent molecular biology methods:(i) The highly parallel 454 GS-FLX pyrosequencing based on 16SrRNA gene with 100 times the throughput of a traditional Sangersequencing used in low throughput methods (e.g. clone librariesand denaturing gradient gel electrophoresis [DGGE]) and (ii) thetranditional 16S rRNA gene clone libraries.

2. Methods

2.1. MEC construction and operation

Single-chamber, membraneless MECs (liquid volume 26 mL)were constructed as previously described (Lu et al., 2009) withgraphite brush anodes (2.5 cm diameter � 2.5 cm length; fibertype: T700–12 K, Toray, Inc.) and carbon cloth air cathodes(7 cm2, type B, E-TEK, Inc.) containing 0.5 mg cm�2 Pt catalyst.MEC anodes were initially enriched in similarly constructed micro-bial fuel cells (MFCs) with cathodes being exposed to air. MFC reac-tors were frequently inoculated (intervals of 2–3 days) with a50:50 mixture of the two effluents from a room temperature glu-cose-fed MFC reactor (23 ± 2 �C, operating about 6 months) and

an acetate-fed psychrotolerant MEC (Lu et al., 2011) (operatingmore than 12 months at 4 �C), with 11.1 mM (2 g L�1) glucose asthe fuel. Acclimation was conducted in fed-batch mode at 4 �C(three reactors) except two MFCs were enriched at 25 �C (as con-trols) for community structure comparison to psychrotolerantelectroactive biofims. Anodes were considered fully enriched whena reproducible maximum voltage (over 1 kX resistor) was ob-tained, and then they were transfer to MECs for H2 production.

All MECs were fed 11.1 mM (2 g L�1) glucose in a 50 mM nutri-ent phosphate buffer solution (NPBS) (Na2HPO4, 4.58 g L�1; NaH2-

PO4�H2O, 2.45 g L�1; NH4Cl, 0.31 g L�1; KCl, 0.13 g L�1; NH4Cl,0.31 g L�1; trace minerals and vitamins) and operated at 4 �C ex-cept control reactors (25 �C). Prior to being fed into reactors, thesolution was sparged with nitrogen gas (99.999%) for 15 min to re-move oxygen, then a fixed voltage of 0.6 V or 0.8 V was applied tothe MECs by a programmable power source (3645A, Array, Inc.).

Current was determined by measuring the voltage over a high-precision resistor (10 X) using a multimeter/data acquisition sys-tem (model 2700 with 7702 module, Keithley, Inc.) at 10 min inter-vals. A reference electrode (Ag/AgCl, 0.197 V vs. SHE; RE-5B, BASi,Inc.) was inserted into the chamber to measure the anode andcathode potentials. When the current decreased below 0.3 mA,the reactors were refilled with fresh medium. MECs were operatedcontinuously over a period of about two months.

2.2. Analyses and calculations

The gas produced by the MECs was collected in a gas bag (0.1 LCali-5-Bond, Calibrate, Inc.), and the total volume was measuredusing a glass syringe. Gas composition was analyzed using a gaschromatograph (4890D, Agilent, Inc.). Liquid samples from MECswere immediately filtered through 0.22 lm pore-diameter cellu-lose acetate filters, and analyzed for volatile fatty acids (VFAs), eth-anol and chemical oxygen demand (COD). The concentrations ofVFAs (including acetic, propionic, butyric and valeric acid) and eth-anol were analyzed using another gas chromatograph (7890A, Agi-lent, Inc.). The COD was measured according to standard methods(APHA, 1998).

All electrochemical experiments were conducted using a multi-channel potentiostat (WMPG-1000S, WonATech Co., Ltd). If notstated otherwise, all potentials provided refer to a Ag/AgCl refer-ence electrode (0.197 V vs. a standard hydrogen electrode, SHE).Cyclic voltammetry (CV) was conducted at a scan rate of 1 or5 mV s�1, in the potential range from �0.8 to 0.2 V using a three-electrode arrangement in anaerobic conditions. Linear sweep vol-tammetry (LSV) was performed to determine the power densities(normalized to cathode surface area) of MFCs from �0.8 to�0.05 V using a scan rate of 0.1 mV s�1.Hydrogen production rateQ (m3 H2 m�3 day�1), volumetric current density IV (A m�3), hydro-gen yield YH2 (mol H2 mol�1 glucose) and Coulombic efficiency (CE)calculated on the basis of COD, cathodic hydrogen recovery (rcat),overall hydrogen recovery (RH2 ¼ CErcat) and energy recovery rela-tive to the electrical input (gE) were used to evaluate the perfor-mance of MECs according to previously described (Logan et al.,2008).

2.3. Bacterial community analysis

At the end of MECs operation, the graphite fibers were cut fromthe anodes of psychrotolerant MECs and those in controls (25 �Creactor), and were fragmented using sterile scissors. Graphite fi-bers were sampled equably from three different sections of eachbrush and were combined for DNA extraction. Total genomicDNA was extracted using a PowerSoil DNA Isolation Kit (MoBioLaboratories, Inc., Carlsbad, CA) according to the manufacturer’sinstructions. The bacterial 16S rRNA-gene clone libraries were con-

Fig. 1. Current densities of MECs operated at 4 �C using glucose at applied voltagesof 0.6 V and 0.8 V. Inset: Anode and cathode potential over two batch-cycles atdifferent applied voltages.

70 L. Lu et al. / Bioresource Technology 124 (2012) 68–76

structed according to our previous work (Lu et al., 2011). The se-quences were deposited to the GenBank with accession numbersJN680051�JN680078.

The 16S rRNA gene for 454 pyrosequencing was amplified byPCR using a 10-nucleotide barcoded forward primer 8F (50-AGA-GTTTGATCCTGGCTCAG-30) and the reverse primer 533R (50-TTACCGCGGCTGCTGGCAC-30). After being purified and quantified,a mixture of amplicons was used for pyrosequencing on a Rochemassively parallel 454 GS-FLX according to standard protocols.Raw pyrosequencing data that obtained from this study weredeposited to the NCBI Sequence Read Archive (SRA, http://trace.nc-bi.nlm.nih.gov/Traces/sra/sra.cgi?) with accession No. SRA047640.Low-quality sequences were removed by eliminating those thatdid not match the primer sequence, lengths shorter than 200nucleotides and those that contained any ambiguous base calls(Ns). Pyrosequencing produced 6160 and 5478 high-quality V1–V3 tags of the 16S rRNA gene for psychrotolerant MECs and thereactor of control, respectively, with an average read length ofabout 430 bp. Operational taxonomic units (OTUs), rarefactioncurve, species richness estimators (Chao1 and the abundance-based coverage estimator ACE), and Shannon diversity index weregenerated using MOTHUR program (http://www.mothur.org/wiki/Main_Page). Sequences were phylogenetically assigned to phylum,class, and genus taxonomic groups using a RDP naïve BayesianrRNA classifier with a confidence threshold of 80% (http://rdp.cme.msu.edu/classifier/classifier.jsp). Hierarchical cluster anal-ysis was performed using gplots package of R (http://www.r-pro-ject.org/) in Linux.

3. Results and discussion

3.1. Enrichment of anodes at a low temperature

About two months after start-up (inoculation was stopped afterone month), all MFCs used for enrichment of anodes at 4 �C dem-onstrated an average reproducible maximum voltage of494 ± 4 mV (external resistance of 1 kX; three reactors). This accli-mation time is longer than that needed for enrichment of acetate-fed MFCs at 4 �C (27–30 days) (Lu et al., 2011) despite the use ofthe same start-up means based on inoculating reactors frequentlywith pre-acclimatized bacteria from other MFCs. Previous studieshave shown that glucose fermentation provided the substratesused by exoelectrogens at mesophilic temperatures. However,the growth of fermentative bacteria would be expected to be veryslow at low temperatures. The reactors operated at 25 �C (control)required only 16 days (inoculation was stopped after 6 days) toreach a maximum voltage (564 ± 3 mV), which is close to the timeneeded for start-up of acetate-fed MFC at the same temperature(Lu et al., 2011). The mature anode biofilms produced maximumpower densities of 554 ± 3 mW m�2 (4 �C) and 982 ± 11 mW m�2

(25 �C).

3.2. Hydrogen production from glucose at a low temperature

When anodes from the MFCs were switched to MECs, there wasimmediate current generation and H2 production at 4 �C (Fig. 1).Low temperature MECs exhibited consistent and reproducible cy-cles of current generation with an average current density of38 ± 5 A m�3 at an applied voltage (Eap) of 0.6 V. When the Eap

was set at 0.8 V, the current density increased as expected(50 ± 7 A m�3). All of these values were ca. 56�67% less than theaverage current densities produced at the mesophilic temperature.The H2 production rate was also reduced at a low temperature(Table 1).

The overpotential was mainly increased on the anode (Fig. 1, In-set), when Eap was increased from 0.6 to 0.8 V. Anodic potential in-creased from �0.221 V (Eap = 0.6 V) to �0.055 V (Eap = 0.8 V) with apotential difference of �0.166 V. However, the cathodic potentialchanged by only �0.034 V (from �0.821 V at Eap = 0.6 V to�0.855 V at Eap = 0.8 V). The energy recovery relative to the electri-cal input decreased from 152 ± 11% (Eap = 0.6 V) to 128 ± 8%(Eap = 0.8 V). These results indicated that although a large Eap couldincrease hydrogen production rate, more energy was lost due tothe overpotential produced on the anode. Combining this increasein overpotential with the nearly constant of hydrogen yield at dif-ferent Eap values (Table 1), suggested that the lower Eap was morefavorable for H2 production at a lower temperature than the higherEap.

H2 was not detected in the headspaces of any open circuit MECsat 4 �C during a batch-cycle time of 80 h. Only CO2 (0.6�1.0%) wasobserved in the headspace of the open circuit MECs, at a concentra-tion that was much less than that measured in the closed circuitMECs (6.4�9.5%). These results indicated a lack of fermentation-produced H2, and therefore that there was no significant amountof current generation in closed-circuit tests from H2 produced byfermentation.

Hydrogen yields obtained at 4 �C were similar to those observedin controls (at 25 �C), as well as at a higher temperature (30 �C)using the same MECs (Selembo et al., 2009) (Table 1). Consideringthat fermentative H2 yields cannot exceed 4 mol H2 mol�1 glucose,a MEC yield of about 6 mol H2 mol�1 glucose obtained in this studyat such low temperature is considerable (thermodynamic value is12 mol H2 mol�1 glucose). The main barrier for a further enhance-ment of H2 yields at low temperatures was a relative low cathodichydrogen recovery (61–68%). We don not know the real reasons forlow efficiency due to no hydrogen loss to methane at 4 �C. Lowcathodic hydrogen recovery could be due to a lower rate of protonreduction under psychrotolerant condition with decreased currentdensity and longer batch-cycle time. These results are differentfrom those in a previous MEC study at a higher temperature of30 �C (Selembo et al., 2009), where the main loss of hydrogenwas through methanogenesis, resulting in a lower cathodic hydro-gen recovery of 51%. In all tests at 4 �C here, methane was notdetected. For reactors operated at 25 �C, the low H2 yield was pri-marily due to low Coulombic efficiencies (59%). In these reactors,

Table 1Coulombic efficiencies, hydrogen recoveries, volumetric current densities, hydrogen production rates, hydrogen yields and energy recoveries obtained in this study (reportedaverages and standard deviations are based on three reactors at 4 �C and two reactors for controls at 25 �C over four successive cycles) and another data reported in the literatureusing glucose with the same reactor.

Temp (�C) Conc (mM) Eap (V) CE (%) rcat (%) RH2 (%) IV (A m�3) Q (m3 H2 m�3 d�1) YH2 (mol H2 mol�1 glucose) gE (%) Source

4 11.1 0.6 82 ± 13 61 ± 4 50 ± 11 38 ± 5 0.25 ± 0.03 6.0 ± 0.3 152 ± 11 This study4 11.1 0.8 74 ± 8 68 ± 4 51 ± 4 50 ± 7 0.37 ± 0.04 6.1 ± 0.5 128 ± 8 This study

25 11.1 0.6 59 ± 6 82 ± 5 48 ± 3 113 ± 4 1.01 ± 0.05 5.8 ± 0.3 241 ± 10 This study30 5.6 0.5 127 ± 23 51 ± 4 N/A 115 ± 4 0.83 ± 0.18 6.4 159 ± 12 (Selembo et al., 2009)

L. Lu et al. / Bioresource Technology 124 (2012) 68–76 71

the use of a high glucose concentration (11.1 mM) relative to thatused in previous study (5.2 mM) (Selembo et al., 2009) could leadto the loss of electrons donors through acetoclastic methanogene-sis (He et al., 2005). It was observed here in tests at 25 �C thatmethane was produced after the first cycle, reaching concentra-tions of 6% to 12% in the headspace. High organic loading rates alsowould result in increased growth of fermentative bacteria (Rabaeyet al., 2003), accumulation of an undesirable intermediates such asbutyric acid (Freguia et al., 2010) and a lower pH, all of whichwould decrease the flow of electrons from the original substrateinto current.

3.3. Intermediates produced during electrohydrogenesis at a lowtemperature

Glucose was completely consumed within 10 h and it was ob-served that intermediates (fermentation products) rapidly accu-mulated in solution after the start of the MECs at 4 �C (Fig. 2).Acetic, butyric, propionic, and valeric acids, and ethanol, were themain intermediates formed during MEC operation with glucose.Fermentation products initially reached concentration peaks at4 h (acetate and ethanol), 10 h (propionate) and 8 h (butyrateand valerate). This observation of intermediate accumulation withglucose was similar to that found in 30 �C MEC test (Selembo et al.,2009) (Eap = 0.5 V) where it observed that glucose (5.6 mM) was

Fig. 2. Intermediates formation and current production at 4 �C with glucose aselectron donor at an applied voltage of 0.6 V (based on two batch cycles).

depleted within 5 h and the production of formate, acetate and lac-tate reached peaks in concentration within 10 h, and propionatewas produced with a longer lag-time and 16 h to reach a peak.After these first peaks in concentration, most intermediates de-creased over the subsequent 4–16 h.

Several intermediates reached a second peak at 24 h (ethanol,propionate and valerate) and 33 h (butyrate), but acetic acid didnot show a second peak. In contrast to the two peaks observed herefor most intermediates, in a previous study only one peak wasfound in an MEC at 30 �C (Selembo et al., 2009). Two intermediatepeaks might indicate a mismatch in the rates of two reactions forfermentation and current generation at low temperatures. At thebeginning of the batch cycle the current production increased withthe accumulation of intermediates (first peak) and these weremore rapidly consumed, presumably as fermentation could notprovide sufficient substrates for exoelectrogens resulting in a de-creased current. Decreased electricity generation would make thenallow a re-accumulation of intermediates (second peak), with cor-responding changes in current. This relationship is supported by aclose match between current generation and the formation ofintermediates (Fig. 2), consistent with that reported in a glucose-fed MFC (Freguia et al., 2008). Current generation could be also af-fected by the accumulation of protons and organic acids during fer-mentation. Although proton transfer between electrodes is nothindered by the use of a single-chamber MEC (only an average de-crease of 0.47 ± 0.04 in pH during reactions) that lacks a membranebetween the electrodes, lower pH conditions could occur withinthe biofilm (Torres et al., 2008) at the high initial glucose concen-trations used in this study (11.1 mM or 2 g L�1). The low pH canhave an adverse effect on exoelectrogens and thus currentgeneration.

Acetic acid was the only intermediate that always decreased inconcentration and was fully removed (33 h) over the whole cycle at4 �C, suggesting that acetate was the most favorable substrate forexoelectrogens. Except for acetic acid, other fermentation productswere not completely removed by the end of batch cycle (70 h). Thereason for the lack of removal of these other intermediates is notknown, but this accumulation has been observed in other glu-cose-fed MFC studies (Lee et al., 2008).

3.4. Voltammetric characteristics of psychrotolerant biofim

CVs of mature psychrotolerant biofilm with different substrateavailabilities provides valuable information on the electron trans-fer mechanisms at low temperatures. Under conditions of turn-over (electron donor provided) and maximum biofilm activity,the voltammogram showed a typical sigmoidal shape with a singleinflection point (only one maximum in the first derivative, Fig. 3Aand inset) at a potential of �0.347 ± 0.005 V (�0.15 V vs. SHE), atwhich the rate of increase of the catalytic current reached a max-imum (Srikanth et al., 2008). This potential was relatively consis-tent for different scan rates even with the increase in peakcurrent with an increased scan rate.

Under non-turn-over conditions (no electron donor) additionaldetails on the electrochemical features of the biofilm could be ob-

Fig. 3. Cyclic voltammogram of psychrotolerant biofims. (A) The voltammogram was recorded under turn-over condition at maximum biofilm activity. The scan rate was1 mV s�1 (dash line) and 5 mV s�1 (solid line). Inset: first derivatives of the CV. (B) The voltammogram was recorded under non-turn-over condition at the end of a batch cycle(solid line) and in a fresh medium without substrate (dashed line). The scan rate was 1 mV s�1. Raw curves are shown in black, with baseline-subtracted curves in red. (Forinterpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. Rarefaction curves based on pyrosequencing of MEC communities at 4 �C(MEC-4) and 25 �C (MEC-25) with glucose as electron donor. The OTUs were definedby 5% and 3% distances.

72 L. Lu et al. / Bioresource Technology 124 (2012) 68–76

tained (Fig. 3B). At the end of a batch cycle, in the medium thatreadily degraded electron donors were depleted, the voltammo-gram revealed complex redox behavior, which consisted of two re-dox systems (Fricke et al., 2008), with formal potential atEf1 = �0.311 ± 0.006 V and Ef2 = �0.378 ± 0 V. The arithmetic meanof both potentials is �0.345 V (�0.148 V vs. SHE), was very close tothe potential of the inflection point of the catalytic curve underturn-over conditions, suggesting two redox systems were activefor anodic electron transfer. When substrate-depleted solutionwas replaced by a fresh medium without substrate, all these redoxpeaks remained present at the constant potentials. This indicatedthat these reactions were due to biofilm-bound redox compounds(e.g. outer-membrane bound cytochromes) rather than solubleelectron shuttles. The voltammograms of G. sulfurreducens biofilmsfed with acetate showed the same midpoint potential (�0.15 V vs.SHE) (Richter et al., 2009; Srikanth et al., 2008) as those observedin our psychrotolerant mixed culture biofilms. This suggests that(i) the outer membrane redox proteins (cytochromes) and the elec-tron transfer mechanisms are identical or at least very similar tothose of the exoelectrogens in our community (Richter et al.,2009), or that (ii) the microorganisms belonging to the genus Geob-acter, or that were strongly related to it, were likely the dominantexoelectrogens in the biofilm when glucose was used as the initialelectron donor. This similarity of the voltammetric behavior, how-ever, was not conclusive and therefore additional data was neededon the composition of the microbial communities.

3.5. Pyrosequencing reveals bacterial community structure inpsychrotolerant biofilm

Pyrosequencing produced 6160 and 5478 high-quality se-quences for the communities of MECs operated at 4 �C (MEC-4)and 25 �C (MEC-25), respectively. These sequences were assignedto 250 (MEC-4) and 366 (MEC-25) OTUs with a distance limit of0.03. The more conservative approach (0.05 distance) yielded 173(MEC-4) and 266 (MEC-25) OTUs. The estimators of Chao1 andACE as well as the Shannon index were used to estimate the rich-ness and diversity of a community, respectively. Although the rar-efaction curves trended to plateau after 6000 samples at the 0.05distance (Fig. 4), additional sequences would be needed to capturethe difference between the observed number of OTUs and satu-rated ones estimated by the Chao1 or ACE (Table 2). However,pyrosequencing results still showed more diverse populations thanSanger sequencing used in previous studies with glucose (Chung

and Okahe, 2009; Xing et al., 2009). These previous studies sam-pled 93–149 clones, but no more than 36 phylotypes were detectedin each sample based on a sequence similarity of >97%. MEC-25had the higher richness (higher Chao1 and ACE) and diversity(higher Shannon) than MEC-4 (Table 2). Since the Shannon diver-sity index provides not only species richness (i.e., the number ofspecies present) but also information on how the abundance ofthe species is distributed (the evenness of the species) among allthe species in the community, the distribution of OTUs in MEC-4community was concluded to be more concentrated than that inMEC-25, which could be intuitively represented in a heat mapusing a hierarchical cluster analysis (Fig. 5). The lower communitydiversity in MEC-4 could be attributed to the low temperatureenvironment, which likely reduced mesophilic or thermophilicspecies richness compared to the original inoculum.

To identify the phylogenetic diversity of MEC communities, weassigned qualified reads to known phyla, classes and genera(Fig. 6). Six and nine bacterial phyla were identified in communi-ties of MEC-4 and MEC-25, respectively, and the sum of total ob-served phyla in both MECs was ten. In general, two communitiesformed at different temperature possess an approximately similarstructure at the phylum level with the predominance of Bacteroide-tes, Proteobacteria and Firmicutes phyla (Fig. 6A). Other phyla were

Table 2Similarity-based OTUs and species richness and diversity estimates.

Sample Reads 0.03 distance 0.05 distance

OTU Chao1 ACE Shannon OTU Chao1 ACE Shannon

MEC-4 6160 250 438 726 3.43 173 307 414 3.22MEC-25 5478 366 688 922 4.11 266 409 505 3.87

Fig. 5. Hierarchical cluster analysis of MEC communities at 4 �C (MEC-4) and 25 �C(MEC-25). The y-axis is the clustering of the 300 most abundant OTUs (3% distance)in reads. The OTUs were ordered by taxonomy in level of phylum. Samplecommunities were clustered based on complete linkage method. The color intensityof scale indicates relative abundance of each OTU reads. Relative abundance wasdefined as the number of sequences affiliated with that OTU divided by the totalnumber of sequences per sample.

L. Lu et al. / Bioresource Technology 124 (2012) 68–76 73

diverse, but formed very small fractions of the total bacteria.Although the psychrotolerant community had a lower diversitythan that formed at 25 �C, more novel bacteria were found in it(4.3% unclassified bacteria in MEC-4 vs. 2.5% in MEC-25), suggest-ing that some unique species survived in low temperatures. At the

class level, two communities still held the relatively similar struc-ture, and the majority of sequences belonged to classes of Gamma-proteobacteria, Deltaproteobacteria, Bacteroidia and Clostridia(Fig. 6B) despite a total of 20 bacterial classes detected by pyrose-quencing. Four dominant classes accounted for the 88.4% and83.5% of the total reads in MEC-4 and MEC-25, respectively, andwere all affiliated with three primary phyla referred above(Fig. 6A).

The classification of sequence tags at the genus level allowed usto further examine the anode biofilms communities based on bac-terial function (Fig. 6C). The difference of two community struc-tures was generally reflected in the discrepancy of genusdistribution. Remarkably, Dysgonomonas was the most predomi-nant genus in the community of MEC-4 (36.6% of the total reads),while community in MEC-25 was mostly represented by the genusBacteroides (21.5% of the total reads). The Dysgonomonas genus isassociated with glucose fermentation with production of acid butno gas under facultative anaerobic conditions. However, this genuswas isolated from human clinical specimens, with growth obtainedbetween 25 and 37 �C (Hofstad et al., 2000). Its presence in the psy-chrotolerant biofilm could be attributed to the complex syntrophicinteractions that existed in biofilm, which are possible to acceler-ate the metabolic rate of Dysgonomonas to oxidize substrates atlow temperatures. The mixed culture biofilm may provide protec-tion for Dysgonomonas against adverse environmental conditions.The Bacteroides genus is an important mesophilic fermentativebacterium involved in sugar catabolism (Smith et al., 2006), withmajor products of glucose fermentation that include propionate,acetate, lactate, formate, succinate and fumarate. Both of Dysgono-monas and Bacteroides are generally found in carbohydrate-fedMFC communities (Chung and Okahe, 2009; Zhang et al., 2011,2009). Why the bacterial community at low temperature was dom-inated by the genus Dysgonomonas, in contrast to that formed atroom temperature, remains unknown. To date, no research hasshown that these two genera were related to the direct oxidationof glucose for current generation. Only Zhang et al. isolated an elec-trochemically active Fe(III) reducing bacterium with 99% identityto Dysgonomonas wimpennyi ANFA2 (Zhang et al., 2009).

Interestingly, other two dominant genera Paludibacter and Petri-monas that belong to the class Bacteroidia were found separately intwo communities (Fig. 6C). Paludibacter (5.7% of the total reads)was exclusively observed in community at 4 �C, all with 95–96%identities to Paludibacter propionicigenes. Petrimonas (1.5% of thetotal reads) was only found in biofilm formed at 25 �C. These twogenera are classified as mesophilic fermentative bacteria, withgrowth at 15–35 �C, and propionate and acetate as the major prod-ucts from glucose fermentation (Grabowski et al., 2005; Ueki et al.,2006). P. propionicigenes can grow at 4 �C by developing sphericalcells (Ueki et al., 2006), which may explain its abundance in psy-chrotolerant biofilms.

The genera belonged to class Gammaproteobacteria were mainlydominated by Aeromonas, Enterobacter, Raoultella and Yersinia.Raoultella was the most abundant genus accounting for 11.7%(MEC-4) and 6.1% (MEC-25) of the total reads. Raoultella share sev-eral common traits including growth at 10 �C and fermentation ofglucose to produce acid with Raoultella ornithinolytica can grow at4 �C (Sakazaki et al., 1989). Both genera Yersinia and Enterobacter

Fig. 6. Taxonomic classification of pyrosequences from MEC communities at 4 �C (MEC-4) and 25 �C (MEC-25) with glucose as electron donor at the phylum (A), class (B) andgenus (C) levels. Relative abundance is defined as the number of sequences affiliated with that phylum, class and genus divided by the total number of sequences per sample.In genus level classification the genera that are less than 1% of total composition in both libraries were classified into others.

74 L. Lu et al. / Bioresource Technology 124 (2012) 68–76

are glucose-fermentating bacteria. However, Yersinia (1.3% of thetotal reads) can be only found in the bacterial community at 4 �C,with two species identified of Y. kristensenii (100% identity) andY. frederiksenii (99% indentity). These two species of Yersinia cangrow at temperatures ranging from 4 �C to 41 �C (Bercovier et al.,1980; Ursing et al., 1980), likely explaining their exclusive pres-ence in psychrotolerant biofilms. A. hydrophila has previously been

associated with electricity generation in mixed culture anodic bio-films using glucose (Park et al., 2008). In the bacterial communityat 4 �C, Aeromonas accounted for only 0.68% of the total reads andno sequences similar to A. hydrophila were detected. In contrast, se-quences similar to A. hydrophila (1.8% of the total reads, 100% iden-tity) were found in the MEC-25 biofilm, suggesting it may have hadexoelectrogenic activity in this biofilm.

Table 3Phylogenic affiliation of bacterial 16S rRNA gene clones retrieved from two parallelMECs (MEC1 and MEC2) operated at 4 �C using glucose.

Nearest species or clones No. ofclones

Similarity

MEC1a MEC2b

BacteroidetesDysgonomonas mossii (NR_025484) 10 8 99Dysgonomonas capnocytophagoides

(AB548674)17 5 96

Dysgonomonas gadei (AB548675) 1 96Dysgonomonas sp. A1 (HQ659694) 2 1 95Paludibacter propionicigenes (CP002345) 3 1 95Parabacteroides sp. Lind7H (HQ020488) 1 1 92Uncultured Bacteroidetes bacterium

(CU925625)3 91

FirmicutesLactococcus raffinolactis (NR_044359) 3 7 99Sedimentibacter sp. MO-SED (AB598275) 1 98Acidaminococcus fermentans (CP001859) 2 1 94Acidaminococcus intestinalis (EF028685) 1 92Uncultured Megasphaera sp. (AY995255) 2 91AlphaproteobacteriaUncultured Magnetospirillum sp.

(FJ823930)2 97

BetaproteobacteriaAcidovorax defluvii (AM943035) 1 98Massilia sp. (FJ005057) 1 94GammaproteobacteriaRaoultella terrigena (HQ242728) 28 12 99Raoultella planticola (JF431273) 3 98Pseudomonas antarctica (HQ825082) 1 99Pseudomonas sp. (FJ005063) 2 96Yersinia kristensenii (FN908459) 2 1 99Yersinia aldovae (EF179125) 24 99Yersinia frederiksenii (AJ639879) 7 99Uncultured Aeromonas sp. (EF679189) 1 97Enterobacter aerogenes (FJ799902) 1 93DeltaproteobacteriaDesulfovibrio desulfuricans (FJ873799) 1 99Geobacter sp. CdA-2 (Y19190) 9 19 97SpirochaetesSphaerochaeta sp. TQ1 (DQ833400) 1 99Uncultured Spirochaeta sp. (HM041922) 5 98Total clones 95 96

OTUs were defined at a cutoff value of 0.03. GenBank accession numbers for thesequences (average length of 810–860 bp) are JN680051�JN680078.

a A total of 95 clones in 20 OTUs in MEC1.b A total of 96 clones in 18 OTUs in MEC2.

L. Lu et al. / Bioresource Technology 124 (2012) 68–76 75

Geobacter spp. was the most predominant known exoelectrogenin both bacterial communities, accounting for 12.1% (MEC-4) and16.1% (MEC-25) of the total reads. This result supports findingsbased on potentials observed in CVs. Other known exoelectrogensfound in the psychrotolerant biofilm included Desulfovibrio desulfu-ricans (Park et al., 1997) (0.24% of the total reads, 99% identity) andK. pneumoniae (Zhang et al., 2008) (0.02%, 99% identity). In the bac-terial community at 25 �C, Geobacter and A. hydrophila were mostlikely responsible for electron transfer from donors to theelectrode.

Another independent molecular technique, 16S rRNA geneclone library, was used to capture most members in the bacterialcommunity from two 4 �C MECs (Table 3). The information pro-vided by clone library and pyrosequencing were substantially inagreement, indicating that our analysis for anodic microbial com-munities was reliable.

3.6. Syntrophic interactions in glucose-fed psychrotolerant biofilm

The observed accumulation and removal of intermediates (or-ganic acids) indicated that two steps, fermentation and oxidization

of its by-products, were needed for glucose conversion. Analysis ofbacterial community on the basis of 16S rRNA gene based pyrose-quencing and clone libraries identified the existence of two abun-dant populations: psychrotolerant glucose-fermenting bacteria,and exoelectrogens (Geobacter spp.). Observation that the CV fea-ture of the anode biofilms exactly like that obtained from a pure-culture Geobacter, combined with the indicated abundance ofGeobacter spp. in the microbial communities, indicated that Geob-acter spp. were directly involved in current production at the an-ode. Therefore, there was only one main pathway throughsyntrophic interaction between fermentative bacteria and exoelec-trogens for indirect electron transfer from glucose to the anode at4 �C. No methane was detected in any of MECs at 4 �C, indicating anabsence of hydrogenotrophic and acetoclastic methanogenesis. Thepossibility that methane was taken up by bacteria in biofilms wasexcluded, as it is the end product of anaerobic reactions. Thus, syn-trophic processes that occur between fermentative bacteria andmethanogens in mesophilic cultures was absent in glucose-fed bio-films at 4 �C. No homoacetogens were found in bacterial commu-nity at 4 �C despite the use of pyrosequencing. In contrast,known homoacetogens such as Eubacterium aggregans (98% iden-tity, accounted for 0.4% of the total tags) (Mechichi et al., 1998)and Holophaga foetida (97% identity, 0.04% of the total tags) (Lie-sack et al., 1994) were observed in the MEC-25 biofilms. The ab-sence of this syntrophic interaction among fermenters,homoacetogens and exoelectrogens can be attributed to the lackof H2 production inside biofilms or inactivity of those homoaceto-gens at low temperature (the temperature for the growth of E.aggregans and H. foetida is larger than 10 �C). All abundant fermen-tative bacteria (e.g. Dysgonomonas, Paludibacter, Raoultella, Yersinia,Anaerofilum) in psychrotolerant biofilms are not known to produceH2 through fermentation of glucose.

4. Conclusions

It was demonstrated that H2 can be electrochemically producedfrom glucose via electrohydrogenesis at 4 �C at yields comparableto those obtained under mesophilic conditions. Glucose-fermenta-tion bacteria were abundant and active at a low temperature, andthrough syntrophic interactions with exoelectrogens allowed cur-rent generation from glucose. Direct glucose oxidation into currentwas not a predominant pathway. Other syntrophic processes suchas methanogenesis and homoacetogenesis were negligible at lowtemperatures. The use of MECs for H2 production from carbohy-drates is simple to achieve at low temperatures and it could there-fore contribute significantly to renewable energy production in thefuture.

Acknowledgements

This research was funded by the National Natural Science Foun-dation of China (Nos. 51178140 and 30900046), Fok Ying-TongEducation Foundation for Young Teachers in the Higher EducationInstitutions of China (No. 131076), the Science Fund for CreativeResearch Groups of the National Natural Science Foundation ofChina (No. 51121062), and award KUS-I1-003-13 from King Abdul-lah University of Science and Technology (KAUST).

References

American Public Health Association (APHA), American Water Works Association,and Water Environment Federation, 1998. Standard Methods for theExamination of Water and Wastewater, 20th ed. APHA, Washington, DC.

Bercovier, H., Ursing, J., Don Brenner, J., Steigerwalt, A., Fanning, G., Carter, G.,Mollaret, H., 1980. Yersinia kristensenii: a new species of enterobacteriaceaecomposed of sucrose-negative strains (formerly called atypical Yersiniaenterocolitica or Yersinia enterocolitica-Like). Curr. Microbiol. 4, 219–224.

76 L. Lu et al. / Bioresource Technology 124 (2012) 68–76

Cheng, S., Xing, D., Logan, B.E., 2011. Electricity generation of single-chambermicrobial fuel cells at low temperatures. Biosens. Bioelectron. 26, 1913–1917.

Chung, K., Okahe, S., 2009. Characterization of electrochemical activity of a strainIS02-3 phylogenetically related to Aeromonas sp. isolated from a glucose-fedmicrobial fuel cell. Biotechnol. Bioeng. 104, 901–910.

Freguia, S., Rabaey, K., Yuan, Z., Keller, J., 2008. Syntrophic processes drive theconversion of glucose in microbial fuel cell anodes. Environ. Sci. Technol. 42,7937–7943.

Freguia, S., Teh, E.H., Boon, N., Leung, K.M., Keller, J., Rabaey, K., 2010. Microbial fuelcells operating on mixed fatty acids. Bioresour. Technol. 101, 1233–1238.

Fricke, K., Harnisch, F., Schroder, U., 2008. On the use of cyclic voltammetry for thestudy of anodic electron transfer in microbial fuel cells. Energy Environ. Sci. 1,144–147.

Grabowski, A., Tindall, B.J., Bardin, V., Blanchet, D., Jeanthon, C., 2005. Petrimonassulfuriphila gen. nov., sp nov., a mesophilic fermentative bacterium isolatedfrom a biodegraded oil reservoir. Int. J. Syst. Evol. Microbiol. 55, 1113–1121.

He, Z., Minteer, S.D., Angenent, L.T., 2005. Electricity generation from artificialwastewater using an upflow microbial fuel cell. Environ. Sci. Technol. 39, 5262–5267.

Hofstad, T., Olsen, I., Eribe, E.R., Falsen, E., Collins, M.D., Lawson, P.A., 2000.Dysgonomonas gen. nov. to accommodate Dysgonomonas gadei sp. nov., anorganism isolated from a human gall bladder, and Dysgonomonascapnocytophagoides (formerly CDC group DF-3). Int. J. Syst. Evol. Microbiol.50, 2189–2195.

Lee, H.S., Parameswaran, P., Kato-Marcus, A., Torres, C.I., Rittmann, B.E., 2008.Evaluation of energy-conversion efficiencies in microbial fuel cells (MFCs)utilizing fermentable and non-fermentable substrates. Water Res. 42, 1501–1510.

Liesack, W., Bak, F., Kreft, J.-U., Stackebrandt, E., 1994. Holophaga foetida gen. nov.,sp. nov., a new, homoacetogenic bacterium degrading methoxylated aromaticcompounds. Arch. Microbiol. 162, 85–90.

Logan, B.E., 2009. Exoelectrogenic bacteria that power microbial fuel cells. Nat. Rev.Microbiol. 7, 375–381.

Logan, B.E., Call, D., Cheng, S., Hamelers, H.V.M., Sleutels, T.H.J.A., Jeremiasse, A.W.,Rozendal, R.A., 2008. Microbial electrolysis cells for high yield hydrogen gasproduction from organic matter. Environ. Sci. Technol. 42, 8630–8640.

Lu, L., Ren, N., Xing, D., Logan, B.E., 2009. Hydrogen production with effluent from anethanol–H2-coproducing fermentation reactor using a single-chambermicrobial electrolysis cell. Biosens. Bioelectron. 24, 3055–3060.

Lu, L., Ren, N., Zhao, X., Wang, H., Wu, D., Xing, D., 2011. Hydrogen production,methanogen inhibition and microbial community structures in psychrophilicsingle-chamber microbial electrolysis cells. Energy Environ. Sci. 4, 1329–1336.

Mechichi, T., Labat, M., Woo, T.H.S., Thomas, P., Garcia, J.L., Patel, B.K.C., 1998.Eubacterium aggregans sp. nov., a new homoacetogenic bacterium from olivemill wastewater treatment digestor. Anaerobe 4, 283–291.

Parameswaran, P., Torres, C.I., Lee, H.S., Rittmann, B.E., Krajmalnik-Brown, R., 2011.Hydrogen consumption in microbial electrochemical systems (MXCs): the roleof homo-acetogenic bacteria. Bioresour. Technol. 102, 263–271.

Park, D.H., Kim, B.H., Moore, B., Hill, H.A.O., Song, M.K., Rhee, H.W., 1997. Electrodereaction of Desulfovibrio desulfuricans modified with organic conductivecompounds. Biotechnol. Tech. 11, 145–148.

Park, H.I., Sanchez, D., Cho, S.K., Yun, M., 2008. Bacterial communities on electron-beam Pt-deposited electrodes in a mediator-less microbial fuel cell. Environ. Sci.Technol. 42, 6243–6249.

Patil, S.A., Harnisch, F., Kapadnis, B., Schroder, U., 2010. Electroactive mixed culturebiofilms in microbial bioelectrochemical systems: the role of temperature forbiofilm formation and performance. Biosens. Bioelectron. 26, 803–808.

Rabaey, K., Lissens, G., Siciliano, S.D., Verstraete, W., 2003. A microbial fuel cellcapable of converting glucose to electricity at high rate and efficiency.Biotechnol. Lett. 25, 1531–1535.

Richter, H., Nevin, K.P., Jia, H.F., Lowy, D.A., Lovley, D.R., Tender, L.M., 2009. Cyclicvoltammetry of biofilms of wild type and mutant Geobacter sulfurreducens onfuel cell anodes indicates possible roles of OmcB, OmcZ, type IV pili, and protonsin extracellular electron transfer. Energy Environ. Sci. 2, 506–516.

Sakazaki, R., Tamura, K., Kosako, Y., Yoshizaki, E., 1989. Klebsiella ornithinolytica sp.nov., formerly known as ornithine-positive Klebsiella oxytoca. Curr. Microbiol.18, 201–206.

Scherer, S., Neuhaus, K., 2006. Life at low temperatures. In: Dworkin, M., Falkow, S.,Rosenberg, E., Schleifer, K.H., Stackebrandt, E. (Eds.), The Prokaryotes, vol. 2.Springer, New York, pp. 210–262.

Selembo, P.A., Perez, J.M., Lloyd, W.A., Logan, B.E., 2009. High hydrogen productionfrom glycerol or glucose by electrohydrogenesis using microbial electrolysiscells. Int. J. Hydrogen Energy 34, 5373–5381.

Smith, C.J., Rocha, E.R., Paster, B.J., 2006. The medically important Bacteroides spp.in health and disease. In: Dworkin, M., Falkow, S., Rosenberg, E., Schleifer, K.H.,Stackebrandt, E. (Eds.), The Prokaryotes. Springer, New York, pp. 81–427.

Srikanth, S., Marsili, E., Flickinger, M.C., Bond, D.R., 2008. Electrochemicalcharacterization of Geobacter sulfurreducens cells immobilized on graphitepaper electrodes. Biotechnol. Bioeng. 99, 1065–1073.

Torres, C.I., Marcus, A.K., Rittmann, B.E., 2008. Proton transport inside the biofilmlimits electrical current generation by anode-respiring bacteria. Biotechnol.Bioeng. 100, 872–881.

Ueki, A., Akasaka, H., Suzuki, D., Ueki, K., 2006. Paludibacter propionicigenes gen. nov.,sp. nov., a novel strictly anaerobic, Gram-negative, propionate-producingbacterium isolated from plant residue in irrigated rice-field soil in Japan. Int.J. Syst. Evol. Microbiol. 56, 39–44.

Ursing, J., Don Brennert, J., Bercovier, H., Richard Fanning, G., Steigerwalt, A., Brault,J., Mollaret, H., 1980. Yersinia frederiksenii: a new species of Enterobacteriaceaecomposed of rhamnose-positive strains (formerly called atypical Yersiniaenterocolitica or Yersinia enterocolitica-Like). Curr. Microbiol. 4, 213–217.

Xing, D., Cheng, S., Regan, J.M., Logan, B.E., 2009. Change in microbial communitiesin acetate- and glucose-fed microbial fuel cells in the presence of light. Biosens.Bioelectron. 25, 105–111.

Zhang, L., Zhou, S., Zhuang, L., Li, W., Zhang, J., Lu, N., Deng, L., 2008. Microbial fuelcell based on Klebsiella pneumoniae biofilm. Electrochem. Commun. 10, 1641–1643.

Zhang, Y., Min, B., Huang, L., Angelidaki, I., 2011. Electricity generation andmicrobial community response to substrate changes in microbial fuel cell.Bioresour. Technol. 102, 1166–1173.

Zhang, Y., Min, B., Huang, L., Angelidaki, I., 2009. Generation of electricity andanalysis of microbial communities in wheat straw biomass-powered microbialfuel cells. Appl. Environ. Microbiol. 75, 3389–3395.